What's new

Possible EON vs FreeNas comparison

  • SNBForums Code of Conduct

    SNBForums is a community for everyone, no matter what their level of experience.

    Please be tolerant and patient of others, especially newcomers. We are all here to share and learn!

    The rules are simple: Be patient, be nice, be helpful or be gone!

Got a chance to test out the latest EON version.

Along with test the new version I added another GB of RAM to my server. The Marvell driver detected and configured my network cards just fine. My first couple of file copies (same test files I used before) to the server averaged around 80-85 MB/sec with just a single drive which was a good improvement over my previous tests. After this I deleted the files I copied and tried writing them again and realized that now I was seeing the network spikes/stalls again. Did some read tests and those were about the same speed as before at around 60 MB/sec.

So all of this got me wondering what is causing the network stalls. I recall running into a similar problem using Win XP PRO on the server and coping files to a disk on the server that is much slower than the one on the client. For example I am blasting a file across the network at 100 MB/sec but the disk being written to can only write at 50 MB/sec. If the server only has around 512 MB of RAM I would most likely only be able to buffer a total of 400 MB before running out of memory. At that point the file transfer would stall until some of the buffered data could be written to the disk. Then once there was a bit of room the transfer would resume again. In some cases when the transfer resumes the OS throttles transfer speed to match the actual speed the data can be written to the disk. Other times the transfer resumes at max speed and then ends up stalling the file transfer again and again. My thought was maybe this was happening to me on EON.

I used the diskspeed.sh command to check disk speeds and that came back at 115 MB/sec for the disk in question. Then I decided to test actual read/write speeds using dd. I timed how long it took to complete this command dd if=/dev/zero of=/thing/test bs=1048576 count=1000. Which should write a 1000 MB file to my disk. Time was 26 seconds. 1000/26 = 38.46 MB/sec. Then I timed how long it took to complete this command dd if=/thing/test of=/dev/null. Which should read that same file. Time was 24 seconds. 1000/24 = 41.66 MB/sec. Now these tests were done after I noticed the lower file transfer speeds. During my first couple of file transfers I used "zpool iostat 1" to view write speeds and it was consistantly around 100 MB/sec. Based on these weird results I think that my disk controller might not be working correctly in EON. What do you think?

00Roush
 
Which eon was it (CIFS or Samba)

00Roush,

Thanks again for your time and effort and great detail. Did you use the CIFS or Samba version? I have a suspicion you saw this with CIFS? If am right would you mind trying with Samba and tell me if you observe the same behavior? You would'nt have to re-install just grab the iso and extract the /boot/x86.eon from the the samba package. Then place it on your current install in /mnt/eon0/boot/x86smb.eon. Then edit /mnt/eon0/boot/grub/menu.lst and duplicate the 3 lines for x86.eon changing it to x86smb.eon.

I don't think it's your controller as I've seen the behavior you described so accurately. I'm still trying to track down the intermittent pause go behavior with full throttle.
 
I was using the SAMBA 64 bit version. With a slimmed down smb.conf file similar to what I posted in post #16. The only difference was I removed the buffer sizes. To my knowledge these set the buffer size per connection and actually are reducing the buffer size compared to what is set by default for TCP connections.

Glad to help. I hope overall your original request in the thread was somewhat fulfilled.

00Roush
 
I've been watching this thread go by and you guys have done a ton of work.

Could one of you bottom line the results of the EON vs FreeNas comparison?
 
I did a performance comparison in post #38. Were you looking for that or more of an overall comparison?

00Roush
 
I did a performance comparison in post #38. Were you looking for that or more of an overall comparison?
I'm looking more for a bottom line summary. Basically read/write performance comparison, pros/cons of each and what your conclusion is for using one vs. the other.
 
Thanks

Thanks to all for their time and effort!! Especially 00Roush for all the hard work.
 

Support SNBForums w/ Amazon

If you'd like to support SNBForums, just use this link and buy anything on Amazon. Thanks!

Sign Up For SNBForums Daily Digest

Get an update of what's new every day delivered to your mailbox. Sign up here!
Top