2545 results

How Canara HSBC Life Insurance Optimized Costs and Claims Processing with MongoDB

Since 2008, Canara HSBC Life Insurance has focused relentlessly on bringing a fresh perspective to an industry known more for stability and conservatism rather than innovation. Since its inception in 2008 as a joint venture between Canara Bank and HSBC Insurance, Canara HSBC Life Insurance has strived to differentiate itself from the competition through enhanced customer interactions, launching cutting-edge digital products, and integrating digital services that cater to the evolving needs of customers. For the past six years Chief Operating Officer, Mr. Sachin Dutta, has been on a mission to bring this customer-first mindset to the digital products and touchpoints his team creates. Speaking at MongoDB’s annual .local developer conference in Delhi, Dutta outlined Canara HSBC Life Insurance’s ongoing digital transformation journey, and how his team's focus on customer success and business efficiency led them to work with MongoDB for improved efficiencies and results. “I truly value the partnership we have with MongoDB. We are building a future-ready organization, and this partnership clearly helps us achieve our aim of reaching the last mile possible in customer servicing. Mr. Sachin Dutta, Chief Operating Officer, Canara HSBC Modernizing the architecture and driving developer efficiency Canara HSBC’s digital transformation was centered on three technical pillars: the cloud, analytics, and mobility. The company focused on creating a more integrated organization and automating manual processes within the system. “We try to remove human intervention with a life insurance policy delivered in seconds and claims that are settled virtually in seconds,” Dutta says. To get there, Canara HSBC Life Insurance had to move on from its existing architecture, which required multifaceted changes and several new implementations: Monolithic applications made alterations a time-consuming process A reliance on rigid relational databases prolonged development timelines, forcing developers to spend time wrangling data when they could be building better products for customers. The fully on-premises system had supported the organization in the past but required future-proofing to support growth and deliver a better customer experience. Because of this valuable development time and money were spent managing, patching, and scaling databases, rather than getting new products into the hands of customers. These technical issues impacted the speed of business, particularly during month-end and year-end data processing, when the volumes were high. In addition, batch processing stood in the way of creating the real-time availability of information customers wanted. Dutta and his senior team also realized that their existing infrastructure would make it more challenging to find the right talent in the market, as the existing infrastructure was increasingly becoming outdated. Dutta realized early on that, in order for Canara HSBC to attract and retain the best and brightest developers, the insurer had to offer the chance to work with the latest technologies. Platforms like MongoDB would be integral to this effort. “I want to create an organization that is attracting talent and where people start to enjoy their work, and that benefit then gets passed on to the customers, ” Mr. Dutta says. Looking to overhaul its existing infrastructure, Canara HSBC Life Insurance wanted to move fast and hire the talent required to best serve its end customers. Dutta summarized the situation succinctly: "We found that some of those relational structures that had worked for us would not take us through the next 10 years.” Migrating to a secure, fully managed database platform After evaluating the solutions on the market, the team decided to transition from their existing on-premises relational databases, like IBM DB2, MySQL, and Postgres, to MongoDB Atlas . In the last six years of my work, I’m pleased to say that MongoDB has seamlessly integrated all the processes in the backend. We migrated from a completely legacy-based setup to the new fully managed MongoDB service to enhance IT productivity Mr. Sachin Dutta The first stage of the journey was moving from monolithic applications and relational databases to a microservices architecture. With its flexible schema and capabilities for redundancy, automation, and scalability, MongoDB served as the best partner to help facilitate the transition. Next, the team moved to modernize key parts of the business, such as underwriting, freeing their data to power more automation in straight-through processing (STP) of policies and faster claims processing. The adoption of a hybrid cloud model shifted Canara HSBC Life Insurance away from on-premises databases to MongoDB Atlas. As a fully managed cloud database, MongoDB Atlas solves issues related to scalability, database management, and overall reliability. MongoDB Atlas is also cloud agnostic, giving the insurance company an option to work with Azure, AWS, and Google Cloud. Mongo Atlas’ BI Connector bridged the gap between MongoDB and traditional BI tools. This seamless integration allowed Canara HSBC Life Insurance to deploy its preferred reporting tools and, when coupled with MongoDB Atlas’ real-time analytics capability, made batch processing a thing of the past. Halving delivery times and driving business efficiencies Moving to MongoDB Atlas has had a profound impact on the breadth of digital experiences Canara HSBC Life Insurance can offer customers and the speed at which new products can be developed. Something that used to take months, with the implementation of our new tools could be completed in a couple of weeks or days Mr. Sachin Dutta And it’s not only the customer experience and product delivery that has benefited from the partnership. Canara HSBC Life Insurance has also realized substantial efficiency gains and savings as a result of working with MongoDB. We are leveraging artificial intelligence as a core capability to predict human behavior and auto-underwrite policies wherein around half of the policies issued today are issued by the system Mr. Sachin Dutta Highlighted results include: Straight-through processing (STP) surged from 37% to an impressive 60%. This is set to increase further with AI/ML integrations and rule suggestions. Policy issuance turnaround time improved by 60%. Efficiency in operations led to a 20% cost-saving per policy issuance. Canara HSBC experienced 2x top-line growth due to seamless integration with analytical tools. Looking ahead, Canara HSBC Life Insurance has already outlined three key areas where the MongoDB partnership will grow. First, Dutta wants to take advantage of MongoDB Atlas’ flexible document data model to collect and organize data on customers from across the business, making MongoDB Atlas the sole database at Canara HSBC Life Insurance and creating a true customer 360 data layer to power sophisticated data analytics. In financial services, this capability is referred to as know your customer (KYC). “We want to build a data layer that provides a unique experience to the customer after getting to know them,” he says. “That’ll help the company generate better NPS scores and retain customers.” Second, the adoption and integration of AI and machine learning tools also factor heavily into future plans. MongoDB Atlas, with its flexible schema, compatibility with various machine learning platforms, and AI-specific features — such as Vector Search and storage — is a good fit for the company. In Dutta's words, "We are going to scale up and capture the GenAI space.” Last, Dutta wants to take advantage of the MongoDB Atlas SQL interface, connectors, and drivers to augment business intelligence for reporting and precise SQL-based report conversions. Learn More about how MongoDB Works with global Insurers

December 4, 2023

MongoDB Doubles Down on Aotearoa as Part of Continued APAC Expansion

MongoDB is expanding its business in New Zealand to help Kiwi organisations build modern applications and take advantage of the AI opportunity that exists today. With hundreds of customers already in Aotearoa, including Pathfinder, Rapido, and Tourism Holdings, we're continuing to hire and invest to continue to grow our community in the country. Powering the next generation of modern applications Interest and excitement in AI, and particularly generative AI, has exploded. With a proud history of Innovation, it's not a surprise that many New Zealand companies are early adopters of this incredible technology. In fact, an AI Forum report has revealed that AI has the potential to increase New Zealand's GDP by as much as $54 billion by 2035. No matter what you think of the veracity of those bold predictions, one thing is sure: Almost every company is trying to figure out how to take advantage of data and software, to help them build better products, more efficiently and more quickly. Jake McInteer speaking at MongoDB.local Auckland As organisations transform into digital-first businesses, they’re faced with a growing list of application and data requirements. Modern applications are complex – they need to handle transactional workloads, app-driven analytics, full-text search, AI-enhanced experiences, stream data processing, and more. Companies are being asked to do this all while reducing data infrastructure sprawl, complexity and often also cut costs. What we are seeing globally is our developer data platform solves this challenge and complexity since it integrates all of the data services organisations need to build modern applications in a unified developer experience. Additionally, we also allow our customers to easily run anywhere in the world with over 110+ locations making us uniquely placed to enable Kiwi companies to adapt to a multicloud future. We also have strong local partnerships with all three cloud hyperscalers, all of which plan to open new cloud regions in New Zealand in the coming years. With the support of our cloud partners, in New Zealand we've already seen great adoption of MongoDB Atlas, including the largest established enterprises, through to cutting-edge startups. Here are a couple of examples. Pathfinder: Protecting vulnerable children Pathfinder , headquartered in Auckland, is a global leader in software development specialising in protecting vulnerable children. The company's mission centres on empowering law enforcement agencies with state-of-the-art technology, meticulously designed to combat the reprehensible crime of child exploitation. "We are committed to delivering investigators the most advanced tools. We cannot accept delays in removing a child from harm due to investigations being overwhelmed by large amounts of disparate data. In situations where every minute impacts a child's well-being, these tools must enable investigators to swiftly navigate data challenges, and rapidly apprehend perpetrators" said Bree Atkinson, CEO of Pathfinder Labs. Pathfinder’s Paradigm service is being built on MongoDB Atlas, running on AWS, and takes advantage of the wider developer data platform features in order to enable the next generation of data-driven investigative capabilities. By using MongoDB Atlas Vector Search , a native part of the MongoDB Atlas platform, the Pathfinder team are also able to match images and details within images (such as people and objects), classify documents and text, and build better search experiences for their users via semantic search. This enables Paradigm to efficiently aid law enforcement in identifying victims and apprehending offenders. Bree Atkinson, CEO of Pathfinder Labs, and Peter Pilly, DevOps Architect at Pathfinder Labs, with the MongoDB team in Auckland at the recent .local event "MongoDB Atlas allows our team to focus on our strengths: developing outstanding technology. It works with us not against us, enhancing integration which enables us to build better user experiences," said Peter Pilley, DevOps Architect at Pathfinder Labs. "Take MongoDB Atlas Vector Search, for example. Before MongoDB, we would have needed to incorporate multiple tools to achieve that functionality. Now we can handle it all from a single platform removing complexity and architecture that wasn't needed. With MongoDB Atlas, we're able to make data-driven decisions swiftly, boosting our productivity and decision-making speed." Peter's team at Pathfinder also uses MongoDB's performance advisor. They say it's like having an extra team member who suggests the best indexes for accessing their data, which is critical in an industry where getting to a specific piece of data could make all the difference. Rapido: Optimising B2B revenue and distribution Rapido has been utilising MongoDB Atlas for over five years. The team was originally part of MongoDB for Startups , a programme that offers startups free credits and technical advice to help them build faster and scale further. Their eagerness to adopt new technologies has enabled them to effectively harness MongoDB Atlas's evolving features. Working with the Accredo ERP system, Rapido has harnessed MongoDB Atlas to innovate in business-to-business (B2B) transactions. Using features like MongoDB Atlas Vector Search, the ' moreLikeThis ' operator, and MongoDB App Services, they've transformed business interactions, offering precise product recommendations and improved real-time visibility via change streams. Rapido's platform, which has processed orders collectively worth more than $100m to date, is essential for many wholesale businesses in New Zealand. Adam Holt, CEO of Rapido, summarises their experience: "Our journey with MongoDB Atlas has been transformative. By building on a cohesive developer data platform, we don't need to bolt-on and learn special technologies for every requirement. Continuously integrating new features keeps our platform advanced in the fast-paced B2B market. It's about leveraging technology to innovate and deliver better solutions to our clients." MongoDB expands in Aotearoa The increased demand from Kiwi organisations who are looking to innovate faster and take advantage of cutting-edge technologies, like AI, means MongoDB is now doubling down on its New Zealand footprint. Earlier this month, MongoDB established its local operations in Aotearoa, New Zealand. Jake McInteer , a native Kiwi, has officially transferred from MongoDB’s Australia business to lead the organisation in New Zealand. MongoDB already has a large, engaged community, more than 200 customers, and an extensive partner network. CEO of Lumin Max Ferguson presents at the Christchurch MongoDB user group We are incredibly excited about the opportunity to invest in and contribute to the Kiwi tech ecosystem, both to support local companies and help kiwi startups like Lumin and Marsello as well as established companies like Tourism Holdings , Figured , and Foster Moore . To support our growth, we have roles open on our Sales and Solutions Architecture team. If you are based in NZ and interested in joining our incredible team, working in our hybrid environment, please check out and apply for the roles here: Enterprise Account Executive, Acquisition Senior Solutions Architect Additionally, read here about the massive opportunity at MongoDB in APAC from our SVP Simon Eid.

