Files saved to S3

[Pavan B]

Hello,

We are considering deploying M3 in AWS cloud and would like to know if there is an option to store the files in S3 instead of local files or block storage? Ideally, we would like to use AWS instances with large memory that can be used to return results for data in cache and when the data is not in cache, read from S3 to load the cache and serve the queries.

We are hoping this model would perform better than (if not similar to) Thanos because most queries would hit memory, while giving us the ability to store the metrics persistently in S3 to save cost.

If this is not currently supported, is there a plan to support it in the future?

Thanks

-Pavan

[Rob Skillington]

Hi Pavan,

Thanks for the query. At this time there is no native integration with S3.

That being said, there is the ability to mount different directories to different file systems and since M3 does not perform any significant amount of compactions on data sets you could feasibly mount the historical “data” and “index” directories to a fuse filesystem that is backed by S3 and leave the “commitlogs” and “snapshots” directories on an SSD and still get great write performance.

Please note: I have not tried this, although I would love to hear people’s results.

Here’s a popular S3 fuse filesystem:

https://github.com/s3fs-fuse/s3fs-fuse

I have heard from users who have used spinning disks for “index” and “data” directories with “commitlogs” and “snapshots” on SSDs and they have had good success saving costs on disk with that model.

Best,

Rob

--
You received this message because you are subscribed to the Google Groups "M3" group.
To unsubscribe from this group and stop receiving emails from it, send an email to m3db+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/m3db/0a936cff-e682-4212-bec4-2bd10332cd0d%40googlegroups.com.