Server Fault is a question and answer site for system and network administrators. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Currently my setup is 1 main server which holds all the mp4 videos and 3 other servers which holds portions of this video. The process is: when a user uploads a video, it is stored into the main server, then synced (rsync) oh one of the 3 servers. So the main server acts as a storage server to sync these videos to their respective servers. The slave servers have 12x HDDs on them with raid 10 configuration. Maximum bandwidth peaks at 2GBps per each server on a dedicated 10GBps line. I believe I believe its the HDD bottleneck because my iowait goes up to 40% at peak hours. When the HDDs are set to 2048 readahead, Im able to maintain below 5% iowait.

I'm planning to replace the HDDs with SSDs but discard the raid setup. Instead, go with a bunch of SSDs for each server. Would this be better than my raid 10 setup? currently the HDDs are 7200 only. Another question, why is that setting the HDD read ahead to 2048 lowers the iowait so much and it doesnt affect the bandwidth throughout? it still averages the same or even better than read ahead at default 128k.

each slave server has similar/the same specs:

  • DUAL XEON E5 2603
  • 128GB RAM
  • 12x3TB HDD 7200 Raid 10
  • 10Gbps fiber optic NIC
  • 10Gbps dedicated port
share|improve this question

Keep in mind first that a 10Gbps link is gigaBITS, not gigaBYTES. 2-2.5 gigabytes per second transfer is pretty much maxed out for a bonded dual 10Gbps link. I'm going to continue answering the question assuming that this is not the case and you meant that you are not getting full throughput over the network.

You should be getting plenty of throughput on the slave server end with that many disks in raid 10. Make sure your hardware controller supports 3Gb/s for the drives you are using (you don't want to be limited to 1.5Gb/s or less) and that the interface for the controller is at its full bandwidth (I've seen PCIe slots restricted to x4 when the card supports x8).

You have a pretty good amount of RAM in each slave server. You might consider setting up a big ramdisk (75-100GB) and copying data to that as a test case. If you get similar transfer speeds, then the slave servers are not at fault; you need more read speed on the master server. A 4 or 6 SSD raid 10 in this server can help with that.

If you are bent on upgrading to SSDs, you will also want to ensure your controller will go up to 6GB/s per disk.

share|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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