November 30, 2023

Atlas Vector Search obtient le meilleur taux de satisfaction des développeurs dans l'enquête Retool State of AI 2023

Retool vient de publier son tout premier rapport sur l'état de l'IA , qui vaut la peine d'être lu. Inspirée du rapport State of Internal Tools, très populaire, l'enquête State of AI a pris le pouls de plus de 1 500 techniciens issus de divers secteurs d'activité : ingénieurs logiciels, dirigeants, chefs de produit, concepteurs, etc. L'objectif de l'enquête est de comprendre comment ces passionnés de technologie utilisent et construisent avec l'intelligence artificielle (IA). Dans le cadre de l'enquête, Retool a étudié les outils les plus populaires, notamment les bases de données vectorielles les plus fréquemment utilisées dans le domaine de l'IA. L'enquête a révélé que MongoDB Atlas Vector Search affichait le meilleur score de promotion (NPS) sur .NET et était la deuxième base de données vectorielles la plus utilisée, cinq mois seulement après sa mise en service. Il devance ainsi les solutions concurrentes qui existent depuis des années. Dans cet article de blog, nous examinerons l'essor phénoménal des bases de données vectorielles et la façon dont les développeurs utilisent des solutions telles qu'Atlas Vector Search pour créer des applications alimentées par l'IA. Nous aborderons également d'autres points essentiels du rapport Retool. Consultez notre page de ressources sur l'IA pour en savoir plus sur la création d'applications alimentées par l'IA avec MongoDB. Adoption d'une base de données vectorielle : Sur le site Charts (enfin presque...) De curiosité mathématique à superpuissance derrière l'IA générative et les LLM, les vector embeddings et les bases de données qui les managed ont parcouru un long chemin en très peu de temps. Consultez les tendances de DB-moteur dans les modèles de bases de données au cours des 12 derniers mois et vous verrez que les bases de données vectorielles dépassent de loin toutes les autres en termes de popularité. Il suffit de regarder la trajectoire "vers le haut et vers la droite" de la ligne rose dans le site Charts ci-dessous. Capture d'écran avec l'aimable autorisation de DB-moteur, 8 novembre 2023 Mais pourquoi les bases de données vectorielles sont-elles devenues si populaires ? Ils constituent un élément clé d'un nouveau modèle architectural appelé "génération augmentée par la recherche" ( RAG ). Il s'agit d'un mélange puissant qui combine les capacités de raisonnement de LLM préformés et polyvalents et les alimente en temps réel avec des données spécifiques à l'entreprise. Les résultats sont des applications alimentées par l'IA qui servent de manière unique l'entreprise - qu'il s'agisse de créer de nouveaux produits, de réimaginer l'expérience client ou de porter la productivité et l'efficacité internes à des niveaux sans précédent. Les encastrements vectoriels sont l'un des composants fondamentaux nécessaires pour libérer la puissance de RAG. Les modèles d'intégration vectorielle codent les données de l'entreprise, qu'il s'agisse de texte, de code, de vidéo, d'images, de flux audio ou de tableaux, sous forme de vecteurs. Ces vecteurs sont ensuite stockés, indexés et interrogés dans une base de données vectorielles ou un moteur de recherche vectoriel, fournissant les données d'entrée pertinentes en tant que contexte pour le LLM choisi. Il en résulte des applications d'IA fondées sur les données et les connaissances de l'entreprise qui sont pertinentes pour l'activité, exactes, fiables et à jour. Comme le montre l'enquête Retool, le paysage des bases de données vectorielles est encore largement vierge. Moins de 20 % des personnes interrogées utilisent des bases de données vectorielles aujourd'hui, mais avec la tendance croissante à la personnalisation des modèles et de l'infrastructure d'IA, l'adoption est assurée de croître. Pourquoi les développeurs adoptent-ils Atlas Vector Search ? L'étude de Retool sur l'état de l'IA présente quelques grandes bases de données vectorielles qui ont ouvert la voie au cours des deux dernières années, en particulier pour les applications nécessitant une recherche sémantique contextuelle. Pensez aux catalogues de produits ou à la découverte de contenu. Cependant, le défi auquel les développeurs sont confrontés lorsqu'ils utilisent ces bases de données vectorielles est qu'ils doivent les intégrer avec d'autres bases de données dans la pile technologique de leur application. Chaque couche de base de données supplémentaire dans la pile technologique de l'application ajoute encore une autre source de complexité, de latence et de frais généraux opérationnels. Cela signifie qu'ils doivent se procurer une autre base de données, l'apprendre, l'intégrer (pour le développement, les tests et la production), la sécuriser et la certifier, la répartir, la surveiller et la sauvegarder, tout en maintenant la synchronisation des données entre ces multiples systèmes. MongoDB adopte une approche différente qui évite totalement ces problèmes : Les développeurs stockent et recherchent les encastrements vectoriels natifs dans le même système que celui qu'ils utilisent comme base de données opérationnelle. Grâce à l'architecture distribuée de MongoDB, ils peuvent isoler ces différentes charges de travail tout en gardant les données entièrement synchronisées. Search Nodes fournissent un calcul dédié et une isolation de la charge de travail, ce qui est vital pour les charges de travail de recherche vectorielle à forte intensité de mémoire, permettant ainsi d'améliorer les performances et la disponibilité. Grâce au schéma de document flexible et dynamique de MongoDB, les développeurs peuvent modéliser et faire évoluer les relations entre les vecteurs, les métadonnées et les données d'application comme d'autres bases de données ne peuvent le faire. Ils peuvent traiter et filtrer les données vectorielles et opérationnelles selon les besoins de l'application grâce à une API d'interrogation expressive et à des pilotes qui prennent en charge tous les langages de programmation les plus courants. L'utilisation de la plateforme de données de développement MongoDB Atlas entièrement gérée permet aux développeurs d'obtenir le répartir, la sécurité et la performance que les utilisateurs de leurs applications attendent. Que signifie cette approche unifiée MEAN pour les développeurs ? Cycles de développement plus rapides, application plus performante offrant une latence plus faible avec des données plus fraîches, associée à une réduction des frais généraux et des coûts d'exploitation. Des résultats qui se reflètent dans le score NPS de MongoDB, le meilleur de sa catégorie. Atlas Vector Search est robuste, rentable et extrêmement rapide ! Saravana Kumar, CEO, Kovai parle du développement de l'assistant d'intelligence artificielle de son entreprise Consultez notre série Construire l'IA avec MongoDB blog (rendez-vous dans la section Prise en main pour consulter les anciens numéros). Ici, vous verrez Atlas Vector Search utilisé pour des applications alimentées par GenAI couvrant l'IA conversationnelle avec des chatbots et des voicebots, des co-pilotes, l'intelligence des menaces et la cybersécurité, la gestion des contrats, la réponse aux questions, la conformité des soins de santé et les assistants de traitement, la découverte et la monétisation de contenu, et bien d'autres choses encore. MongoDB stockait déjà des métadonnées sur les artefacts dans notre système. Avec l'introduction d'Atlas Vector Search, nous disposons désormais d'une base de données vectorielles complète qui a été testée pendant plus d'une décennie et qui répond à nos besoins en matière de recherche dense. Il n'est pas nécessaire de déployer une nouvelle base de données que nous aurions dû managed et apprendre. Nos vecteurs et les métadonnées des artefacts peuvent être stockés les uns à côté des autres. Pierce Lamb, ingénieur logiciel principal au sein de l'équipe "Données et apprentissage automatique" de VISO TRUST Que peut-on apprendre sur l'état de l'IA à partir du rapport Retool ? Au-delà de la découverte des bases de données vectorielles les plus populaires, l'enquête aborde l'IA sous différents angles. Elle commence par explorer les perceptions de l'IA par les personnes interrogées. (Il n'est pas surprenant de constater que les dirigeants sont plus optimistes que les collaborateurs individuels). Il explore ensuite les priorités d'investissement, l'impact de l'IA sur les perspectives d'emploi futures et la manière dont elle affectera probablement les développeurs et les compétences dont ils auront besoin à l'avenir. L'enquête explore ensuite le niveau d'adoption et de maturité de l'IA. Plus de 75 % des répondants à l'enquête déclarent que leur entreprise s'efforce de commencer à utiliser l'IA, mais environ la moitié d'entre eux déclarent qu'il s'agit encore de projets précoces, principalement axés sur des applications internes. L'enquête se penche ensuite sur la nature de ces applications et sur l'utilité que les personnes interrogées leur prêtent pour l'entreprise. Elle constate que presque tout le monde utilise l'IA au travail, que cela soit autorisé ou non, et identifie ensuite les principaux points problématiques. Il n'est pas surprenant que la précision des modèles, la sécurité et les hallucinations figurent en tête de liste. L'étude se termine par l'examen des principaux modèles utilisés. Encore une fois, il n'est pas surprenant que les offres d'Open AI ouvrent la voie, mais cela indique également une intention croissante d'utiliser des modèles open source avec l'infrastructure et les outils d'IA pour la personnalisation à l'avenir. Vous pouvez approfondir tous les détails de l'enquête en lisant le rapport . Prise en main avec Atlas Vector Search Vous souhaitez découvrir notre offre de recherche vectorielle ? Rendez-vous sur notre page produit Atlas Vector Search . Vous y trouverez des liens vers des tutoriels, de la documentation et des intégrations clés de l'écosystème de l'IA afin que vous puissiez vous plonger directement dans la construction de votre propre application alimentée par genAI . Si vous souhaitez en savoir plus sur les possibilités de haut niveau de la recherche vectorielle, téléchargez notre livre blanc sur l'intégration de l'IA générative .

