Take the 2-minute tour ×
Server Fault is a question and answer site for system and network administrators. It's 100% free, no registration required.

We have to buy a number of servers in my organisation - and after some thinking about blade vs rack, we rather want to go in direction rack - mostly due to cost considerations. Our organisation is a university with about 3000 - 4000 Students - and we are planning to offer E-Learning Services (including video streaming). For that purpose we want to start with 5 servers - and virtualise everything. Our Budget is about 50,000 Dollar. We are calculating with 200 concurrent sessions - it might be less in the beginning - but it should grow, once our services become well-developed and known.

Our current idea is to buy 3 rather strong servers that will host our e-learning environment as well as streaming. Through virtualisation we are able to relocate a server to a different machine, in case of any trouble with one machine - of course loosing some performance - but at least continuing with our services.

Further we want to have 1 Server for backups, which would be the slowest server - and one for files - that would be a bit faster than the backup server - but not super-fast.

We are looking at SAS Drives for all servers but the backup server.

So all together we have the following plan:

  • Web Server 128 GB Faster CPU 2 12 TB SAS
  • Web Server 128 GB Faster CPU 2 12 TB SAS
  • Multimedia Server 128 GB Fastest CPU 2 12 TB SAS
  • Backup Server 16 GB Slowest CPU 1 36 TB SATA (needs extension)
  • File Server 32 GB Okay-CPU 1 12 TB SAS

The faster CPUs should be 8 core. Of course we are using not only HMTL-Websites but content management systems (DRUPAL and also JAVA-Applications, Flash-Streaming, etc.), that eat up some good CPU and RAM. But expect the servers to be super fast for our purpose - and strong enough to keep up with our growing demands for some time.

Offsite Backup we want to do with tape - as another storage unit would be just too expensive...

We want to use RAID 1 - mirroring - so we will lose 50% space - but the system should be very reliable.

Also power supply, cooling and network cards ought to be redundant.

In terms of RAM we are looking at Quad Rack RDIMM (8GB each)

In terms of CPU we look at the Intel 26 Family, by going in the 8 core-ranks for the active server (2965 for multimedia-server - and similar CPU for web-server) and staying with the 4 core 2609 for file and backup server.

No our disaster recovery strategy is as follows:

We cannot have full redundancy due to cost considerations, so we understand our strategy as follows:

if file-server goes down, we can switch to backup server. if backup server goes down, we have tapes - and we are also planning another offsite backup at the other data centre. So backup is there.

if a one of the three musketeers goes down, they may replace each other. The virtual machine can go to the right server - and immediately be up. So we should be able to operate with short breaks (copying virtual machine) as long as we have one server running. Of course that well hit our performance very heavily - but in any case, we can do our work.

One big headache is the storage. SAN is very expensive - so we are now going for onboard hard drives and storage extension shelves. That is not ideal - but also gives us additional redundancy and independence from a unified storage device.

But my questions are for recommendations from experienced admins who have built up data centres:

a) Recommendations for CPU - does it make sense to go up in the ranks of the INTEL 26 Family - where it gets very costly - or shall we rather stay on the level where 8 core starts... Maybe for multimedia, ast CPU is required, as the server also shall be able to transform video-files into flash-files for streaming.

b) Recommendations for storage: What storage options should we consider other than onboard storage?

c) Any general recommendations/ideas, that you get, when you look at our specs?

d) Does our backup/reliability strategy make sense for you? Any recommendations?

e) We require systems, that work with opensource virtualisation like XEN, so we look in the direction of IBM. But they are costly. Experiences with DELL or HP?

I know, this are many questions in one, if you have some valuable tips and hints, do not hesitate to just answer part of it...

thanks and best regards Christian

share|improve this question

closed as not constructive by Magellan, Ward, David Schwartz, Iain Dec 3 '12 at 7:53

As it currently stands, this question is not a good fit for our Q&A format. We expect answers to be supported by facts, references, or expertise, but this question will likely solicit debate, arguments, polling, or extended discussion. If you feel that this question can be improved and possibly reopened, visit the help center for guidance.If this question can be reworded to fit the rules in the help center, please edit the question.

1 Answer 1

a) Expand horizontally if you can, not vertically. That is, get a lot of cost efficient CPU/servers, not stupidly expensive ones that make little sense cost wise unless your application can't expand horizontally. This way, you can theoretically expand infinitely in a cost effective mean rather than having a hard ceiling that's very expensive.

b) Opting for Raid 10, 50 and 60 (depending on importance of fault tolerance, read/write ratio behavior) with big beefy 7.2k drives will be a much better solution than raid 1 of SAS (10k/15k) drives given your scenario. If one server is too overloaded, get another server, not higher performing disk. For video streaming, 15K drives are not significantly better than 7.2k drives anyway, they're more of a response time improvement.

c) See A & B

d) See A & B

e) That would be a shopping question and frankly they make no difference... You also have more options than 3 manufacturers.

share|improve this answer

Not the answer you're looking for? Browse other questions tagged or ask your own question.