[sheepdog-users] Sheepdog 0.6.0 schedule and todos

[Alexandre DERUMIER]

>>I got 80x times faster even with another HDD as journal device on my test. 
 
Seem great :)

>>See more info at 
>>http://lists.wpkg.org/pipermail/sheepdog/2012-November/007136.html 
>> Total write throughput: 33875968.0B/s (32.3M/s), IOPS 8270.5/s.

How many sas drives ?


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

De: "Liu Yuan" <namei.unix at gmail.com> 
À: "Alexandre DERUMIER" <aderumier at odiso.com> 
Cc: "MORITA Kazutaka" <morita.kazutaka at gmail.com>, sheepdog at lists.wpkg.org, sheepdog-users at lists.wpkg.org 
Envoyé: Mardi 6 Novembre 2012 09:32:06 
Objet: Re: [sheepdog-users] Sheepdog 0.6.0 schedule and todos 

On 11/06/2012 04:10 PM, Alexandre DERUMIER wrote: 
> One thing that could be great with journal, is something like slog 
> device with zfs or nvram with netapp san. 
> 

Currently I am writing it. To be honest, I have no idea of zfs and 
netapp san. 

The idea is that we log ahead all the updates of object in another 
device (be it a HHD or SSD or nvram) sequentially and use buffered 
IO/direct IO for actual data updates without 'O_DSYNC' flag. 

I got 80x times faster even with another HDD as journal device on my test. 

See more info at 
http://lists.wpkg.org/pipermail/sheepdog/2012-November/007136.html 

> random writes are going to a fast journal (ssd, nvram,...) then are 
> flushed sequentially each X seconds to slower disks. 
> 

The log write will currently is with O_DSYNC | O_DIRECT flag to assure 
data integrity. I will consider group commit to accelerate log write 
later when basic function is stable. 

Thanks, 
Yuan