November 30, 2023

Atlas Vector Search erzielt den höchsten Entwickler-NPS in der Retool State of AI 2023-Umfrage

Retool hat gerade seinen allerersten State of AI-Bericht veröffentlicht und es lohnt sich, ihn zu lesen. Basierend auf dem äußerst beliebten Bericht „State of Internal Tools“ erfasste die „State of AI“-Umfrage den Puls von über 1.500 Technikleuten aus den Bereichen Softwareentwicklung, Führung, Produktmanager, Designer und mehr aus verschiedenen Branchen. Der Zweck der Umfrage besteht darin, zu verstehen, wie diese Technikleute künstliche Intelligenz (KI) nutzen und damit bauen. Im Rahmen der Umfrage untersuchte Retool, welche Tools beliebt sind, einschließlich der Vektordatenbanken, die am häufigsten mit KI verwendet werden. Die Umfrage ergab, dass MongoDB Atlas Vector Search den höchsten .NET Promoter Score (NPS) erzielte und die am zweithäufigsten genutzte Vektordatenbank war – und das innerhalb von nur fünf Monaten nach ihrer Veröffentlichung. Damit liegt es vor konkurrierenden Lösungen, die es schon seit Jahren gibt. In diesem Blog untersuchen wir den phänomenalen Aufstieg von Vektordatenbanken und wie Entwickler Lösungen wie Atlas Vector Search nutzen, um KI-gestützte Anwendungen zu erstellen. Wir werden auch andere wichtige Highlights aus dem Retool-Bericht behandeln. Schauen Sie sich unsere KI-Ressourcenseite an, um mehr über die Erstellung KI-gestützter App mit MongoDB zu erfahren. Einführung von Vektordatenbanken: Aus den Charts (naja, fast ...) Von mathematischer Neugier bis zur Superkraft hinter generativer KI und LLMs haben Vektoreinbettungen und die Datenbanken, die sie managed , in sehr kurzer Zeit einen langen Weg zurückgelegt. Schauen Sie sich die DB-Engine Trends bei Datenbankmodellen in den letzten 12 Monaten an und Sie werden sehen, dass Vektordatenbanken bei der Beliebtheitsänderung alle anderen um Längen übertreffen. Schauen Sie sich einfach die Flugbahn der rosa Linie „nach oben und rechts“ in den Charts unten an. Screenshot mit freundlicher Genehmigung von DB-Engine, 8. November 2023 Aber warum sind Vektordatenbanken so beliebt geworden? Sie sind eine Schlüsselkomponente in einem neuen Architekturmuster namens Retrieval-Augmented Generation – auch bekannt als RAG – einer leistungsstarken Mischung, die die Argumentationsfähigkeiten vorab trainierter Allzweck-LLMs kombiniert und ihnen unternehmensspezifische Echtzeitdaten zuführt. Das Ergebnis sind KI-gestützte App , die dem Unternehmen auf einzigartige Weise dienen – sei es bei der Entwicklung neuer Produkte, der Neugestaltung der Kundenerfahrung oder der Steigerung der internen Produktivität und Effizienz auf beispiellose Höhen. Vektoreinbettungen sind eine der grundlegenden Komponenten, die erforderlich sind, um die Leistungsfähigkeit von RAG freizusetzen. Vektor-Einbettungsmodelle kodieren Unternehmensdaten, egal ob Text, Code, Video, Bilder, Audio- Stream oder Tabellen, als Vektoren. Diese Vektoren werden dann gespeichert, Index und in einer Vektordatenbank oder Engine abgefragt, wodurch die relevanten Eingabedaten als Kontext für das ausgewählte LLM bereitgestellt werden. Das Ergebnis sind KI- App , die auf Unternehmensdaten und -wissen basieren, die für das Unternehmen relevant, genau, vertrauenswürdig und aktuell sind. Wie die Retool-Umfrage zeigt, ist die Vektordatenbanklandschaft noch weitgehend auf der grünen Wiese. Weniger als 20 % der Befragten nutzen heute Vektordatenbanken, aber mit dem wachsenden Trend zur individuellen Anpassung von Modellen und KI-Infrastruktur wird die Akzeptanz garantiert zunehmen. Warum übernehmen Entwickler Atlas Vector Search? Die State of AI-Umfrage von Retool enthält einige großartige Vektordatenbanken, die in den letzten Jahren eine Vorreiterrolle gespielt haben, insbesondere bei Anwendungen, die eine kontextbewusste semantische Suche erfordern. Denken Sie an Produktkataloge oder Content Discovery. Die Herausforderung für Entwickler bei der Verwendung dieser Vektordatenbanken besteht jedoch darin, dass sie sie zusammen mit anderen Datenbanken in den Tech-Stack ihrer Anwendung integrieren müssen. Jede zusätzliche Datenbankschicht im Anwendungstechnologie-Stack fügt eine weitere Quelle für Komplexität, Latenz und betrieblichen Overhead hinzu. Das bedeutet, dass sie über eine weitere Datenbank verfügen, die sie beschaffen, erlernen, integrieren (für Entwicklung, Tests und Produktion), sichern und zertifizieren, skalieren, überwachen und sichern müssen – und das alles, während die Daten über diese mehreren Systeme hinweg synchron bleiben. MongoDB verfolgt einen anderen Ansatz, der diese Herausforderungen vollständig vermeidet: Entwickler speichern und durchsuchen native Vektoreinbettungen in demselben System, das sie als Betriebsdatenbank verwenden. Mithilfe der verteilten Architektur von MongoDB können sie diese verschiedenen Arbeitslasten isolieren und gleichzeitig die Daten vollständig synchronisieren. Search Nodes bieten dedizierte Rechen- und Workload- Isolation , die für speicherintensive Vektorsuch-Workloads von entscheidender Bedeutung ist, und ermöglichen so eine verbesserte Leistung und höhere Verfügbarkeit Mit dem flexiblen und dynamischen Schema von MongoDB können Entwickler Beziehungen zwischen Vektoren, Metadaten und Anwendungsdaten auf eine Weise modellieren und weiterentwickeln, die andere Datenbanken nicht können. Mit einer ausdrucksstarken Abfrage-API und Treibern, die alle gängigen Programmiersprachen unterstützen, können sie Vektor- und Betriebsdaten auf jede von der Anwendung benötigte Weise verarbeiten und filtern. Mithilfe der vollständig verwalteten Entwicklerdatenplattform MongoDB Atlas können Entwickler die Skalierbarkeit, Sicherheit und Leistung erreichen, die ihre Anwendungsbenutzer erwarten. Was bedeutet dieser MEAN Ansatz für Entwickler? Schnellere Entwicklungszyklen, eine leistungsstärkere App mit geringerer Latenz und aktuelleren Daten, gepaart mit geringerem Overhead und geringeren Kosten. Ergebnisse, die sich im erstklassigen NPS-Score von MongoDB widerspiegeln. Atlas Vector Search ist robust, kostengünstig und unglaublich schnell! Saravana Kumar, CEO, Kovai spricht über die Entwicklung des KI-Assistenten seines Unternehmens Schauen Sie sich unsere Blog Reihe „ Building AI with MongoDB “ an (gehen Sie zum Abschnitt „Erste Schritte“, um frühere Ausgaben zu sehen). Hier sehen Sie, wie Atlas Vector Search für GenAI-basierte Anwendungen verwendet wird, die Konversations-KI mit Chatbots und Voicebots, Co-Piloten, Bedrohungsinformationen und Cybersicherheit, Vertragsmanagement, Fragebeantwortung, Compliance- und Behandlungsassistenten im Gesundheitswesen, Inhaltserkennung und Monetarisierung usw. umfassen mehr. MongoDB speicherte bereits Metadaten zu Artefakten in unserem System. Mit der Einführung von Atlas Vector Search verfügen wir nun über eine umfassende Vektor-Metadaten-Datenbank, die sich über ein Jahrzehnt im Kampf bewährt hat und unsere Anforderungen an die Suche nach dichten Daten erfüllt. Es ist nicht erforderlich, eine neue Datenbank bereitzustellen, die wir managed und erlernen müssten. Unsere Vektoren und Artefaktmetadaten können direkt nebeneinander gespeichert werden. Pierce Lamb, Senior Software Engineer im Team für Daten und maschinelles Lernen bei VISO TRUST Was können Sie aus dem Retool-Bericht über den Stand der KI lernen? Die Umfrage deckt nicht nur die beliebtesten Vektordatenbanken auf, sondern befasst sich auch mit KI aus verschiedenen Perspektiven. Zunächst wird untersucht, wie die Befragten KI wahrnehmen. (Es überrascht nicht, dass die C-Suite optimistischer ist als einzelne Mitwirkende.) Anschließend werden Investitionsprioritäten, die Auswirkungen von KI auf zukünftige Berufsaussichten und die voraussichtlichen Auswirkungen auf Entwickler und die von ihnen in Zukunft benötigten Fähigkeiten untersucht. Anschließend untersucht die Umfrage den Grad der KI-Einführung und -Reife. Über 75 % der Umfrageteilnehmer geben an, dass ihre Unternehmen Anstrengungen unternehmen, um mit KI zu beginnen. Etwa die Hälfte gab an, dass es sich hierbei noch um frühe Projekte handelte, die hauptsächlich auf interne Anwendungen ausgerichtet waren. In der Umfrage wird anschließend untersucht, um welche Anwendungen es sich handelt und welchen Nutzen die Befragten für das Unternehmen davon halten. Es stellt fest, dass fast jeder KI bei der Arbeit einsetzt, unabhängig davon, ob es ihm erlaubt ist oder nicht, und identifiziert dann die größten Schwachstellen. Es ist keine Überraschung, dass Modellgenauigkeit, Sicherheit und Halluzinationen ganz oben auf dieser Liste stehen. Den Abschluss der Umfrage bildet die Untersuchung der im Einsatz befindlichen Top-Modelle. Auch hier ist es keine Überraschung, dass die Angebote von Open AI wegweisend sind, aber es deutet auch auf eine wachsende Absicht hin, in Zukunft Open Source Modelle zusammen mit KI-Infrastruktur und Tools für individuelle Anpassungen zu nutzen. Sie können sich alle Einzelheiten der Umfrage ansehen, indem Sie den Bericht lesen . Erste Schritte mit Atlas Vector Search Möchten Sie einen Blick auf unser Vektorsuchangebot werfen? Besuchen Sie unsere Atlas Vector Search-Produktseite . Dort finden Sie Links zu Tutorials, Dokumentationen und wichtigen AI Umgebung-Integrationen, sodass Sie direkt mit der Entwicklung Ihrer eigenen GenAI-basierten App beginnen können . Wenn Sie mehr über die umfassenden Möglichkeiten der Vektorsuche erfahren möchten, laden Sie unser Whitepaper zum Einbetten generativer KI herunter .

