[pve-devel] Ceph Octopus is out!

[Alexandre DERUMIER]

Also, ceph server side, the new io_uring seem exciting too :)

https://github.com/ceph/ceph/pull/27392

----- Mail original -----
De: "Thomas Lamprecht" <t.lamprecht at proxmox.com>
À: "pve-devel" <pve-devel at pve.proxmox.com>, "aderumier" <aderumier at odiso.com>
Envoyé: Mercredi 25 Mars 2020 12:00:21
Objet: Re: [pve-devel] Ceph Octopus is out!

On 3/25/20 11:17 AM, Alexandre DERUMIER wrote: 
> I'm pretty excited by the new write-around cache policy for librbd && io scheduler:) 
> 
> (better than writeback, no read latency impact :) 
> 

Better for high perf. write workloads: 
> Writes return immediately under both the write-around and write-back policies, 
> unless there are more than rbd cache max dirty unwritten bytes to the storage 
> cluster. The write-around policy differs from the write-back policy in that it 
> does not attempt to service read requests from the cache, unlike the write-back 
> policy, and is therefore faster for high performance write workloads. Under the 
> write-through policy, writes return only when the data is on disk on all 
> replicas, but reads may come from the cache. 
-- https://docs.ceph.com/docs/octopus/rbd/rbd-config-ref/#cache-settings 

Not sure if all workloads benefit from that. 


> librbd now uses a write-around cache policy be default, 
> replacing the previous write-back cache policy default. 
> This cache policy allows librbd to immediately complete 
> write IOs while they are still in-flight to the OSDs. 
> Subsequent flush requests will ensure all in-flight 
> write IOs are completed prior to completing. The 
> librbd cache policy can be controlled via a new 
> “rbd_cache_policy” configuration option. 
> 
> librbd now includes a simple IO scheduler which attempts to 
> batch together multiple IOs against the same backing RBD 
> data block object. The librbd IO scheduler policy can be 
> controlled via a new “rbd_io_scheduler” configuration 
> option. 

Let's see how much this brings in performance :)