[PVE-User] Poor CEPH performance? or normal?
athompso at athompso.net
Sun Jul 29 16:41:24 CEST 2018
On July 28, 2018 6:00:15 AM CDT, Mark Adams <mark at openvs.co.uk> wrote:
>Thanks for your great round up there - Your points are excellent.
>What I have ended up doing a few days ago (apologies have been too busy
>respond..) was set rbd cache = true under each client in the ceph.conf
>This got me from 15MB/s up to about 70MB/s. I then set the disk holding
>zfs dataset to writeback cache in proxmox (as you note below) and that
>bumped it up to about 130MB/s -- Which I am happy with for this setup.
>On 27 July 2018 at 14:46, Adam Thompson <athompso at athompso.net> wrote:
>> On 2018-07-27 07:05, ronny+pve-user at aasen.cx wrote:
>>> rbd striping is a per image setting. you may need to make the rbd
>>> image and migrate data.
>>> On 07/26/18 12:25, Mark Adams wrote:
>>>> Thanks for your suggestions. Do you know if it is possible to
>>>> existing rbd pool to striping? or does this have to be done on
>> Please be aware that striping will not result in any increased
>> performance, if you are using "safe" I/O modes, i.e. your VM waits
>> successful flush-to-disk after every sector. In that scenario, CEPH
>> never give you write performance equal to a local disk because you're
>> limited to the bandwidth of a single remote disk [subsystem] *plus*
>> network round-trip latency, which even if measured in microseconds,
>> adds up.
>> Based on my experience with this and other distributed storage
>> believe you will likely find that you get large write-performance
>> 1. use the largest possible block size during writes. 512B sectors
>> the worst-case scenario for any remote storage. Try to write in
>> *at least* 1 MByte, and it's not unreasonable nowadays to write in
>> of 64MB or larger. The rationale here is that you're spending more
>> sending data, and less time waiting for ACKs. The more you can tilt
>> in favor of data, the better off you are. (There are downsides to
>> sector/block/chunk sizes, though - this isn't a "free lunch"
>> 2. relax your write-consistency requirements. If you can tolerate
>> small risk with "Write Back" you should see better performance,
>> during burst writes. During large sequential writes, there are not
>> ways to violate the laws of physics, and CEPH automatically amplifies
>> writes by (in your case) a factor of 2x due to replication.
>> 3. switch to storage devices with the best possible local write
>> OSDs. OSDs are limited by the performance of the underlying device
>> virtual device. (e.g. it's totally possible to run OSDs on a
>> RAID6 controller)
>> 4. Avoid CoW-on-CoW. Write amplification means you'll lose around
>> your IOPS and/or I/O bandwidth for each level of CoW nesting,
>> workload. So don't put CEPH OSDs on, ssy, BTRFS or ZFS filesystems.
>> worst-case scenario would be something like running a VM using ZFS on
>> of CEPH, where the OSDs are located on BTRFS filsystems, which are in
>> virtual devices hosted on ZFS filesystems. Welcome to 1980's storage
>> performance, in that case! (I did it without realizing once...
>> 5 MBps sequential writes was a good day!) FWIW, CoW filesystems are
>> generally awesome - just not when stacked. A sufficiently fast
>> NAS running ZFS with VMs stored over NFS can provide decent
>> *if* tuned correctly. iX Systems, for example, spends a lot of time
>> effort making this work well, including some lovely HA NAS
>> 5. Remember the triangle. You can optimize a distributed storage
>> for any TWO of: a) cost, b) resiliency/reliability/HA, or c) speed.
>> is a specific case of the traditional good/fast/cheap:pick-any-2
>> I'm not sure I'm saying anything new here, I may have just summarized
>> discussion, but the points remain valid.
>> Good luck with your performance problems.
>> pve-user mailing list
>> pve-user at pve.proxmox.com
>pve-user mailing list
>pve-user at pve.proxmox.com
That's a pretty good result. You now have some very small windows where recently-written data could be lost, but in most applications not unreasonably so.
In exchange, you get very good throughout for spinning rust.
(FWIW, I gave up on CEPH because my nodes only have 2Gbps network each, but I am seeing similar speeds with local ZFS+ZIL+L2ARC on 15k SAS drives. These are older systems, obviously.)
Thanks for sharing your solution!
Sent from my Android device with K-9 Mail. Please excuse my brevity.
More information about the pve-user