Iperf unMenu Package


59 posts in this topic Last Reply

Recommended Posts

Of course - 20Gb won't go down a 1Gb network ... but, perhaps, it indicates that my machines could make good use of faster interfaces?

As long as you are transferring into and out of memory, the minute you start actually writing to a disk its I/O speed will become the bottleneck ;)  10Gb hardware is out there and I've heard of people using Infiniband for IP networks but both are pretty expensive options currently.

 

These results also lead me to believe that the 50MB/sec maximum I see on writes to the cache drive are  being constrained by the disk drive.

Yep, there is some overhead created by CIFS or NFS, but odds are your drive speed is the limiter.  There's a post floating around somewhere where someone tested an SSD as cache drive and compared it to other rotational media.  Might want to track that down if you are interested in trying to boost your transfer speeds to the cache drive.  IIRC they got in the 70MB/s transfer range, and others have gotten close to that with rotational drives.  Don't quote me on that though, I may be wrong!

 

from my laptop running w7 enterprise 64 bit

 

------------------------------------------------------------

Server listening on TCP port 5001

TCP window size: 85.3 KByte (default)

------------------------------------------------------------

[  4] local 192.168.1.130 port 5001 connected with 192.168.1.2 port 56228

[ ID] Interval      Transfer    Bandwidth

[  4]  0.0- 1.0 sec  3115 KBytes  25516 Kbits/sec

[  4]  1.0- 2.0 sec  3416 KBytes  27985 Kbits/sec

[  4]  2.0- 3.0 sec  3453 KBytes  28288 Kbits/sec

[  4]  3.0- 4.0 sec  3456 KBytes  28312 Kbits/sec

[  4]  4.0- 5.0 sec  3056 KBytes  25039 Kbits/sec

[  4]  5.0- 6.0 sec  3216 KBytes  26342 Kbits/sec

[  4]  6.0- 7.0 sec  2552 KBytes  20903 Kbits/sec

[  4]  7.0- 8.0 sec  3312 KBytes  27134 Kbits/sec

[  4]  8.0- 9.0 sec  3224 KBytes  26412 Kbits/sec

[  4]  9.0-10.0 sec  3473 KBytes  28452 Kbits/sec

[  4]  0.0-10.1 sec  32512 KBytes  26308 Kbits/sec

:P

 

I smell wifi :o

Link to post
  • Replies 58
  • Created
  • Last Reply

Top Posters In This Topic

  • 4 weeks later...

Thanks for posting this! I've been wondering how I could test my network speed because data transfers seem to be slow.

 

I haven't set up a parity drive yet and writing seems to go at about 19MB/s and I've read from the unRaid at about 39MB/s.

 

When I run iperf, I get 935Mb/s, which is over 110MB/s:

iperf-2.0.5 peterchung$ iperf -c 192.168.1.3 -P 1 -i 1 -p 5001 -f k -t 10 -T 1

------------------------------------------------------------

Client connecting to 192.168.1.3, TCP port 5001

TCP window size: 65.0 KByte (default)

------------------------------------------------------------

[  3] local 192.168.1.2 port 55588 connected with 192.168.1.3 port 5001

[ ID] Interval      Transfer    Bandwidth

[  3]  0.0- 1.0 sec  114560 KBytes  938476 Kbits/sec

[  3]  1.0- 2.0 sec  114688 KBytes  939524 Kbits/sec

[  3]  2.0- 3.0 sec  114560 KBytes  938476 Kbits/sec

[  3]  3.0- 4.0 sec  114688 KBytes  939524 Kbits/sec

[  3]  4.0- 5.0 sec  114688 KBytes  939524 Kbits/sec

[  3]  5.0- 6.0 sec  113536 KBytes  930087 Kbits/sec

[  3]  6.0- 7.0 sec  112384 KBytes  920650 Kbits/sec

[  3]  7.0- 8.0 sec  113792 KBytes  932184 Kbits/sec

[  3]  8.0- 9.0 sec  114688 KBytes  939524 Kbits/sec

[  3]  9.0-10.0 sec  114560 KBytes  938476 Kbits/sec

[  3]  0.0-10.0 sec  1142400 KBytes  935723 Kbits/sec

 

When I was preclearing my drives, they were going between 70-110MB/s depending on the progress.

 

Any ideas as to what I can do to improve performance? I am wanting to use unRaid mainly to backup my massive video editing projects so performance is a big priority. My main computer is a Mac Pro if that factors into anything.

 

Thanks!

Link to post

Nice results.  Obviously not a bandwidth issue.  Curious - what was the source OS?

 

You should be getting >19MB/s without parity.  That is a normal value with parity (I get ~17-29MB/s, dependent on other network activity)

 

Note:  I initially read your post and missed the "without" part and was about to suggest a cache drive.

