set up a linux server
 

hi, I am interested in setting up my company server on a linux machine. before i can install and configure any process on a server machine, I first need to install it. what linux release you think is most suitable for server ? in my case, I would be managing up to 300 client machines in my company. secondly, would the server installation be straight forward? or I need to apply some things during linux server instalation?

If you want to purchase, go for something like this
new version may have been arrived , I am giving you just an idea.
But if you want to install it yourself, and you think you have proper configuration (enough memory, enough RAM and other important things)
then you can go for centOS.
centOS is quite robust as a server and it can bear weight of 300 client machines easily.

I'd recommend RHEL (latest is 6.4) if you want paid support, or the Centos equiv ( free rebuild of RHEL) if you are willing to do your own support.
Both options come with updates included.
You can read the manuals here http://www.linuxtopia.org/online_boo...ion_index.html

and I also want to know what partitioning would be suitable in my case to manage 300 clients. I may also use clustering in the future, as I know centos is good in clustering. my questions is about, number of partitions I may need to set up. I am asking this bcoz i do not want to reinstall the system in the future.

That is hard to say. You should accept that you are going to reinstall at some time anyway.

The problem is that we don't know your environment, we don't know what your 300 clients are going to be doing, and we don't know your environment (both political and technical).

And clustering does things differently than what a single system will do - there are MANY different kinds of clustering - disk? compute server? high availability?... We have no idea.

the main use of it is:
1) to assign folders to each of the members in the environment. So that they could back up their files n reports in their own folders on the server.

2) use application like sharepoint and microsoft project. this application is powerful in such a way that it analyzes the docs n reports and draws charts so that required number of resources and ... could be determined.

as the second choice may be accomplished with Ms applications, do you think it could also be done in Linux?
I hope we can come to a reasonable partitioning for installation.

#1 calls for some understanding of the users - 300 * how much storage do they each need?

I always recommend putting home directories on a separate disk, and in your case, I would suggest looking into a hardware raid solution to make it easier to repair when disks fail. Normally, the controller should fail over to using a hot spare, allowing you to replace the failed drive without downtime. Hardware raid allows you to do raid management external to the system, and again, usually not affecting the users other than during initial installation.

I think sharepoint uses (mandates) a MS database for storage, and does require a MS server.

I think jpollard has pretty much nailed it :)

Sounds like all the clients are MSWin, in which case you'll want the Samba pkg.
NB: user's are notoriously slack at backing up onto a separate drive. The best thing is to put their work areas on the Linux server ie Samba shares and make it VERY clear that anything on their local C: drive will not be backed-up.
You'll need your mgr to put that in writing because sooner or later someone will lose something important on their C: drive and blame you....

If you do want/need to consider backing up not just the server, but also some part of the client drives, see Amanda/Zmanda http://www.zmanda.com/quick-backup-setup.html.
You could just use that anyway, but in any case you MUST have a bullet-proof backup soln eg http://www.techrepublic.com/blog/10t...-utilities/895

how can I set up a hardware rate to externally control users?
and as far as it is concerned with MS applications, what are the advantages of setting up a linux server along a MS Server, so that for accessing MS application users may be redirected from the linux server to MS Server. this may encapsulate different tasks somehow. what are the other advantages?

Can you re-phrase; I've no idea what that means; sorry.

The point about MS Sw is that it will not run on Linux; the internal format of an executable is incompatible (& vice versa).
Therefore, for things that require an MS service eg Sharepoint, you'll need an MS Server to run it on.
The servers can share the data files via Samba, as can the users.

Really, you need to research what services you'll need to provide to your users and then tell us, so that we can provide more specific advice.

Yes, I will definitely ask for details and we'd discuss over the solution.
for the quoted statement, I asked how to externally store data on a raid disk. I've never done that before.

An external raid is treated as a disk drive. That raid controller can present the appearance of one disk, or many, it is up to the configuration as to what is shown. The host only sees disks, and can the partition those disks up however it wants.

Exactly how has to be determined from the vendor of the raid.

I am NOT fond of software raid (not very portable), nor of the fake raid controllers cheaply available (again, not very portable, and not very easily managed either - the one I tried using could only be managed from the BIOS level, and thus not very conducive to uptime requirements).

Certainly if you're going with HW RAID, get a good one; the extra money is worth it.
EG I worked at a place that had an HP MSA2000 series storage array, dual ported to 2 servers for redundancy if one server died; pretty solid kit.
System was required to be up 24 x 7 :)

EDIT: we also used HP DL380 servers; in both cases (hosts & array) the HW supported hot-swap of disks, a highly recommended technique for avoiding down time :)

http://skrypuch.com/raid/

It's time to understand with CPU power we have nowadays hardware RAID is not the best solution in most cases.

That isn't exactly accurate.

IT is true for "fake raid", and software implemented raid.

The problem with both is portability. You cannot move the software raid disks from platform to platform AND expect the data to survive. It is possible, but it depends entirely on platform to support EXACTLY the same format.

External raid controllers come with external support. You still can't move disks from one raid controller to another, unless the two controllers support the same format.

Another disadvantage for RAID in general is that once you start dealing with large disks (larger than 200GB) the recovery time from a failed disk can take several hours, to several days. This exposes you to multiple disk failures... and the need to implement more extensive error recovery (supporting multiple disk failures calls for multiple parity...) and still longer recovery times.

The advantage for external raid controllers here is that the recovery activity is OUTSIDE the host computer.

Another advantage external controllers have is that one controller can support multiple systems. Hence the use of fibre channel, and SANs.