SEO

I have a question about the usage of disk arrays. I have not
had the chance/need for building a "larger" system. I admin
about 6 Dell Poweredge servers running Linux. One is a database
server, and the rest are web, file and app servers.

My Sun/Solaris usage is limited at the moment to using my U60 and U5
as desktops for daily usage.

For my own education, how does using a disk array differ from just
using a server in its place? I see an array as a massive storage
facility, whether that be a file server or running a database. Rather
than a disk array, why not have an actual server with a ton of space?
Either way you can have multiple web/app servers accessing the data?

Could this be because such a server with high storage capacity would
likely have high processing power, and therefore you are paying for
power rather than storage?

Are all disk arrays connected to their servers via Fiber Channel
cables,
making for better through-put?

Can someone explain to me where disk arrays fit into the larger
system configurations?

Thanks for your help!

p.s. I am aiming at our next server purchase to be a simple Sun
box like a V120.....just because! :-)

Robert Mazur wrote:

> I have a question about the usage of disk arrays. I have not
> had the chance/need for building a "larger" system. I admin
> about 6 Dell Poweredge servers running Linux. One is a database
> server, and the rest are web, file and app servers.
>
> My Sun/Solaris usage is limited at the moment to using my U60 and U5
> as desktops for daily usage.
>
> For my own education, how does using a disk array differ from just
> using a server in its place? I see an array as a massive storage
> facility, whether that be a file server or running a database. Rather
> than a disk array, why not have an actual server with a ton of space?
> Either way you can have multiple web/app servers accessing the data?
>
> Could this be because such a server with high storage capacity would
> likely have high processing power, and therefore you are paying for
> power rather than storage?
>
> Are all disk arrays connected to their servers via Fiber Channel
> cables,
> making for better through-put?
>
> Can someone explain to me where disk arrays fit into the larger
> system configurations?
>
> Thanks for your help!
>
> p.s. I am aiming at our next server purchase to be a simple Sun
> box like a V120.....just because! :-)

I think you need to redefine your terminology here.

A disk array is simply a lot of disks accessible by a system. This can be as
simple as the built in disks on your poweredges. If you have a disk array
which can be attached to only one machine then this is normally known as
"DAS" or Direct Attached Storage.

If you have a number of disks which can be attached to by several machines
at once then this is normally a SAN or Storage Area Network. All this means
is that several machines can access the disks at once. Normally this access
is by means of FC-AL which is Fibre Channel Arbitrated Loop. This is a
means of attaching lots of disk controllers in a chain and allowing any
unit on the chain to talk to any other. Units can either be storage devices
(disk drives) or hosts (servers). In higher end SANs the access is via
switched Fibre Channel which works much like switched ethernet in that any
device can ask to speak to any other device and they will have a dedicated
path betweent them for the duration of the communication. This switching is
normally achieved by Director Switches.

If you have a number of disks which can be accessed over a TCP network then
this is normally known as NAS or Network Attached Storage. This normally
has a head unit attached to some kind of SAN with the head unit providing
network, NFS, CIFs and the like. In the case of NetApp the head unit is
effectively a high spec PC running an OS (DataOnTap) which is optimised to
hell and beyond for network operation. In the case of BlueArc most of the
services are performed in hardware.

You seem to be asking about NAS storage and the answer is that there is very
little difference between a NAS filer and a server with a ton of disk
attached to it. The only thing the filer tends to have going for it is that
its software is better optimised for the duty than a general purpose OS.
This is something of a generalisation that only actually applies at the
high end, but it's a useful idea.

SAN storage tends to be attached via FC running at either 1Gb/s or 2Gb/s.
The main thing that FC has going for it is that it has minimal overhead as
all it transfers is basic SCSI III commands and a few control commands.
Nowadays there is a concept coming up called iSCSI which is blurring the
lines. Where FC is SCSI running over dedicated fibre iSCSI is SCSI
encapsulated in TCP which allows you to use your existing IP network for
your storage infrastructure.

Where SANs start becoming useful is when you have lots of machines with
varying disk space requirements. You could just attach storage directly to
each machine, but this is very inefficient as you can end up with machines
with not enough disk space and machines with far more than they can use and
no way to exchange the space. With a SAN you can dynamically allocate the
space so that each machine has the storage it needs and no more nor less.
You also get the ability for several machines to attach to the same logical
disk which allows for such things as failover of filesystems between
machines and application clustering such as Oracle 9i RAC provides.

SANs are basically complicated networks of disks and controllers with some
intelligence built in. Most SAN systems provide support in the hardware for
RAID - with EMC kit this tends to be RAID 1 (mirrored disks) which is as
resilient as you can get. HDS kit supports RAID 0,1 and 5 which gives you
an option (never run critical applications on RAID0 - it's fast, but if you
lose one disk you have lost all of your data).

P.

Very few servers have the physical capacity to hold large numbers of disks.
Using 2 or 4 large capacity disks can prove to be slow because when the disk
is busy it can't do another job. Then what if one disk fails you lose a
chunks of your storage. That's why 12 smaller capacity disks are better
than 4 large. Fiber channel is faster than SCSI and can be a distance away
from the host too
"Robert Mazur" <mazur@raincode.net> wrote in message
news:9448144d.0402051516.2b41272@posting.google.co m...
> I have a question about the usage of disk arrays. I have not
> had the chance/need for building a "larger" system. I admin
> about 6 Dell Poweredge servers running Linux. One is a database
> server, and the rest are web, file and app servers.
>
> My Sun/Solaris usage is limited at the moment to using my U60 and U5
> as desktops for daily usage.
>
> For my own education, how does using a disk array differ from just
> using a server in its place? I see an array as a massive storage
> facility, whether that be a file server or running a database. Rather
> than a disk array, why not have an actual server with a ton of space?
> Either way you can have multiple web/app servers accessing the data?
>
> Could this be because such a server with high storage capacity would
> likely have high processing power, and therefore you are paying for
> power rather than storage?
>
> Are all disk arrays connected to their servers via Fiber Channel
> cables,
> making for better through-put?
>
> Can someone explain to me where disk arrays fit into the larger
> system configurations?
>
> Thanks for your help!
>
> p.s. I am aiming at our next server purchase to be a simple Sun
> box like a V120.....just because! :-)



x-- 100 Proof News - http://www.100ProofNews.com
x-- 3,500+ Binary NewsGroups, and over 90,000 other groups
x-- Access to over 800 Gigs/Day - $8.95/Month
x-- UNLIMITED DOWNLOAD