 

 

Link to post

I'm running unRAID Version 4.7. I've got 3 Hitachi 5K3000 2TB drives but only 2 assigned as data drives and the third idle and ready to be used as parity once the data is populated.

 

My Mac Pro is running OSX 10.6.7

 

My WinXP laptop only has 100Mbps ethernet so that only goes at 83.6Mb/s and only about 20Mb/s on wireless :(

Link to post

[sOLVED] I had to set my share as NFS and connect the share as NFS. Now it's writing at around 50MB/s and reading at around 74MB/s... much faster! :)

 

While I'm glad I am getting my unRaid up to speed, so to speak, I am having to read a lot to find out that there are different connection types: NFS, AFP, SMB, etc... I don't know what any of it means and just follow what works for others. I'm glad others have posted their solutions, though!

 

Just for reference, to set up your shares for NFS:

  • Go to the unRaid web interface and go to Shares:
  • In the shares you want to access via NFS, in each "Export (NFS):" input box, type: "*(rw,insecure)" then click "Apply"
  • On your Mac, open "Disk Utility" and go to "File->NFS Mounts..."
  • Click the plus sign and under "Remote NFS URL:" type in "nfs://IPaddress/mnt/user/ShareName" and under "Mount location:" type "/Volumes/ShareName" making sure that you type the same ShareName for each input. For some reason, it doesn't like me putting in "tower" and only finds the NFS with the IP address. I know it will create problems when I restart my switch and router and it assignes a different IP to the unRaid :( Any solutions for this?
  • Click Verify and you should get confirmation that your Mac can connect to the NFS
  • In Finder, click "Go->Connect to Server..." and type in "nfs://IPaddress/mnt/user/ShareName" exactly as you did in Disk Utility
  • And voila! You are now connected via NFS and should experience faster data transfers!

 

Also, for my own reference, it looks like AFP might be better but you have to wait for version 5 of unRaid as it is currently in beta right now and not recommended for production use (unRaid 5beta, that is, not AFP, to clear up any confusion)

 

Hope my headaches are helping others!

Link to post

If you don't want to mess with setting up your shares as NFS, there is a simpler "hack" I came across: just enable "Internet Sharing." You can even turn off your AirPort after setting up Internet Sharing through your AirPort.

 

I was getting over 40MB/s both reading and writing just by enabling "Internet Sharing" and connecting to smb://tower/ShareName

Link to post

If you don't want to mess with setting up your shares as NFS, there is a simpler "hack" I came across: just enable "Internet Sharing." You can even turn off your AirPort after setting up Internet Sharing through your AirPort.

 

I was getting over 40MB/s both reading and writing just by enabling "Internet Sharing" and connecting to smb://tower/ShareName

 

Interesting.  Sounds like protocol overhead or implementation to me.  I just made a test copy of of my array and I'm getting 60MB/s + sustained transfer writing to an SSD on my Win7 workstation.  I tried copying from a WD green drive and a 7200 RPM drive and got about the same results.

 

OSX may be using the SMB protocol rather than the CIFS protocol.  You could be losing some transfer speed to the overhead in the protocol.  That kind of makes sense because you are getting much faster reads/writes using NFS.  NFS is a Unix derived file system and OSX does indeed play nice with it.  That's the route I would go sense you are running a Mac :)

 

You are correct that AFP may be a good solution for you if when V5 stabilizes and gets out of beta.

 

Here are some links if you want to read up on file systems:

SMB/CIFS

NFS

AFP

A bunch more that you've never heard of

 

Glad the iperf package helped you out ;D

Link to post

Not sure what it takes for OSX to resolve names via NETBIOS, it's doable but I have no idea where to even begin but someone else here may be able to point you down the right track.  In the mean time running it with a static IP or setting up a DHCP reservation on the DHCP server would be advisable.  Then you could use the host file to map the name to the IP address.  The instructions for the host file are in this article.  Should be /etc/hosts on a mac...

Link to post

TCP window size: 8.00 KByte (default)

------------------------------------------------------------

[344] local 192.168.0.247 port 14043 connected with 192.168.0.149 port 5001

[ ID] Interval      Transfer    Bandwidth

[344]  0.0- 1.0 sec  53480 KBytes  53480 KBytes/sec

[344]  1.0- 2.0 sec  56656 KBytes  56656 KBytes/sec

[344]  2.0- 3.0 sec  56000 KBytes  56000 KBytes/sec

[344]  3.0- 4.0 sec  56320 KBytes  56320 KBytes/sec

[344]  4.0- 5.0 sec  55720 KBytes  55720 KBytes/sec

[344]  5.0- 6.0 sec  56712 KBytes  56712 KBytes/sec

[344]  6.0- 7.0 sec  56288 KBytes  56288 KBytes/sec

[344]  7.0- 8.0 sec  56280 KBytes  56280 KBytes/sec

[344]  8.0- 9.0 sec  56752 KBytes  56752 KBytes/sec

[344]  9.0-10.0 sec  57424 KBytes  57424 KBytes/sec

[344]  0.0-10.0 sec  561640 KBytes  56079 KBytes/sec

Done.

