If I have a collection with fields A, B and C indexed, will those indexes be preserved when data gets chopped off and moved to Online Archive? In other words, if I later query the data in Online Archive by those fields will the queries be just as fast as if it were in the original db?
Short answer: absolutely not
Longer answer. Think about it. If they could make it as fast, normal db with the complexity of tree based indexes, working set in RAM with cache, would be useless.
Cold data == only on S3 == Hard Drives.
Hot data == Atlas cluster == RAM + CPU + Hard Drives.
Because of these partitions, the data in S3 will be stored in folders so if your query contains these fields, the query resolver can eliminate quickly many many “cold” files and find the right documents WAY faster instead of scanning everything.
Thanks for links Maxime. However the one about Online Archive gives me:
404 Not Found Code: NoSuchKey Message: The specified key does not exist. Key: online-archive RequestId: HJDNS7FBJK3BNCV7 HostId: zdYTTiasTSMSlnCRSsXfHBzijVLRwGyw9iTsb7xpIPE30poBzgNB3U3rb5H52bi7p0V2jzWPVYo=
There is a lot of things that I don’t know
B-(. The concept of partitions is very nice.
I fixed the link!