Date: Wed, 2 Oct 1996 08:27:44 -0500 (CDT) From: Joe Greco <jgreco@brasil.moneng.mei.com> To: cassy@loop.com (Cassandra Perkins) Cc: freebsd-isp@freebsd.org Subject: Re: RAID Controller Product Message-ID: <199610021327.IAA05083@brasil.moneng.mei.com> In-Reply-To: <Pine.BSF.3.91.961001155258.14781B-100000@patty.loop.net> from "Cassandra Perkins" at Oct 1, 96 04:05:41 pm
next in thread | previous in thread | raw e-mail | index | archive | help
> Hello....
>
> Does anyone know of a good scsi-to-scsi RAID controller that can do RAID
> level 5 and that would be great for a file server application. I'm
> presently using disk ccd for our news servers. Each server carries the
> same information, so I'm looking to consoladate the information on one
> fileserver.
I am curious as to why you would choose to do this.
If you currently have your I/O spread amongst several machines, you gain
from having the data replicated. If you have N machines, you have
approximately N times the I/O bandwidth available as compared to a
solution where you only have the main system's I/O bandwidth available.
Put another way:
Consider a system with redundant disks (let's say two). For read
operations, each disk is allowed to perform operations simultaneously -
and they certainly do not have to be participating in the _same_
operation. Assuming that you are using GigaWonder disk drives capable
of fetching 1000 articles per second.. You have essentially:
1) increased your potential I/O transfers per second (2 drives = 2000
articles per second).
2) Reduced your average time to complete a request ("latency") by
reducing contention. One drive would be busier than two and
ultimately gets saturated.
3) The saturation point essentially doubles.
These are all important factors in news processing.
Now... this on the surface should appear to have no real relation to
your problem.
However... I contend that if you currently have two complete news
servers with two complete sets of disks, you have the _exact_ _same_
_advantages_... but you have even replicated the hardware, so you
gain a hidden bonus - reliability. If one news server fails, you
still have fallback ability with the second one.
Therefore: I do not think what you are trying to do is a particularly
good idea. Easier, yes, better, well I remain unconvinced.
Incidental: I have set up a client with such a distributed news system
and it works extremely well, is very scalable, and it is really cool
to be able to take a machine out of the DNS roundrobin when it needs
to be serviced, etc... (the cluster is carefully overengineered so that
they have 'N+1' news servers where N is the number they really need).
Can you say 'zero service interruption'.
... JG
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199610021327.IAA05083>