November 30, 2023

Atlas Vector Search comanda o NPS de desenvolvedor mais alto na pesquisa Retool State of AI 2023

A Retool acaba de publicar seu primeiro relatório sobre o estado da IA e vale a pena lê-lo. Com base no relatório extremamente popular sobre o estado das ferramentas internas, a pesquisa sobre o estado da IA mediu o pulso de mais de 1.500 profissionais de tecnologia, abrangendo engenharia de software, liderança, gerentes de produto, designers e muito mais, provenientes de uma variedade de setores. O objetivo da pesquisa é entender como esse pessoal da tecnologia usa e constrói com inteligência artificial (IA). Como parte da pesquisa, a Retool investigou quais ferramentas eram populares, incluindo os bancos de dados vetoriais usados com mais frequência com IA. A pesquisa descobriu que o MongoDB Atlas Vector Search comandou o .NET Promoter Score (NPS) mais alto e foi o segundo banco de dados de vetores mais usado - apenas cinco meses após seu lançamento. Isso o coloca à frente de soluções concorrentes que já existem há anos. Nesta postagem do blog, examinaremos o aumento fenomenal dos bancos de dados vetoriais e como os desenvolvedores estão usando soluções como o Atlas Vector Search para criar aplicativos com tecnologia de IA. Também abordaremos outros destaques importantes do relatório Retool. Confira nossa página de recursos de IA para saber mais sobre como criar aplicativos baseados em IA com MongoDB. Adoção de banco de dados vetorial: fora do Charts (bem, quase...) Da curiosidade matemática ao superpoder por trás da IA generativa e dos LLMs, os embeddings de vetores e os bancos de dados que os managed percorreram um longo caminho em muito pouco tempo. Confira as tendências dos DB-Engines em modelos de banco de dados nos últimos 12 meses e você verá que os bancos de dados vetoriais estão muito acima de todos os outros em termos de mudança de popularidade. Basta observar a trajetória “para cima e para a direita” da linha rosa nos Charts abaixo. Captura de tela cortesia de DB-engines, 8 de novembro de 2023 Mas por que os bancos de dados vetoriais se tornaram tão populares? Eles são um componente-chave em um novo padrão de arquitetura chamado geração aumentada de recuperação – também conhecido como RAG – uma combinação potente que combina os recursos de raciocínio de LLMs de uso geral pré-treinados e os alimenta com dados específicos da empresa em tempo real. Os resultados são aplicativos baseados em IA que atendem exclusivamente aos negócios, seja criando novos produtos, reimaginando as experiências do cliente ou elevando a produtividade e a eficiência internas a níveis sem precedentes. Os embeddings vetoriais são um dos componentes fundamentais necessários para desbloquear o poder do RAG. Os modelos de incorporação de vetores codificam dados corporativos, sejam eles texto, código, vídeo, imagens, fluxos de áudio ou tabelas, como vetores. Esses vetores são então armazenados, indexados e consultados em um banco de dados de vetores ou mecanismo de busca de vetores, fornecendo os dados de entrada relevantes como contexto para o LLM escolhido. O resultado são aplicativos de IA baseados em dados e conhecimentos corporativos relevantes para os negócios, precisos, confiáveis e atualizados. Como mostra a pesquisa Retool, o cenário dos bancos de dados vetoriais ainda é em grande parte inexplorado. Atualmente, menos de 20% dos entrevistados usam bancos de dados vetoriais, mas com a tendência crescente de personalização de modelos e infraestrutura de IA, a adoção certamente crescerá. Por que os desenvolvedores estão adotando o Atlas Vector Search? A pesquisa State of AI da Retool apresenta alguns excelentes bancos de dados vetoriais que abriram caminho nos últimos dois anos, especialmente em aplicações que exigem pesquisa semântica sensível ao contexto. Pense em catálogos de produtos ou descoberta de conteúdo. No entanto, o desafio que os desenvolvedores enfrentam ao usar esses bancos de dados vetoriais é que eles precisam integrá-los a outros bancos de dados na pilha de tecnologia de seus aplicativos. Cada camada adicional de banco de dados na pilha de tecnologia de aplicativos adiciona outra fonte de complexidade, latência e sobrecarga operacional. Isso significa que eles têm outro banco de dados para adquirir, aprender, integrar (para desenvolvimento, teste e produção), proteger e certificar, dimensionar, monitorar e fazer backup, e tudo isso enquanto mantêm os dados sincronizados entre esses vários sistemas. O MongoDB adota uma abordagem diferente que evita totalmente esses desafios: Os desenvolvedores armazenam e pesquisam incorporações de vetores nativos no mesmo sistema que usam como banco de dados operacional. Usando a arquitetura distribuída do MongoDB, eles podem isolar essas diferentes cargas de trabalho enquanto mantêm os dados totalmente sincronizados. Os nós de pesquisa fornecem computação dedicada e isolamento de carga de trabalho que é vital para cargas de trabalho de pesquisa vetorial com uso intensivo de memória, permitindo assim melhor desempenho e maior disponibilidade Com o esquema de documento flexível e dinâmico do MongoDB, os desenvolvedores podem modelar e desenvolver relacionamentos entre vetores, metadados e dados de aplicativos de uma forma que outros bancos de dados não conseguem. Eles podem processar e filtrar dados vetoriais e operacionais da maneira que o aplicativo precisar, com uma API de consulta expressiva e drivers que suportam todas as linguagens de programação mais populares. O uso da plataforma de dados de desenvolvedor MongoDB Atlas totalmente managed permite que os desenvolvedores alcancem a escala, a segurança e o desempenho que os usuários de seus aplicativos esperam. O que essa abordagem unificada MEAN para os desenvolvedores? Ciclos de desenvolvimento mais rápidos, aplicativos de maior desempenho, proporcionando menor latência com dados mais atualizados, juntamente com menores custos e despesas operacionais. Resultados que são refletidos na melhor pontuação NPS do MongoDB. Atlas Vector Search é robusto, econômico e extremamente rápido! Saravana Kumar, CEO, Kovai discutindo o desenvolvimento do assistente de IA de sua empresa Confira nossa série de blog Construindo IA com MongoDB (vá para a seção Primeiros passos para ver as edições anteriores). Aqui você verá o Atlas Vector Search usado para aplicativos baseados em GenAI, abrangendo IA de conversação com chatbots e voicebots, copilotos, inteligência de ameaças e segurança cibernética, gerenciamento de contratos, resposta a perguntas, conformidade de saúde e assistentes de tratamento, descoberta e monetização de conteúdo, e mais. O MongoDB já armazenava metadados sobre artefatos em nosso sistema. Com a introdução do Atlas Vector Search, agora temos um banco de dados abrangente de metadados vetoriais que foi testado em batalha ao longo de uma década e que resolve nossas densas necessidades de recuperação. Não há necessidade de implantar um novo banco de dados que teríamos que managed e aprender. Nossos vetores e metadados de artefatos podem ser armazenados um ao lado do outro. Pierce Lamb, engenheiro de software sênior da equipe de dados e aprendizado de máquina da VISO TRUST O que você pode aprender sobre o estado da IA no relatório Retool? Além de revelar os bancos de dados de vetores mais populares, a pesquisa abrange a IA sob diversas perspectivas. Começa explorando as percepções dos entrevistados sobre IA. (Sem surpresa, o C-suite é mais otimista do que os colaboradores individuais.) Em seguida, explora as prioridades de investimento, o impacto da IA nas futuras perspectivas de emprego e como provavelmente afectará os programadores e as competências de que necessitarão no futuro. A pesquisa explora então o nível de adoção e maturidade da IA. Mais de 75% dos entrevistados afirmam que as suas empresas estão a envidar esforços para começar a utilizar a IA, com cerca de metade a dizer que estes ainda são projetos iniciais e orientados principalmente para aplicações internas. A pesquisa prossegue examinando quais são esses aplicativos e quão úteis os entrevistados consideram que eles são para os negócios. Ele descobre que quase todo mundo usa IA no trabalho, quer tenham permissão ou não, e então identifica os principais pontos problemáticos. Não é nenhuma surpresa que a precisão, a segurança e as alucinações do modelo estejam no topo dessa lista. A pesquisa termina explorando os principais modelos em uso. Mais uma vez, não é surpresa que as ofertas de IA aberta estejam liderando o caminho, mas também indica uma intenção crescente de usar modelos de código aberto juntamente com infraestrutura e ferramentas de IA para personalização no futuro. Você pode se aprofundar em todos os detalhes da pesquisa lendo o relatório . Introdução ao Atlas Vector Search Quer dar uma olhada em nossa oferta de pesquisa de vetores? Acesse nossa página de produto Atlas Vector Search . Lá você encontrará links para tutoriais, documentação e integrações importantes do ecossistema de IA para que possa mergulhar diretamente na construção de seus próprios aplicativos com tecnologia genAI . Se você quiser saber mais sobre as possibilidades de alto nível da pesquisa vetorial, baixe nosso whitepaper Incorporação de IA generativa .

