Re: [vserver] NFS shares or iSCSI

From: PÁSZTOR György <pasztor_at_linux.gyakg.u-szeged.hu>
Date: Mon 12 Dec 2011 - 14:57:17 GMT
Message-ID: <20111212145717.GC20493@linux.gyakg.u-szeged.hu>

Hi,

"Michael S. Zick" <mszick@morethan.org> írta 2011-12-12 08:22-kor:
> Why not use a small SATA-3 SSD on each of the host machines
> as a cache in front of your network file or block system?

How could linux host machine use it as "cache"?
ZFS can have ZIL, and L2ARC. What can you do on linux?
You mean some local blockdev as external journal for some fs, eg. xfs?

> Caching locally about 5% of the network data would greatly
> reduce your network load.

Usually the load is because of the iowait. If the iSCSI box can signal
back, that it wrote the data onto disk (eg.: It's cached on zil), than the
blocked io can continue imediately.

> Unless you can upgrade your networking to 10Gbs fiber, I think
> this would give you about the next best performance.

Why fiber? The SFP+ DAC cables are much cheaper, however, on 10G with DCE
compliant devices, he could use FCoE, which contains less needless layers
than iSCSI, and if use a CNA, than it will do the "trick" to show the
things through a virtual hba. But those are just theory, I never see myself
a real working FCoE environment. But it could happen, to start with iSCSI,
than later it evolves to FCoE.
And... The problem is usually not the througput, but the latency!

Cheers,
György Pásztor
Received on Mon Dec 12 14:57:25 2011

[Next/Previous Months] [Main vserver Project Homepage] [Howto Subscribe/Unsubscribe] [Paul Sladen's vserver stuff]
Generated on Mon 12 Dec 2011 - 14:57:26 GMT by hypermail 2.1.8