 

Windows 7 64bit machine G-Bit lan with about 15mtr Cat5e to Gbit router,, , The Unraid machine is connected directly to this router with a 2 mtr Cat5e patchcable.

I have a feeling that this could be better, reallife writing to the unraid server gives 18 tot 22Mb/sec , reading 38 tot 42Mb/sec

 

a few weeks ago i connected 2 single disks windows 7 machines instead of the Unraid machine and got a 70000 to 90000Kbytes/sec troughput

 

Is this because of the difference in protocols and overhead, or the parity drive slowing the speeds down?

 

 

Link to post

Is this because of the difference in protocols and overhead, or the parity drive slowing the speeds down?

Not the parity drive, but the process of calculating parity.  It is slowed by the slower rotational speed of EITHER/BOTH the data drive and the parity drive.

 

Both the data and parity drive must first read the sector being written, then write them after waiting a full rotation of their disk platters.  The rotation is what slows the process.  7200RPM disks will rotate 1/3 faster than 5400 RPM disks, and users can get close to 35MB/s.  If a 5400 RPM drive is involved, your speeds are very typical.

 

Joe L.

Link to post

What Joe said :) I wil add in that iperf is only measuring raw bandwidth. Any disk activity will slow things down. That said your numbers are very low. 55 KB/s is < 1 Mbps; based on your results I think there is definitely something wrong. I would start with the cables and/or try diff machines to try to isolate the problem.

Link to post

Both Joe and Spectrum thnx for the tips and explanation,

 

My main desktop is a X58 mainbord , i7 920 , 6GB DDR3 and sata 2 7200 rpm disk in combination with a onboard Gbit Nic (Realtek RTL8168C(P)/8111C(P) Family PCI-E Gigabit Ethernet NIC (NDIS 6.20)

 

It's connected to the G-bit router with a 2mtr Cat5E patchcable to Wall-socket > 8 to 10mtr's cat5E in the wall > out of the wall in garage to the g-bit router.

 

I will try a direct cable, without the wall-socket and try another desktop.

 

will post results

 

 

Link to post

;D ;D ;D, No it's not bad,, i only noticed the big difference writing speeds with a single disk windows environment. i thought something was wrong,, but now realising that it's not so bad at all with a WD-Green as a parity drive.

 

did the test,, Win7 pc directly to G-bit router with patchcable of 2 mtr, and UNRAID server directly with patchcable of 2mtr ,, i did noticed some improvement in write speeds, but not much.

 

i had to pump over 3TB of data and that took ages ,but even SATA to SATA would take long,  now that's done.

 

 

 

Link to post
  • 1 year later...
  • 1 year later...
  • 3 weeks later...

I'll second that request. I scoured the Interwebs looking for that package, and even tried to compile my own, but without success. I have no issues doing it on my Ubuntu machine, but I think I would need C++ on my unRAID server and I'm not sure if that's a good idea.

 

Does anyone have a copy of the iPerf 2.05 package for unRAID left over they would care to post? Please?

Link to post

I'll second that request. I scoured the Interwebs looking for that package, and even tried to compile my own, but without success. I have no issues doing it on my Ubuntu machine, but I think I would need C++ on my unRAID server and I'm not sure if that's a good idea.

 

Does anyone have a copy of the iPerf 2.05 package for unRAID left over they would care to post? Please?

Does this work for you?
Link to post

I see I have the package on my flash drive here's a link to the unMENU package: https://dl.dropboxusercontent.com/u/189963/iperf-2.0.5.zip

 

In the zip file consists of two files, iperf-2.0.5-i486-1.tgz and iperf-unmenu-package.conf

 

If I understand the process correctly, one would extract the two files in place them in the /boot/packages  folder. I think, then from unMENU you would click on the "Select iperf-2.0.5-i486-1.tgz" button. It should be seen as local and install. However, it might be a good idea to restart unMENU before trying to select it.

 

I hope this helps! It works on my unRAID server; I'm running 5.0.4.

Link to post

I tried to install the package on my V6b5a test server  and ran into problems. I did notice the file iperf-unmenu-package.conf was there already in the /etc/packages folder, but iperf is not being listed in unMENU's web page. So there's more to it. But, I guess for V6  one could run a VM, etc. but that's adding too much overhead to get a believable number.

Link to post

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.