November 30, 2023

Atlas Vector Search obtiene el NPS más alto para desarrolladores en la encuesta Retool State of AI 2023

Retool acaba de publicar su primer informe sobre el estado de la IA y merece la pena leerlo. Inspirada en su popular informe State of Internal Tools, la encuesta State of AI tomó el pulso de más de 1,500 técnicos que abarcan ingeniería de software, liderazgo, gerentes de producto, diseñadores y más provenientes de una variedad de industrias. El propósito de la encuesta es entender cómo esta gente de tecnología usa y construye con inteligencia artificial (IA). Como parte de la encuesta, Retool averiguó qué herramientas eran populares, incluidas las bases de datos vectoriales que se utilizan con mayor frecuencia con IA. La encuesta encontró que MongoDB Atlas Vector Search tenía el Net Promoter Score (NPS) más alto y era la segunda base de datos de vectores más utilizada, tan solo cinco meses después de su versión. Esto lo coloca por delante de las soluciones de la competencia que han existido durante años. En esta publicación de blog, examinaremos el fenomenal aumento de las bases de datos vectoriales y cómo los desarrolladores están utilizando soluciones como Atlas Vector Search para crear aplicaciones impulsadas por IA. También cubriremos otros aspectos destacados clave del informe Retool. Consulte nuestra página de recursos de IA para obtener más información sobre cómo crear aplicaciones basadas en IA con MongoDB. Adopción de bases de datos vectoriales: fuera de Charts (bueno, casi...) Desde la curiosidad matemática hasta la superpotencia detrás de la IA generativa y los LLM, las incorporaciones de vectores y las bases de datos que las managed han recorrido un largo camino en muy poco tiempo. Consulte las tendencias de DB-Engines en modelos de bases de datos durante los últimos 12 meses y verá que las bases de datos vectoriales están muy por encima de todas las demás en cuanto a cambios de popularidad. Basta con mirar la trayectoria "hacia arriba y hacia la derecha" de la línea rosa en los gráficos a continuación. Captura de pantalla cortesía de DB-engines, 8 de noviembre de 2023 Pero, ¿por qué las bases de datos vectoriales se han vuelto tan populares? Son un componente clave en un nuevo patrón arquitectónico llamado generación aumentada de recuperación, también conocida como RAG , una mezcla potente que combina las capacidades de razonamiento de LLM pre-entrenadas y de propósito general y les proporciona datos específicos de la compañía en tiempo real. Los resultados son una aplicación impulsada por IA que sirven de manera única al negocio, ya sea creando nuevos productos, reinventando la experiencia del cliente o impulsando la productividad y eficiencia internas a niveles sin precedentes. Las incrustaciones vectoriales son uno de los componentes fundamentales necesarios para liberar el poder de RAG. Los modelos de incrustación de vectores codifican datos de la empresa, ya sean texto, código, vídeo, imágenes, transmisión de audio o tablas, como vectores. Luego, esos vectores se almacenan, clasifican y consultan en una base de datos de vectores o en un motor de búsqueda de vectores, proporcionando los datos de entrada relevantes como contexto para el LLM elegido. El resultado es una aplicación de IA (en este contexto móvil); también válido: aplicación basada en datos y conocimientos de la empresa que sean relevantes para el negocio, precisos, confiables y actualizados. Como muestra la encuesta de Retool, el panorama de la base de datos vectorial sigue siendo en gran medida un campo nuevo. Menos del 20% de los encuestados utilizan bases de datos vectoriales hoy en día, pero con la tendencia creciente hacia la personalización de modelos e infraestructura de IA, se garantiza que la adopción crecerá. ¿Por qué los desarrolladores están adoptando Atlas Vector Search? La encuesta sobre el estado de la IA de Retool presenta algunas excelentes bases de datos vectoriales que han abierto un camino en los últimos años, especialmente en aplicaciones que requieren búsqueda semántica consciente del contexto. Piense en catálogos de productos o descubrimiento de contenido. Sin embargo, el desafío que enfrentan los desarrolladores al utilizar esas bases de datos vectoriales es que tienen que integrarlas junto con otras bases de datos en la pila tecnológica de su aplicación. Cada capa de base de datos adicional en la pila tecnológica de la aplicación agrega otra fuente de complejidad, latencia y gastos operativos generales. Esto significa que tienen otra base de datos para adquirir, aprender, integrar (para desarrollo, pruebas y producción), proteger y certificar, escalar, monitorear y realizar copias de seguridad, y todo esto mientras mantienen los datos sincronizados en estos múltiples sistemas. MongoDB adopta un enfoque diferente que evita estos desafíos por completo: Los desarrolladores almacenan y buscan incrustaciones de vectores nativos en el mismo sistema que utilizan como base de datos operativa. Al utilizar la arquitectura distributed de MongoDB, pueden aislar estas diferentes cargas de trabajo mientras mantienen los datos completamente sincronizados. Search Nodes proporcionan computación dedicada y aislamiento de cargas de trabajo que es vital para cargas de trabajo de búsqueda vectorial con uso intensivo de memoria, lo que permite un mejor rendimiento y una mayor disponibilidad. Con el esquema de documento dinámico y flexible de MongoDB, los desarrolladores pueden modelar y desarrollar relaciones entre vectores, metadatos y datos de aplicaciones de maneras que otras bases de datos no pueden. Pueden procesar y filtrar datos vectoriales y operativos de cualquier forma que la aplicación necesite con una API de consulta expresiva y controladores que brindan asistencia técnica a todos los lenguajes de programación más populares. El uso de la plataforma de datos para desarrolladores MongoDB Atlas , totalmente gestionada, permite a los desarrolladores lograr la escalabilidad, la seguridad y el rendimiento que esperan los usuarios de sus aplicaciones. ¿Qué significa este enfoque unificado para los desarrolladores? Ciclos de desarrollo más rápidos, aplicaciones de mayor rendimiento que proporcionan menor latencia con datos más actualizados, junto con menores costos y gastos operativos generales. Resultados que se reflejan en la mejor puntuación NPS de su clase de MongoDB. ¡Atlas Vector Search es sólido, rentable e increíblemente rápido! Saravana Kumar, CEO, Kovai , habla sobre el desarrollo del asistente de inteligencia artificial de su empresa Consulte nuestra serie blog Construyendo IA con MongoDB (diríjase a la sección Introducción para ver los números anteriores). Aquí verá Atlas Vector Search utilizado para aplicaciones impulsadas por GenAI que abarcan IA conversacional con chatbots y voicebots, copilotos, inteligencia de amenazas y ciberseguridad, gestión de contratos, respuesta a preguntas, asistentes de tratamiento y cumplimiento de atención médica, descubrimiento de contenido y monetización, y más. MongoDB ya estaba almacenando metadatos sobre artefactos en nuestro sistema. Con la introducción de Atlas Vector Search, ahora contamos con una base de datos integral de metadatos vectoriales que ha sido probada durante más de una década y que resuelve nuestras densas necesidades de recuperación. No es necesario implementar una nueva base de datos, tendríamos que administrar y aprender. Nuestros vectores y metadatos de artefactos se pueden almacenar uno al lado del otro. Pierce Lamb, ingeniero sénior de software del equipo de datos y aprendizaje automático de VISO TRUST ¿Qué puede aprender sobre el estado de la IA en el informe de Retool? Más allá de descubrir las bases de datos vectoriales más populares, la encuesta cubre la IA desde una variedad de perspectivas. Comienza explorando las percepciones de los encuestados sobre la IA. (Como era de esperar, el C-suite es más alcista que los contribuyentes individuales). Luego explora las prioridades de inversión, el impacto de la IA en las perspectivas laborales futuras y cómo afectará probablemente a los desarrolladores y las habilidades que necesitan en el futuro. Luego, la encuesta explora el nivel de adopción y madurez de la IA. Más del 75% de los encuestados dicen que sus empresas están haciendo esfuerzos para comenzar con la IA, y alrededor de la mitad dice que estos todavía eran proyectos iniciales, y principalmente orientados a aplicaciones internas. La encuesta continúa para examinar cuáles son esas aplicaciones y qué tan útiles piensan los encuestados que son para el negocio. Encuentra que casi todo el mundo usa IA en el trabajo, ya sea que se les permita o no, y luego identifica los principales puntos débiles. No es de extrañar que la precisión del modelo, la seguridad y las alucinaciones encabecen esa lista. La encuesta concluye explorando los mejores modelos en uso. Una vez más, no sorprende que las ofertas de Open AI estén liderando el camino, pero también indica una creciente intención de utilizar modelos de código abierto junto con infraestructura de AI y herramientas para personalización en el futuro. Puedes profundizar en todos los detalles de la encuesta leyendo el informe . Introducción con Atlas Vector Search ¿Deseas echar un vistazo a nuestra oferta de búsqueda vectorial? Dirígete a nuestra página de productos Atlas Vector Search . Allí encontrarás enlaces a tutoriales, documentación e integraciones clave del ecosistema de IA para que puedas sumergirte directamente en la creación de tu propia aplicación impulsada por GenAI . Si desea obtener más información sobre las posibilidades de alto nivel de la búsqueda vectorial, descargue nuestro documento técnico de " Embedding Generative AI ."

