[pve-devel] ceph firefly in pve repos

[Alexandre DERUMIER]

>>Considering the price, you can buy twice as many crucial disk for same money. This would lead to a cluster 
>>with same write performance, but much better read? 

Yes I think. (I'll try to tests with more osds, but I think the cpu will be the bottleneck for reads)


Note that I don't have burned them yet, so I don't known how the performance will be after a lot of writes.

also  for endurance
crucial m550 : 72 TBW
intel s3500 : 450 TBW

(but intel have 200GB overprovising, do maybe crucial with manual overprovision could give same endurance)


Also, note that writes are benched with directio, but in real world application, with cache=writeback in qemu,
it's faster. 



----- Mail original ----- 

De: "Dietmar Maurer" <dietmar at proxmox.com> 
À: "Alexandre DERUMIER" <aderumier at odiso.com> 
Cc: pve-devel at pve.proxmox.com, "Stefan Priebe - Profihost AG" <s.priebe at profihost.ag> 
Envoyé: Vendredi 12 Septembre 2014 17:05:08 
Objet: RE: [pve-devel] ceph firefly in pve repos 

So I try to summarize the results you posted for on ceph-devel: 

Intel s3500 rdb 4k write: 5200 IOPS 
Intel s3500 rdb 4k read: 21634 IOPS 
Crucial m550 rbd 4k write: 3081 IOPS 
Crucial m550 rbd 4k read: 20151 IOPS 

Read performance is about the same, but write is twice as fast with intel. 

Considering the price, you can buy twice as many crucial disk for same money. This would lead to a cluster 
with same write performance, but much better read? 


> results of fio on rbd with kernel patch 
> 
> 
> 
> fio rbd crucial m550 1 osd 0.85 (osd_enable_op_tracker true or false, same 
> result): 
> --------------------------- 
> bw=12327KB/s, iops=3081