Home / server / Questions / 205831
Accepted
GorillaApe
Asked: 2010-11-25 17:36:40 +0800 CST

Building a new server?

  • 772

I dont know if this is the right place to ask the question.

A friend of mine assigned me to create a web application for the company(warehouse) he works in order clients place orders etc.

The software is a COBOL program over 25 years old running on a 15 years old computer.

Because the computer has single disk and has a strange type of disk he takes backups at tape. There are about 400 cobol programs. However the computer now is ok.

But cause of the single disk and that the computer is old he wants to upgrade it. The 10 people who use the programs in the company connect via terminal from windows to the AIX server with cobol programs.

The computer still works great after 15 years with no problem!

The question after this long introduction is:... He want to buy a computer with high integrity that wont cause a damage..

However i am wondering if we should buy something like this: http://www-03.ibm.com/systems/x/hardware/tower/x3200m3/index.html

not because that the software is demanding( a 133 celeron would do the job i believe!) but because it has high quality parts..

I believe that it is extravagant to buy such a computer but i dont know if a custom rig with high quality parts (those that i can find) would be better.

Also we need a second pc for the webserver.. The uplink is 2mbps enough for the simple web interface and few concurrent clients i believe!

So to sum up...

What specifications are recommended? Also is it better to use a seperate pc for webserver and "cobol server" ?

hardware

3 Answers

  1. Best Answer

    If you want something with "high quality parts", then pretty much any off-the-shelf server-class computer will do.

    If you want something with real redundancy, then you might want to buy two servers and some shared storage, and run the OS's inside a virtual machine. This way if one server goes down, you can boot up the exact same OS on the 2nd server and continue working. If you pay mega bucks you can even get fault tolerance so that if a server fails, it's as if nothing happened at all.

    All that said though, a basic COBOL server and a basic Web server I would not bother with two seperate servers, but I would definately investigate your virtualisation options (for single-server deployments most are free) to see if they suit your needs.

    • 2

  2. Raid 10 is a better friend than Raid 1 or raid 0 (basically it is a combination of both. Stick rather RUN away from Raid 5

    There are a number of options - from using a SAN to just getting 2 decent servers and running raid 1 on both with DRDB (linbit.com)

    DRBD® refers to block devices designed as a building block to form high availability (HA) clusters. This is done by mirroring a whole block device via an assigned network. DRBD can be understood as network based raid-1. Best of all DRDB is OPEN-SOURCE DRDB is Network Raid 1

    For the ultimate in high availability you could combine with HeartBeat and have one system come online when the other fails.

    Always make sure to complete backups - AND send them offsite. Simple Rsync can help you here as well - if stuck and want a few scripts to help let the community know.

    • 1

  3. Electronics will fail and I think there's a definite ceiling to how much you can pay for the reasonable assurance that they won't immediatly. Beyond that ceiling your essentially throwing money away for a false promise.

    I think the best plan going forward revolves around putting in a good hardware lifecycle program and including appropriate hardware support contracts on any server you place into production.

    • 0