November 30, 2023

How Atlas Edge Server Bridges the Gap Between Connected Retail Stores and the Cloud

Efficient operations and personalized customer experiences are essential for the success of retail businesses. In today's competitive retail industry, retailers need to streamline their operations, optimize inventory management, and personalize the customer experience to stay ahead. In a recent announcement at MongoDB .local London, we unveiled the private preview of MongoDB Atlas Edge Server , offering a powerful platform that empowers retailers to achieve their goals, even when low or intermittent connectivity issues may arise. What is edge computing, and why is it so relevant for retail? The retail industry's growing investment in edge computing, projected to reach $208 billion by 2023, confirms the strategic shift retailers are willing to take to reach new markets and enhance their offers. And for good reason — in scenarios where connectivity is unreliable, edge computing allows operations to continue uninterrupted. Edge computing is a strategic technology approach that brings computational power closer to where data is generated and processed, such as in physical retail stores or warehouses. Instead of relying solely on centralized data centers, edge computing deploys distributed computing resources at the edge of the network. The evolution of investments in edge computing reflects a journey from initial hesitation to accelerated growth. As edge computing continues to mature and demonstrate its value, retailers are likely to further embrace and expand their focus in bringing applications where the computing and data is as close as possible to the location where it's being used. Let’s dig into how MongoDB addresses the current challenges any retailer would experience when deploying or enhancing in-store servers using edge computing. Connected store: How MongoDB's versatile deployment from edge to cloud powers critical retail applications. Currently, many retail stores operate with an on-site server in place acting as the backbone for several critical applications within the store ecosystem. Having an on-site server means that the data doesn't have to travel over long distances to be processed, which can significantly reduce latency. This setup can often also be more reliable, as it doesn't depend on internet connectivity. If the internet goes down, the store can continue to operate since the essential services are running on the local network. This is crucial for applications that require real-time access to data, such as point-of-sale (POS) systems, inventory management, and workforce-enablement apps for customer service. The need for sync: Seamless edge-to-cloud integration The main driver for retailers taking a hybrid approach is that they want to experience the low latency and reliability of an on-site server coupled with the scalability and power of cloud computing for their overall IT stack. The on-site server ensures that the devices and systems that are critical to sales floor operations — RFID tags and readers for stock management, mobile scanners for associates, and POS systems for efficient checkout — remain functional even with intermittent network connectivity. This data must be synced to the retailer’s cloud-based application stack so that they have a view of what’s happening across the stores. Traditionally this was done with an end-of-day batch job or nightly upload. The aim for the next generation of these architectures is to give real-time access to the same data set, seamlessly reflecting changes made server-side or in the cloud. This needs to be achieved without a lag from the store being pushed to the cloud and without creating complex data sync or conflict resolution that needs to be built and maintained. These complexities may cause discrepancies between the online and offline capabilities of the store's operations. It makes sense that for any retailer wanting to benefit from both edge and cloud computing, it must simplify its architecture and focus on delivering value-added features to delight the customer and differentiate from their competitors. Low-latency edge computing with Atlas Edge Server and its different components to achieve data consistency and accuracy across layers This is when Atlas Edge Server steps in to bridge the gap. Edge Server runs on-premises and handles sync between local devices and bi-directional sync between the edge server and Atlas. It not only provides a rapid and reliable in-store connection but also introduces a tiered synchronization mechanism, ensuring that data is efficiently synced with the cloud. These devices are interconnected through synchronized data layers from on-premises systems to the cloud, simplifying the creation of mobile apps thanks to Atlas Device SDK , which supports multiple programming languages, development frameworks, and cloud providers. Additionally, Atlas Device Sync automatically handles conflicts, eliminating the need to write complex conflict-resolution code. In the below diagram, you can see how the current architecture for a connected store with devices using Atlas Device SDK and Atlas Device Sync would work. This is an ideal solution for devices to sync to the Atlas backend. A high-level overview of the Architecture for connected devices in a retail space with MongoDB Device Sync and MongoDB Atlas when connectivity is unreliable. In a store with Atlas Edge Server, the devices sync to Atlas on-premises. All changes made on the edge or on the main application database are synced bidirectionally. If the store server goes offline or loses connectivity, the devices can still access the database and update it locally. The store can still run its operations normally. Then, when it comes back online, the changes on both sides (edge and cloud) are resolved, with conflict resolution built into the sync server. A high-level overview of the architecture for connected devices in a retail space with MongoDB Device Sync and MongoDB Atlas solving connectivity issues by implementing an on-premises Atlas Edge Server. Deploying Atlas Edge Server in-store turns connected stores into dynamic, customer-centric hubs of innovation. This transformation produces advantageous business outcomes including: Enhanced inventory management — The hybrid model facilitates real-time monitoring of logistics, enabling retailers to meticulously track stock in store as shipments come in and sales or orders are processed. By processing data locally and syncing with the cloud, retailers gain immediate insights, allowing for more precise inventory control and timely restocking. Seamless operational workflows — The reliability of edge computing ensures essential sales tools — like RFID systems, handheld scanners, workforce apps, and POS terminals — remain operational even during connectivity hiccups. Meanwhile, the cloud component helps ensure that all data is backed up and accessible for analysis, leading to more streamlined store operations. Customized shopping experiences — With the ability to analyze data on-the-spot (at the edge) and harness historical data from the cloud, retailers can create highly personalized shopping experiences. This approach enables real-time, tailored product recommendations and promotions, enhancing customer engagement and satisfaction. Conclusion With Atlas Edge Server, MongoDB is committed to meeting the precise needs of modern retail stores and their diverse use cases. Lacking the seamless synchronization of data between edge devices and the cloud, delivering offline functionality that enables modern, next-generation workforce applications, as well as in-store technologies like POS systems, is daunting. Retailers need ready-made solutions so they don't have to deal with the complexities of in-house, custom development. This approach allows them to channel their development efforts towards value-added, differentiating features that directly benefit their customers by improving their in-store operations. With this approach, we aim to empower retailers to deliver exceptional customer experiences and thrive in the ever-evolving retail landscape. Ready to revolutionize your retail operations with cutting-edge technology? Discover how MongoDB's Atlas Edge Server can transform your store into a dynamic, customer-centric hub. Don't let connectivity issues hold you back. Embrace the future of retail with Atlas Edge Server!

November 30, 2023

India: A Cornerstone of Growth for MongoDB Technical Services

India has emerged as a cornerstone in our MongoDB Technical Services growth story, marked by the team’s 100% growth in just two years. Bengaluru has been at the forefront of our expansion, witnessing an incredible increase in personnel and the addition of new teams and functions to support our developer data platform. This highlights Bengaluru’s emerging role in providing critical technical assistance to our customers and partners. Gurugram has also played a crucial role in the growth of Technical Services in India. This growth underscores Gurugram’s increasing significance as a thriving hub for MongoDB Technical Services. MongoDB’s continuing investment in expanding Technical Services in India reflects the substantial impact the team has had on MongoDB's customer success. APAC Technical Services team members A look into Technical Services We have multiple customer-facing Technical Services teams in India, each with unique roles and responsibilities. From specific product support to support for MongoDB services and Atlas cluster deployments, each team seeks individuals with strong critical thinking skills who can quickly detect, resolve, or escalate complex issues that may span various aspects of MongoDB's products and services. Our Technical Services teams are committed to delivering exceptional support to our customers through each team’s unique focus. Building together across teams and departments As part of their role, Technical Services Engineers (TSEs) need to partner with other supporting functions within MongoDB to ensure seamless operations and exceptional customer support. TSEs proactively identify issues that may require escalation and involve Escalation Managers accordingly. They also identify opportunities for improvement within the MongoDB product ecosystem and pass on any feedback, feature requests, and customer insights to our Product Management and Customer Success teams. India Technical Services team members The Technical Services team is highly collaborative and works together to solve customer problems. While each sub-team within Technical Services focuses on specific areas of expertise, there are numerous intersections that require cross-team collaboration. This collaborative approach enables team members to learn and build new skills by exploring different areas of interest. Excellence centers We also take pride in the technical excellence centers that we have built between our teams in India. These centers are part of our global Technical Experts ecosystem. Technical Experts help us train the Technical Services team, liaise with the development teams, and highlight pain points to the Product Management team, amongst other responsibilities. Learning and development We believe that an engineer's journey goes beyond a career and is about gaining knowledge, skills, and experiences. MongoDB is an integral part of this journey, offering a platform for continuous growth and development. New hires undergo comprehensive onboarding and training to gain a holistic understanding of the MongoDB ecosystem. Engineers are encouraged to participate in cross-team rotations and have access to various learning platforms, including O’Reilly Learning and internal Product Readiness training. We collaborate closely with our Technical Services Knowledge and Training teams to deliver training sessions, fostering both technical and presentation skills and an observational learning culture. Engineers are also provided with “protected time” to focus on their individual learning plan, allowing them to focus on delivering projects, prepping for certifications, growing their expertise in a specific area, or working in collaborative cross-skill focus groups. Periodic hackathons offer a platform for innovation, encourage collaboration, and contribute to team-wide problem-solving efforts. In addition, MongoDB User Groups and .local events allow the team to showcase and share their knowledge externally. A TSE speaking at MongoDB.local Mumbai By creating a learning environment where engineers can grow and achieve their goals, both our team members and the business thrive. Enhancing support across time zones An exciting development in our Technical Services journey is the introduction of the Swing Shift in India. This strategic initiative leverages India's engineering talent to enhance support during India and EMEA hours, further augmenting support for the surrounding regions and improving service continuity for APAC and EMEA customers. We continue to hire in both our Bengaluru and Gurugram offices. If you’re someone who is technically talented, enjoys problem-solving, loves working with customers, and values your personal and professional growth, I encourage you to explore open roles on our careers site .

