Documentation #3960
closed
[Document bug]MON and MDS do not need a ssd for data storage.
Added by Xiaoxi Chen over 11 years ago.
Updated over 11 years ago.
Description
From :http://ceph.com/docs/master/install/hardware-recommendations/#data-storage
it says:
Since the storage requirements for metadata servers and monitors are so low, solid state drives may provide an economical opportunity to improve performance.
But actually, MDS do not store any data to local , and MON is not IO intensive, this part of document seems wrong.
- Status changed from New to In Progress
- Assignee set to John Wilkins
You are correct. The machines and processes would only boot a bit faster. The way to accelerate metadata servers is to assign the pool to a CRUSH ruleset. I'll modify the docs accordingly.
- Status changed from In Progress to Resolved
I removed the reference to monitors, added detail on sequential write throughput, and a link to an example for a CRUSH hierarchy that includes SSD devices only.
John Wilkins:
What do you mean by:
One way Ceph accelerates filesystem performance is to segregate the storage of metadata from the storage of >the underlying object data. Ceph provides a default metadata pool. You will never have to create a pool for >metadata, but you can create a CRUSH map hierarchy for your metadata that points only to a host’s SSD >storage media. See Mapping Pools to Different Types of OSDs for details.
The metadata pool aim at store CephFS meta,right? other metadata such like pginfo and pglog go to the pool where data goes.So I doubt ssd_based pool for metadata is really make sense for usages except CephFS
And here may be a typo:
Sequential Writes: When you store multiple journals on an OSD (it should be SSD I suppose) you must consider the sequential write limitations of the SSD too, since they may be handling the requests of multiple OSD journals simultaneously.
Also available in: Atom
PDF