Slow Deploy Speed.
-
Hello,
I’ve been using fog for awhile now and Mostly good however my school system got new laptops and I am having a hell of a time getthing them done. They are HP SB 215 G1 A6-1450 320GB. I get more than 8 on and the transfer speed slows to a crawl. I know they only have 10/100 Base T however should still be faster than 50MB/m - 120MB/M
I was doing some reading on the forms over the weekend and It looks like I may just need to change the kernel to match the laptops more correctly. Is there a listing that explains the difference between the different Kernels or is it more of experimentation till i found the correct one ?
PS. When you download a Kernel is that the Default or do you have to select the kernal you want to use. Trying to understand this section a little bit more
-
I tried to switch and Play around with Kernels today to see if that would help but to no avail still having troubles…
I have 300+ of these laptops to push into the school system this week and would love to do more than 16 at a time per hour.
any suggestions anyone?
-
Not exactly solving the problem of the slow speed, but could you do twice as many with an extra storage node?
I’ve found with multiple clients there is a point at which the slowdown really does ramp up. would be interesting if someone had FOG on SSD and could see if the same thing occurred.
-
I’m willing to bet it is a setting in a switch, are you using wired connections to transfer? Is there a possibility to use a GIG switch to image, say 7 of them at a time? If that works you can daisy chain another hub to give you more space.
Right now, if I was in your shoes, I would isolate FOG and do some further testing to make sure that is it not my network, or networking equipment.
-
Actually using a 2,000 dollar 48 Port Gig switch. and I can do say 10 at a time at 1.2 gb/min however it seems like the moment i add over that amount i loose a lot of speed. Its only been since ive been working on these laptops…
When i was Imaging 40 Lenovo SL510’s at a time was going at least 1.2 GB/Min
-
[quote=“VincentJ, post: 36878, member: 8935”]Not exactly solving the problem of the slow speed, but could you do twice as many with an extra storage node?
I’ve found with multiple clients there is a point at which the slowdown really does ramp up. would be interesting if someone had FOG on SSD and could see if the same thing occurred.[/quote]
Yea i am starting to think machines with only 10/100 ethernet may be my problem. When i was doing the SL510 they had 10/100/1000 ethernet s the speeds seemed good.
-
on machines with 10/100 I would expect 500MB/minute. On my gig machines I can push 5+GB per minute to a few at a time… but if I pile on more it can throw in the towel and really drag the speed down.
Also consider checking the disk IO statistics on your drives… HDDs aren’t good at random IO, and if you have enough sequential transfers at once, it becomes random.
-
Not to waste a post, I hate doing this. But I am in the same boat as Vincent here. My Gig equipment ( so long as I keep the 10/100 out of the way) I can push almost 5 GB/min and sometimes 6.5, but that’s very rare with NO ONE in my building but me.
I have some labs that have not been upgraded and they still sit at 10/100 and I only image them in sets of 10 (30 computers 3 sessions) to keep the load down. I also take my fog machine to this lab to take care of it. I know this isn’t the case in some installs. Just letting you know how I deal with the issue
Keep testing and let us know, we will do what we can to help you to fix the issue.
-
So i dumped using Multicast as the speeds where terrible and I’ve been using UDPCast and I’ve been able to push more at a time.
I am going to duplicate and add a node so i an have 2 different sets going at a time and see if i can push 32 at a time with 2 nodes using udp.
Edit: using a mysql client can i connect to the fog data base so i can export a more custom report…? if so what do i need to do
-
Just as a note; if you are using multicast for a 10/100 set of hosts from a gigabit server, I reckon you will get slow speeds as a result of a whole load of re-transmits. If you were to force the server to run at 10/100, I reckon you will get the same performance as you would for unicast.
-
Hello, after several tests on my FOG QNAP server, I noticed that it must reduce the Gzip compression level in fog settings\fog boot settings. Put 7 instead of default 9. The download speed and upload is much better. Example: 3 minutes to download a backup of a disk of 5500Mo for Windows XP or 15 minutes for a disk 42000Mo for Windows 7
-
your client CPUs low end? decompression is usually quite easy for CPUs to do fast. (faster than gigabit anyway)
have you looked at how much space you lose from doing that change?
-
Peu d’espace perdu, mais énormément de débit gagné. A vous de faire vos tests pour le vérifier
Little wasted space, but a lot of flow won. You make your tests to check -
[quote=“VincentJ, post: 36878, member: 8935”]Not exactly solving the problem of the slow speed, but could you do twice as many with an extra storage node?
I’ve found with multiple clients there is a point at which the slowdown really does ramp up. would be interesting if someone had FOG on SSD and could see if the same thing occurred.[/quote]
I’m running fog and imaging a bunch of HP t820 with 128G SSD drives. They are linux based so the up / down is raw not resizable. I can do on one system in about 15 minutes averaging 7G on the push down. I was able to push 16 systems last week and averaged 4.6G This week the 16 systems thru put drops to about 1G to all.
Similar thru put to traditional window systems but because nfts performance is better. Monitoring the eth0 now to see. I may have issues with the universities network getting udp traffic to and from other ips. -
is your server SSD as well?
the interesting bit would be imaging from a server with an SSD to allow multiple images to get to multiple clients from different parts of the disk at the same time.
-
[quote=“Steven B, post: 37172, member: 24174”]I’m running fog and imaging a bunch of HP t820 with 128G SSD drives. They are linux based so the up / down is raw not resizable. I can do on one system in about 15 minutes averaging 7G on the push down. I was able to push 16 systems last week and averaged 4.6G This week the 16 systems thru put drops to about 1G to all.
Similar thru put to traditional window systems but because nfts performance is better. Monitoring the eth0 now to see. I may have issues with the universities network getting udp traffic to and from other ips.[/quote]
No the server is not SSD. The ubuntu image has four VMs with networking. I have 3 SSD drives in the thin client. 16G windows 7 for horizon view 80G with dual boot windows 8 server 2008.