November 29, 2023

MongoDB助力腾讯游戏 优化游戏开发体验

客户简介 腾讯游戏提倡超级数字场景 连接数亿游戏玩家 作为“超级数字场景”理念的倡导者和实践者,腾讯游戏致力于为用户创造高品质数字生活体验,为产业和社会发展创造更多建设性的价值。 腾讯游戏为全球知名的游戏开发与服务运营商,在全球连接超过8亿的用户。在开放发展的模式下,腾讯游戏采取自主研发和多元化的外部合作相结合的方式,在网络游戏众多细分市场领域形成专业化布局,打造覆盖全品类的产品阵营,为全球网络游戏玩家提供休闲游戏平台、大型网游、中型休闲游戏、桌面游戏、对战平台五大类。另外,腾讯游戏与全球顶级游戏开发公司建立深度合作,将国外优质的前沿产品体验带到中国,也将中国的游戏带向世界。 2021年3月,腾讯游戏针对国际业务推出了在线游戏开发平台Level Infinite PGOS(Level Infinite Game Online Service)。Level Infinite PGOS是一种游戏在线服务解决方案,旨在降低游戏后端开发和维护的难度,同时降低成本,从而使开发者专注于游戏玩法与核心逻辑开发。 业务挑战 产品力主导游戏行业竞争 游戏出海面临多重挑战 全球游戏市场规模不断扩大,游戏产业已成为一种重要的文化产业。据市场调查机构 Newzoo 最新数据显示,2023年全球游戏市场规模预计将超过1877亿美元,同比增长2.6%。 游戏市场潜力巨大、前景看好,与此同时游戏开发竞争也变得异常激烈。游戏品类越来越多、玩法越来越多,给游戏开发带来更高要求:游戏设计和架构越来越复杂,游戏开发成本水涨船高,游戏复用性较低,新项目启动门槛更高等等。 尤其对于腾讯游戏海外业务来说,研发更需要具备全球发行、全球部署的能力,直接拉升了对于底层数据架构的要求,当中包括:需要提供多租户SaaS模式;能够物理隔离每个游戏大区,满足全球各个区域的隐私保护;以及可在全球各个地区进行分布式部署、自动扩容、缩容等。 解决方案 深度契合游戏业务场景 为了应对游戏行业的最新趋势和海外市场的挑战,腾讯游戏推出为海外游戏而设的Level Infinite PGOS通用平台。Level Infinite PGOS是一套多租户SaaS游戏后台解决方案,采用全球化分布式架构,在欧洲、北美、日韩、东南亚等游戏发行热点区域部署运行。 数据库是游戏软件的核心组件,游戏玩家的各种信息、运营数据、游戏场景数据等都需要借助数据库来保存。对于数据存储系统,Level Infinite PGOS根据自身场景进行了严格的测试选型,最终采用MongoDB作为核心存储组件,成为一整套覆盖游戏各个维度、各个生命周期的解决方案。 玩家数据存储 – 与传统游戏开发有所不同,使用MongoDB去存储玩家的基础数据,不会将MongoDB直接暴露给游戏去使用,开发者无需关心底层的数据细节,即可直接灵活定义数据,例如,不同游戏可定义不同数据模板。此外,MongoDB 支持多种数据类型和数据原子运算,易于实现幂等操作;而且基于MongoDB的分片可横向扩容,对于一些爆款游戏来说,这一点很重要,可以不用担心玩家规模的快速增加。 智能对局匹配 – 在很多游戏中,都需要在平台上匹配两名玩家去进行对局竞技。以腾讯游戏为例,如果是一款全球发行的游戏,就有可能在不同游戏大区匹配到两名玩家,这种匹配看似随意,但却需要后台具备强有力的数据处理能力。技术调度要同时满足不同区域的服务器集群,也就是满足不同场景需求下的服务器扩容。腾讯游戏底层通过MongoDB实现原子化操作,经过玩家各种属性的对比,找到一个距离各个玩家最近的服务器,并进行服务器分配,最终形成一个对局。 游戏内经济系统闭环 – 假设把游戏内的经济系统理解为一个特殊的交易场景,在处理游戏交易的过程中,涉及到订单、退款、跟踪回溯等多个环节,而通过MongoDB的原子化、事务性操作可以将整个交易流程一次性完成。虽然游戏内交易是虚拟的、复杂的,但采用MongoDB可以保证交易是规范的、完整的。 数据流能力 – 游戏开发者需要跟踪玩家各种行为事件,以便形成流水日志,同时要保证玩家的所有事件是可追溯、可查询的。MongoDB的数据库实例,可将这些流水日志存储起来,并基于灵活的文档结构,让开发者不论是在开发期间、还是游戏已经发行的期间,都可以快速检索玩家的所有事件流。 客户价值 优化开发体验 拉升运维能力 根据数据显示,2023年一季度,腾讯的游戏领域收入达到483亿元,而其中132亿元来自国际市场,占游戏整体收入的27%,可见腾讯在海外市场的巨大潜力和影响力。对腾讯游戏而言,全球化协作体系已然成型。在多元化布局和全球化视野之下,中国游戏既要在内容创新、玩法创新上学习更多,也要将支持大规模玩家在线的后台技术越做越强。 回顾与MongoDB的合作历程,腾讯海外游戏Level Infinite PGOS平台负责人谢磊谈到,无论在功能还是性能上,MongoDB都很好地契合了游戏业务场景,带给腾讯游戏的不只是功能价值,还有运维价值: 简单、易用的控制台 实现全面托管服务,即时自动扩缩容的专用服务器,为实时游戏提供低延迟和高可靠性。 丰富的可视化监控 提供实时可视化日志、监控面板,研发人员、管理人员可以实时监控业务运行状态。 一键升配、降配能力 在访问量突增时,一键自动扩容保障业务的正常运行;在流量低谷,一键自动缩容以节约成本。 多维告警能力 提供运行时间、状态异常等多维度告警能力,使问题可以在最短的时间内被捕捉并通知到用户。 客户证言 腾讯海外游戏Level Infinite PGOS平台负责人 谢磊: “游戏行业的发展越来越由产品力主导。Level Infinite PGOS平台最大的使命是要能够将更新、更现代化的开发模式与腾讯已有能力结合在一起,形成开发体验良好的平台,其中尤为重要的是需要符合海外开发人员开发习惯。正是在这样的背景之下,我们的底层技术选择使用MongoDB。可以说,MongoDB让腾讯游戏Level Infinite PGOS如虎添翼。”

November 29, 2023

A Year of Thrill: Celebrating the New MongoDB University

Staying ahead in the ever-evolving tech world is like being on a rollercoaster - it’s exciting but it can also make your head spin! When we set out to revamp MongoDB University , we wanted to provide developers with frictionless access to the learning content they needed to conquer their challenges. It’s been one year since the launch and we are over the moon about how far the new MongoDB University has come - a one-stop hub with fresh certifications and new content, all available online. But none of this would have been possible without the incredible support of our engaged learners who have embarked on this ride with us. Our commitment to delivering top-notch educational resources has been nothing short of award-winning, earning us the prestigious Silver Excellence Award from the Brandon Hall Group in the category of ‘Best Advance in Creating an Extended Enterprise Learning Program’. The success of the new University has also been featured at industry events, including Cognition and the Customer Education Management Association Conference. So, let’s buckle up and take a tour through the revamped MongoDB University! New content With over 1,000 learning assets, including videos, hands-on labs, code recaps, quizzes, and courses, there’s something for everyone. Plus, now we have you covered with language subtitles in Chinese (Traditional and Simplified), Korean, Spanish, German, Japanese, Italian, French, and Portuguese. The best part? All of the content is free, online, and you can take your time and learn at your own pace. Let’s explore three of our newest courses: Data Modeling for MongoDB: This course guides you through the foundational steps of creating an effective data model in MongoDB, including identifying entities and workloads, mapping and modeling relationships between entities, and using key schema design patterns. Atlas Essentials: In this course, you’ll gain the foundational knowledge and skills needed to use MongoDB Atlas, the multi-cloud developer data platform. MongoDB for SQL Professionals: This course will help you leverage your SQL skills to get started with MongoDB quickly. You can practice what you learn and gain valuable real-world skills with labs hosted in our in-browser development environment. The new experience allows you to explore hands-on exercises as part of our courses, or you can dive directly into a standalone lab . The labs include step-by-step instructions that guide you through each scenario and even provide hints along the way. And for those looking for nuggets of MongoDB wisdom, explore the catalog of over 30 Learning Bytes . These short videos cover a wide variety of topics and are designed to help you get the knowledge you need quickly. New certifications Our freshly revamped certifications are recognized by professional institutions and are your ticket to having your knowledge and skills formally validated and recognized by MongoDB. They are a great way to elevate yourself in your current role and increase your marketability for future roles. Certifications come with bragging rights, inclusion in the Credly Talent Directory, and a shiny Credly badge that makes it easy for you to share your achievement. So, let’s explore the two new certifications: MongoDB Associate Developer: Certify that you possess the essential skills to create beginner-level applications utilizing MongoDB as a backing database for Java, Python, C#, PHP, or Java applications. MongoDB Associate Database Administrator: Validate your MongoDB database administration skills by certifying your knowledge of building, supporting, and securing MongoDB infrastructure. And if you need a boost, once you complete one of the certification learning paths you will automatically unlock a 50% discount on a certification exam. Educators and students can check out the Academia program to learn how to receive a free exam. All aboard! This is just the beginning of the adventure and we are excited for what is yet to come. So, fasten your seatbelt, and let’s keep learning together! With over 1,000 learning assets, MongoDB University has what you need to pick up new skills and advance your career. Explore free courses, practice with hands-on labs, and earn MongoDB certifications.

