School server environment storage problem

The problem may not be the server, but the network itself. Is this a wired or wireless network (or both)? Is it 100Mbps or 1000Mbps?

Please describe the network environment, including the number of servers and clients.

Sounds like a very capable network. I agree that your current bottleneck seems to be the single server. I assume that the load spikes at the beginning of each class and the server just bogs down.

You may want to consider a second server to split the load. Even with a faster storage sub-system, you will still have a single server (faster drives would help).

I like the 15k HDD option (your original option 2), if you can maintain your needed capacity. If not (capacity-wise), I would expand with the SAS 7200rpm HDDs.

As fast as they are, I am not a huge fan of SSDs in a server environment. The cost of storage space needed can be prohibitive.

You said the storage has a Gb connection to the server?
So basically your servers' storage is a gigabit connected Nas - to me this looks like a huge bottleneck?
Who knows, maybe you've looked into it and the current drives cannot saturate this link.

I have seen major storage issues with ESXi servers and 7200rpm drives. It seems as if ESXi doesn't want to use all the drive bandwidth. I've seen the same with 1Gb networks. (I work in a software test lab with 5 ESXi test hosts, where the testers are working with backup software.) I've seen VMs set up on 8 drive RAID6 sets of 7200rpm drives have a transfer bandwidth of less than 80MB/s, and network transfers of single VMs get limited to 50MB/s. The limitations aren't the hardware, but something in ESXi.

However, many of my testers have gotten past this by running more VMs. Instead of testing with one VM backing up to our disk products at a time, they run 4 or 6 and are pushing 200MB/s for their tests. Perhaps your answer isn't in expanding the hardware, but running 4 VMs instead of 2.

There's several pieces here that are causing your problem I believe, it can't ONLY be narrowed down to drive speed.

First, you have three hundred or so end client computers, and ALL of these are concurrently requesting access to the same storage server. Split your gigabit connection into that many client computers and each one is crawling. Now of course I'm not saying they are all simultaneously pulling data through that server, but still you are talking about a ton of access having to go through a single gigabit channel.

Even if you do increase the throughput of your server storage, you're still limited by that gigabit connection. A single 7k SATA hard drive can saturate the throughput of a single gigabit ethernet channel, so while increasing the speed of your hard drives will give you greater overall data throughput at the hard drive, it's not increasing the bottleneck that's really causing an issue at your network interface. You should be looking at also increasing your network throughput for individual connections by either utilizing link aggregation and a network team of ports, or investing in 10 GbE technology.

While a group of 7k SAS hard drives isn't great performance, it's still pretty good depending upon the other underlying hardware (RAID controller, OS, RAID type, etc.) You may wish to consider separating out to additional storage servers (or NAS) to help distribute the load. Putting additional storage or faster storage into a single server can get you by, but that depends on your other underlying hardware again. The reason I might recommend an additional server instead of utilizing the same server for that kind of workload is because you're still running into a problem with bottlenecks. Even with more 7k or even faster 15k drives, you're still running everything through that single RAID controller, that single processor, that single network adapter, etc. If you set up two separate servers you have twice the flexibility and half the load across each server. This means you can also specialize your servers for their storage needs. For example, you can utilize your current storage server exactly how it is, but put in a second server with new faster 15k SAS drives for greater throughput and store shared video or database files on there instead, while having photos, documents, and other "lightweight" data that doesn't require as much throughput on your original server with the 7k SAS drives.

I would trouble-shoot the problem 1st - Here are things that I would do:
Test RANDOM access of your SAS array using IO meter,
You need over 100MB/s in RANDOM read

Assume you get 100MB/s random, and each video stream requires 1.5MB (DVD duality)
You are looking at about streaming 60+ clients - in theory but in real application you are lucky to have 40+, without ANY other computers access to the network

Like other posts, you need to know where is the bottle neck - IMO shot-gun is not a good approach

Work a bit crazy! could not check back in

Can you confirm the trxfer is Mbps not MBps - it's a 1 to 8 ratio between Mb and MB.
If your data is Mb <--- you do have a major issue in your network
I would plug direct (bypass) the router/switches further trouble-shoot
can you list your Router and Switched model?