Super User is a question and answer site for computer enthusiasts and power users. 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

Background: I'm building a quiet HTPC + NAS that is also supposed to be used for general computer usage. I'm so far generally happy with things, it was just that I was expecting a little better IO performance. I have no clue if my expectations are unreal. The NAS is there as a general purpose file storage and as a media server for XBMC and other devices. ZFS is a requirement.

Question: Where is my bottle-neck, and is there anything I can do config wise, to improve my performance? I'm thinking VM-disk settings could be something but I really have no idea where to go since I'm neither experienced with FreeNAS nor VMware-WS.

Tests: When I'm on the host OS and copy files (from the SSD) to the CIFS share, I get around 30 Mbytes/sec read and write. When I'm on my laptop laptop, wired to the network, I get about the same specs. The test I've done are with a 16 GB ISO, and with about 200 MB of RARs and I've tried avoiding the RAM-cache by reading different files than the ones I'm writing (> 10 GB). It feels like having less CPU cores is a lot more efficient, since the resource manager in Windows reports less CPU-usage. With 4 cores in VMware, CPU usage was 50-80%, with 1 core it was 25-60%.

EDIT:

HD ActiveTime was quite high on SSD so I moved the page file, disabled hibernate and enabled Win DiskCache both on SSD and RAID. This resulted in no real performance difference for one file, but if i transferred 2 files the total speed went up to 50 Mbytes/s vs ~40. The ActiveTime avg also went down a lot (to ~20%) but has now higher bursts. DiskIO is on ~ 30-35 Mbytes/s avgs, with ~100Mb bursts. Network is on 200-250Mbits/s with ~45 active TCP connections.

Hardware

  • Asus F2A85-M Pro
  • A10-5700
  • 16GB DDR3 1600
  • OCZ Vertex 2 128GB SSD
  • 2x Generic 1tb 7200 RPM drives as RAID0 (in win7)
  • Intel Gigabit Desktop CT

Software

  • Host OS: Win7 (SSD)
  • VMware Worksation 9 (SSD)
  • FreeNAS 8.3 VM (20GB VDisk on SSD)
    • CPU: I've tried 1, 2 and 4 cores.
    • Virtualisation engine, Preferred mode: Automatic
    • 10,24Gb ram
    • 50Gb SCSI VDisk on the RAID0, VDisk is formatted as ZFS and exposed through CIFS through FreeNAS.
    • NIC Bridge, Replicate physical network state

Below are two typical process print-outs while I'm transfering one file to the CIFS share.

last pid:  2707;  load averages:  0.60,  0.43,  0.24  up 0+00:07:05    00:34:26
32 processes:  2 running, 30 sleeping
Mem: 101M Active, 53M Inact, 1620M Wired, 2188K Cache, 149M Buf, 8117M Free
Swap: 4096M Total, 4096M Free

  PID USERNAME    THR PRI NICE   SIZE    RES STATE    TIME   WCPU COMMAND
 2640 root          1 102    0 50164K 10364K RUN      0:25 25.98% smbd
 1897 root          6  44    0   168M 74808K uwait    0:02  0.00% python

last pid:  2746;  load averages:  0.93,  0.60,  0.33  up 0+00:08:53    00:36:14
33 processes:  2 running, 31 sleeping
Mem: 101M Active, 53M Inact, 4722M Wired, 2188K Cache, 152M Buf, 5015M Free
Swap: 4096M Total, 4096M Free

  PID USERNAME    THR PRI NICE   SIZE    RES STATE    TIME   WCPU COMMAND
 2640 root          1  76    0 50164K 10364K RUN      0:52 16.99% smbd
 1897 root          6  44    0   168M 74816K uwait    0:02  0.00% python

I'm sorry if my question isn't phrased right, I'm really bad at these kind of things, and it is the first time I post here at SU. I also appreciate any other suggestions to something, I could have missed.

share|improve this question
    
Is that software/BIOS RAID-0? How much I/O is occurring on the host OS hard disks, the RAID0 volume? RAID in software (or BIOS) is terrible; you shouldn't use RAID without a hardware RAID controller, pretty much end of story. It can cause mostly invisible CPU usage and I/O storms on the host that bog down the entire system. – allquixotic Nov 30 '12 at 17:16
    
Copying data from the SSD to the RAID drive (NoVM) goes at around 150+ Mbytes/s with CPU usage going up about 5%, I have a hardware raid I think (how do I check?). I've set the SATA ports to RAID in the BIOS and I've set the RAID defs when it loads the controllers pre win. As far as the VM is concerned I think it sees the RAID as a really fast disk. Copying data from host to CIFS results in a ~35Mbyte/s SSD read and a ~30Mbyte RAID write. Not sure if it helped. – maka Nov 30 '12 at 17:47
    
If you don't know for a fact that you have specifically purchased a Hardware RAID card (for example an Adaptec 6405E), you aren't using hardware RAID. BIOS RAID is just as bad as software RAID. The CPU is used for coordinating the striping. – allquixotic Nov 30 '12 at 17:50
    
Is this related specifically for VM-ware IO? Since I get more than 150Mbytes/s in windows SSD-->RAID0 with almost no CPU usage and I'm happy with the performance there, but maybe VMWare can't handle the type of raid that I'm using? – maka Dec 1 '12 at 19:18

The bottleneck could be several things:

  • VM and Data is stored in the same ssd
  • in bridged mode -> router could be the bottle neck too

Check These links:

http://www.seancull.co.uk/public/seancull.nsf/dx/is-your-vmware-workstation-network-really-slow-solved-.htm

http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=2019058

share|improve this answer
    
I'll check this out and update how it goes. Thank you. – maka Nov 30 '12 at 17:48

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .