Ubuntu VM, running FOG latest SVN, Images on Raid5 = Very slow deployment

[Huggybearjr]

So I have finally rebuilt our FOG Server as follows.
i7 processor, 24GB Ram, 60GB SSD (Operating systems), 2TB drives in Raid5 =6TB, 2 Intel based Ethernet cards, 1 for internet and the other goes to our 1Gig 24port switch

System is Ubuntu 14.4 LTS running VirtualBox, Fog is running as VM, Ubuntu 12.4 LTS the virtual machine has 4 processors, 12GB Ram, 30GB VDI for operating system (Saved on SSD) and 2TB VDI for /images (Saved on the raid) Network is setup as 1 card = Internet through NAT and the other is FOG which is a bridged adapter.

2TB VDI is mounted successfully as /images, Network setting for FOG in the VM is set to static IP 10.10.3.49, 255.255.255.0 and 10.10.3.1 as Gateway. FOG SVN installed “successfully” with the static IP address, DHCP router address as 10.10.3.1 and DHCP service =yes DNS=no.

FOG works and we have sent a 3 partition image up (Around 80GB in total size) compression set to level 9, this took around 2.5 hours avg 2.5GB a min (started a lot faster) However deployment back down to 10 Dell laptops is much slower than before the rebuild, taking nearly 4 hours avg 500mb a min. Trying to understand why its much slower than our previous build (made by someone else no longer contactable “work related”) I am not sure where to start to fault find. Before it used to take around 1.5 hours to deploy 10 laptops with this kind of sized image. Any tips/help/save myself going greyer

Thanks

Matt

[Wayne Workman]

For starters, it’s not necessary to use 9 for compression. 6 is the default for a reason.
12 Gigs of ram is absurd for FOG - you’re probably only using 500MB of it, and that’s a high ball. Why not set it to 4 GB (overkill still) ?

Set your compression to 5 or 6 and then report back with what you find.

You might ALSO want to give this article a read through: https://wiki.fogproject.org/wiki/index.php/Image_Compression_Tests

Also, 10 clients deploying at once at roughly 500mb per minute means 0.0666666667 Gbps on EACH host, which is cumulatively 0.667Gbps.

I’m guessing you have the virtual server operating on a 1Gbps connection. I’m also going to assume that your bottle-neck in deployment time is your compression in combination with network overhead x10.

Please, if you can, as mentioned earlier, re-upload using compression set to 5 or 6. Then download to just ONE client, and let us know what sort of speeds you get.

Things get sketchy with network overhead and HDD seek times, along with processor overhead.

Everybody that uses fog regularly and frequents the forums can input on what they see on their setups for just ONE host. So, if you give us the data for just ONE host, then we can tell you if you’re under-performing, average, or over-performing. -respectively.

[Huggybearjr]

Tested today, sent an image up to FOG compressed at level 6, Image is 40GB C Drive and 40GB Recovery Partition. Starts off at 4.5GBpm then steadys to 2.5GBpm. For deployment of this image back down to a client again starts around 4GBpm then slows to around 2.5GBpm, added 3 more clients and still around 2.5GBpm, im not at work at the moment so trying to get my colleagues to test and send me the figures over. Does this speed seem normal?

[Wayne Workman]

@Huggybearjr For FOG 1.2.0, yes that seems about right. For FOG Trunk, this would be very slow.

[Huggybearjr]

Its FOG trunk, svn 4363. avg speed to deploy with 8 clients is 750mbpm. The network shows the network cards as been 1GB cards, I/O of the raid seems to be fine, our switch is fine and tested. So much slower than our last build, but can’t figure out why?

[Huggybearjr]

also to add the clients are dell m4800 / m4700 laptops with SSDs.

[Wayne Workman]

@Huggybearjr Multicast or Unicast?

[Huggybearjr]

Unicast, by ipxe booting first laptop then selecting “quick image” then selecting the image, then repeating until 10 clients are “quick imaging”

[Wayne Workman]

@Huggybearjr Is there a bottleneck anywhere between the 10 clients and the server?

A 100 meg link? a switch with a 100meg uplink ? Or even a bad fiber connection that has degraded to 100 meg?

What speeds do you get on a single client if you connect it to the same switch that the fog server is connected to?

[Huggybearjr]

Fog server network card 82540EM Intel Gigabyte controller, into 24 port switch in which all the clients are plugged into. This switch and setup hasnt changed from the last setup which worked fine. Changed the cables, still the same. Would it be worth doing a read / write test on the VDI thats on the raid?

[Huggybearjr]

We are a rental company, so just have a small isolated network been the switch to image machines back to a standard image before going out to the next client. So not trying to deploy over a School IT network or anything like that.

[Wayne Workman]

@Huggybearjr said:

Would it be worth doing a read / write test on the VDI thats on the raid?

Couldn’t hurt.

[Junkhacker]

is there a reason you’re running fog on Ubuntu 12.04, inside a virtualbox VM, on top of a Ubuntu 14.04 server, instead of running it right on the Ubuntu 14.04 or on a proper bare metal hypervisor if you want it in a VM?

[Wayne Workman]

@Junkhacker said:

is there a reason you’re running fog on Ubuntu 12.04, inside a virtualbox VM, on top of a Ubuntu 14.04 server, instead of running it right on the Ubuntu 14.04 or on a proper bare metal hypervisor if you want it in a VM?

12.04 is really stable with 1.2.0 out of the box. Maybe he doesn’t have a Windows Server or VMWare and wants the security of snapshots still? But, you do make a good point. If the VDI speeds are terrible with this setup, might be best to bite the bullet and install on bare-metal 12.04 with 1.2.0

[Junkhacker]

@Wayne-Workman I wasn’t questioning the use of Ubuntu 12.04. I actually still use it on my fog servers. my question was mostly regarding the rational behind the older-OS on VM inside newer-OS arrangement.