XenServer - Fog VM Poor Upload Speeds
-
I think it’s an issue specific to Your VM. I run FOG in Hyper-V and I have really good upload speeds.
It sounds like your vm has only a 100 meg connection… Perhaps try different network adapters. Make sure your HDDs are set to SATA and not IDE.
-
Someone also mentioned to allocate at least 4 cores to it or performance suffers.
-
-
@FordST I can confirm that’s is the bandwith of a 100Mb connection. I have a 100Mb connection on our switchs and I have the same result.
-
This can’t be a 100MB vNIC issue, he receives 4-5GB write speeds from the Ubuntu (fog) server to his laptops. With the current setup.
-
@FordST Run bandwidth throughput test using a 3rd party tool to confirm your interface speed:
Let us know what you find.
-
Hey Guys,
Sorry for the delay, I decided to do a fresh install of Fedora on our XenServer.
Currently I am getting ~1.5GB/min upload.
I haven’t tested pulling an image as I haven’t migrated the images over.
I am going to do a throughput test like @Wayne-Workman suggested. I will update with results.
-
@Wayne-Workman We get 941 Mbits/sec when running a 1GB file through iperf…
-
I hate to ask the obvious, but are your xenserver tools on that VM up to date? XenServer does wonky networking things if the VM tools are out of date- and IIRC reverts to 100mbit if they’re not installed at all.
-
All up-to-date. I am pulling images at 7GB/min currently…
I just turned compression down to 0 and now we are uploading at 3-3.5GB/min, it peaked around 5GB/min at the start.
Anything I can do to up that speed?
It’s not crucial at this point… I was just upset about the 700-800MB/min range.
-
@FordST said:
All up-to-date. I am pulling images at 7GB/min currently…
I just turned compression down to 0 and now we are uploading at 3-3.5GB/min, it peaked around 5GB/min at the start.
Anything I can do to up that speed?
It’s not crucial at this point… I was just upset about the 700-800MB/min range.
The speed/compression depends CPU. If you have good cpu (multi core) on client, you can set a high compression. On poor CPU configuration set compression to 0. There is on wiki something about compression.
-
@ch3i Turning up the compression kills my upload speed. CPU performance doesn’t budge though… It sits under 10% usage. That’s another issue though…
-
-
@ch3i Ohhhhh ok, thanks for that, that makes sense.
So I should see peak upload speeds at 0 compression, correct?
-
-
@FordST said:
we are uploading at 3-3.5GB/min, it peaked around 5GB/min at the start.
Anything I can do to up that speed?With a Core i5 and 1Gbps network connection, that’s as good as it gets pretty much. Your performance is where it should be now.
Faster upload speeds don’t mean faster upload times, as @ch3i pointed out. It’s a balancing of compression and network speed.
-
@Wayne-Workman
Agreed, currently I am running on VM with dual quad xeon core running as fast as a vm most likely can. I am currently running uploads of about 3-5 GBmin and downloads of about 6gbmin. This is running across our WAN of fibre back bone and true gigabit switches. I mean downloading Windows 7 image in 7 minutes. I have not experienced anything faster even with very expensive commercial products.And correct me if I am wrong but quit some time ago the compression switched to the client side of things meaning, the server is not going to be the bottle neck but more the client processing the compression/decompression. So you can have a very fast server but if the client is running an economy processor then thing may become a bit slower.
-
@Tom-S said:
I mean downloading Windows 7 image in 7 minutes. I have not experienced anything faster even with very expensive commercial products.
+1,000,000 points to the @Developers Job well done.
-
@Tom-S said:
@Wayne-Workman
So you can have a very fast server but if the client is running an economy processor then thing may become a bit slower.True, but when you offload the CPU work, you not only free up the server’s CPU, but you free up the server’s bandwidth.
And, while one client with an economy processor might run slow, you can run significantly more concurrently.