November 28, 2023

深入理解用户需求,利用 AI 和 MongoDB 提升内容个性化

内容无处不在。 无论消费者寻找什么,也无论消费者所处任何行业,找到内容并不困难,关键在于如何找到对应的内容。而这就是 Concured 专注的领域。 Concured 是一家来自蒙特利尔的 AI 初创企业,致力于协助市场营销团队对标受众,有的放矢地打造网站和销售内容;同时,帮助内容营销团队脱颖而出,加速基于洞察驱动的内容个性化。 2015 年,首席执行官 Tom Salvat 创立 Concured,旨在助力内容营销机构更深入地理解受众需求,交付更具影响力的内容。 Built with MongoDB 栏目采访了公司成立约一年后加入公司的 首席技术官 Tom Wilson ,话题围绕 Concured 的人工智能使用情况、Wilson 加入 Concured 的过程以及公司未来的发展。 Built with MongoDB: Concured 是做什么的? Tom Wilson: Concured 是一家软件公司,以过去 5-10 年间发展的人工智能技术为基础,帮助市场营销机构了解如何撰写具体领域的宣发物料,发掘自身内容亮点,把握竞争对手以及行业宣发现状。进而打造个性化的客户网站使用体验,最大化内容投资收益比。 Concured 已成功推出一套内容推荐系统,能够为每位访问者提供针对性服务。这套系统注重用户隐私,不使用任何第三方 cookie 或用户监视技术,完全基于网站用户的操作,通过访问者的点击行为勾勒出其兴趣领域。随着用户的兴趣和目的逐渐清晰,这套系统会尝试推荐新的阅读内容,比如博文、产品介绍或其他类型的内容。 Built with MongoDB:您刚刚提到了人工智能,那么 Concured 是如何使用人工智能的呢? Wilson: 我们运用人工智能的场景有很多。有别于其他个性化系统,Concured 的卖点之一是不需要过长的整合期,也无需在日常管理中进行维护。实现的途径是借助 AI 机器人剖析客户网站的内容,发掘相关性,提取文本、标题和其他所有相关元数据,然后完成自动索引。 我们的系统利用先进的自然语言处理 (NLP) 技术,为每个文档生成语义元数据,在多维空间中与特定点相对应。另一方面是理解其中的关键实体,以及同一个网站中某一篇特定文章与其他文章之间的关系。我们利用 AI 支持的网络爬虫找到的所有内容都会自动附上海量元数据。 [图片] Built with MongoDB:AI 并不总是 100% 准确,您在 Concured 打造的 NLP 的准确度怎么样呢? Wilson: 以内容推荐系统而言,很难断定什么是最佳推荐,因为即便是同一个人,根据日期或网络操作的不同,推荐也会有所变化。例如一些知名的推荐系统,如 Netflix、Amazon 和 Spotify,总是在猜我们接下来想看什么,但却没有一个是对应的答案。 正因如此,绩效评价变得非常困难。所以,我们采取的方式是不提供 100% 对应的答案,而是通过改变算法,来看访问者是否会点击更多的文章,是否会进入网站运营商定义的目标页面,比如产品页面或注册表单。在网站访问人员中,最终执行该操作的人数比例越高,说明推荐系统越出色。我们可以对比客户在启用 Concured 个性化系统前后的网站成功率。截至目前,我们已经看到 2-3 倍的提升,算法一直在完善中。 Built with MongoDB:您是何时加入到 Concured 团队的? Wilson: 当时公司已经得到第一笔来自外部的巨额投资,条件之一是引入一位专业的首席技术官。这种情况在企业初创期比较常见,投资方想要介入企业架构,把控资金流向,减少鲁莽行事。所以,有些企业将其戏称为“家长式监督”。我不知道这算不算是我的角色。不过虽然当时团队已经很强大,但我还是从架构入手,从根上上确保我们能够实现后续的目标,以及更长期的战略规划和技术愿景。 Built with MongoDB:您的团队是如何选择 MongoDB 的? Wilson: 我加入时,团队已经在使用 MongoDB。加入后的几个月里,我们讨论过是否要换用结构化的数据库,并且必须制定出一项决议。所以我才参与其中,经过深思熟虑,决定继续使用 MongoDB。事实证明是一项完全正确的决议,有利于我们实现最初的愿景。同时,我们将弃用 Google Cloud Platform 上的社区版本,换用 MongoDB Atlas Serverless。令人欣喜的是,因为新平台无服务器,我们将不再需要管理各种机器,还能够使用 Atlas 上的文本搜索功能,顺便简化一下我们的技术堆栈。作为一家企业,就我们当下所处的位置以及未来五年的发展方向而言,MongoDB 始终是我们正确的选择。 Built with MongoDB: Concured 的未来是什么样的? Wilson: 就在我们交谈的过程中,未来已经被书写。此时此刻,越来越多和我们大客户有着相同需求的企业正在找到我们。这些企业那些有着海量、已存档的内容,需要继续从中盈利,继续发布更多内容。无论是否是大型企业,比如咨询或金融服务行业或传统的出版商,确定的一点是,以相应的 KPI 为基准,产出利益最大化的对应内容。 Built with MongoDB:您收到的最好反馈是什么? Wilson: 我的团队给我的一条正面反馈,是说我有担当。如果他们遇到问题,我会出手解决或者减少阻力,这样他们就可以全力以赴解决问题。这是我的人生观:如果你用心领导团队,事情就会自然顺利推进。 对于任何投资内容宣发的企业而言,最大化投资回报都是无可争议的商业诉求。

November 28, 2023


什么是MongoDB分片? 分片是一种将数据分布或分割到多台计算机上的方法。相较于单个计算机,分片技术允许您进行水平扩展,这在大型现代工作负载的场景下是非常有用的。 水平扩展,也称为横向扩展,是指添加计算机来共享数据集和负载。水平扩展允许进行接近无限的扩展,以处理大数据和强烈的工作负载。 通过分片实现横向扩展 通过分片,您可以自动将MongoDB数据库跨多个节点和区域进行扩展,以处理写入密集型工作负载、不断增长的数据大小以及数据存储要求。 使用MongoDB的分片,您可以在应用程序增长超出单个服务器的硬件限制时,在无需增加应用程序复杂性的情况下,无缝地扩展数据库。 为了响应不断变化的工作负载需求,可以在分片之间迁移文档,并随时向群集中添加或删除节点 - MongoDB将自动根据需要重新平衡数据,无需手动干预。 分片的好处 分片允许您将数据库扩展以处理几乎无限的负载增加。它通过增加读/写吞吐量和存储容量来实现这一点。具体来说: 增加的读/写吞吐量:通过将数据集分布到多个分片上,您可以利用并行处理来增加读/写的吞吐量。假设一个分片可以每秒处理一千次操作,每增加一个分片,您将多获得额外的一千次每秒的吞吐量。 增加的存储容量:同样地,通过增加分片的数量,您还可以增加总体的存储容量。假设一个分片可以容纳4TB的数据。每增加一个分片,您的总存储容量将增加4TB。这样可以实现接近无限的存储容量。 数据本地性:区域分片允许您轻松创建分布式数据库,以支持地理分布的应用程序,并通过强制数据在特定区域内驻留的策略来实现。每个区域可以有一个或多个分片。 MongoDB中的分片策略 大多数分布式数据库在处理数据分布时,是通过简单地对主键值进行散列,将数据随机分布在集群节点中。这在查询跨节点的数据时会带来性能损失,并且在需要将数据本地化到特定区域时会增加应用程序的复杂性。 MongoDB 可以提供多种分片策略,提供对于数据分布更好的方法。。数据可以根据查询模式或数据位置要求进行分布,从而在各种工作负载下实现更高的可扩展性: 范围分片。文档根据分片键值分区到分片上。分片键值彼此接近的文档可能位于同一个分片上。这种方法非常适用于需要优化基于范围的查询的应用程序,例如将特定区域所有客户的数据放置在特定分片上。 散列分片。文档根据分片键值的MD5散列进行分布。这种方法保证了写入在分片上的均匀分布,通常对于摄取时间序列和事件数据流是最优选择。 区域分片。提供了开发人员定义在分片群集中数据放置的特定规则的能力。 MongoDB Atlas中的全局群集 完全托管的云数据库服务MongoDB Atlas允许您使用可视化用户界面或Atlas API快速实现区域分片。您可以轻松创建分布式数据库以支持地理分布的应用程序,并通过强制在特定区域内存储数据的策略来实现数据存储。 使用阿里云MongoDB 分片集群为始终在线、全球分布式的写入应用程序提供服务 要确保充分发挥分片的优势,您需要遵循一系列最佳实践。 确保分片键的均匀分布 当读取和写入的分片键不均匀分布时,操作可能会受限于单个分片的容量。当分片键均匀分布时,没有单个分片会限制系统的容量。 避免散布-汇集查询用于运营工作负载 在分片系统中,不能基于分片键进行路由的查询必须广播到所有分片进行评估。由于这些查询涉及每个请求的多个分片,随着添加更多分片,这些查询不会呈线性扩展,并且需要额外的开销来合并来自多个分片的结果。您应该在查询中包含分片键,以避免散布-汇集查询。 这一规则的例外是大型聚合查询。在这些情况下,散布-汇集可以是一种有用的方法,因为它允许查询在所有分片上并行运行。 在适当的时候使用基于散列的分片 对于发出基于范围的查询的应用程序,基于范围的分片是有益的,因为操作可以路由到最少的分片,通常是一个分片。然而,基于范围的分片需要对数据和查询模式有很好的理解,在某些情况下可能不切实际。 基于散列的分片确保读取和写入的均匀分布,但不提供高效的基于范围的操作。 预分割和分发分片 在创建新的分片集合以加载数据时,首先做集合的预分片,并将它们均匀分布在所有分片上,然后再加载数据。对于基于散列的分片,您可以使用numInitialChunks来自动执行此操作。 下一步 以上就是MongoDB 性能最佳实践中关于分片的内容。接下来我们将介绍事务相关的实践。

November 28, 2023