Multicasting VERY VERY SLOW!

[J_Worth]

(I have posted the below under "Enterprise Software" as well - sorry if this is bad practice but the below is relevant in both!)

 

Hello everyone,

 

We currently are having issues with multicasting. We have SCCM 2012 installed and are imaging computers. Our problem is that when multicasting it takes ages to open a session and to download the image WIM file. Also, when PXE booting the computers once we have about 10 computers imaging it takes ages to contact DHCP for an IP address. Also, it will struggle to obtain an IP address when the computer has booted into WinPE - by struggle I mean it sometimes does not get an IP address before it times out and bombs out!. If we keep retrying it will eventually work - imaging is still very slow.

 

When we open Task Manager whilst the computer is downloading the image the network utilisation does not get anywhere above 1.50%!!!

 

Our network consists of Netgear GS724TP and GSM7224 switches. We have a GSM7224 switch as our core switch in the server room. We also have a couple of HP Procurve switches in the server rack that are part of our high availability server virtualisation setup.

 

Our network is flat - no vLANs. All switch links are connected at 1GB. We have checked the switches to the best of our knowledge and they all have IGMP enabled and spanning tree is set to RSTP.

 

I hope the above makes sense - we can only think that the issue is down to some settings on the switches.

 

Thanks in advance!

 

J.Worth.

[Cache]

I had this with plain WDS - I fixed it by disconnecting any unmanaged switches/hubs (because I haven't been able to get rid of all of them yet) which meant cutting off over half the network at the core, but might be worth a test if there are any on your network. Means I can't use multicasting most of the time, but atleast it got the bulky areas of computers done so that going room to room atleast the image download was staggard enough not to grind to a halt. Edited August 13, 2012 by Cache

[newpersn]

I cheated when I was reimaging one of our it rooms.

 

1gb 24 port switch. I plugged computers and the wds server (ex classroom computer) and link to network.

 

I took less than 10mins to image the computers.

[plexer]

Normally we would expect people to post their question once rather than cross posting a you used to do with newsgroups.

 

Ben

[glennda]

Sounds like you IGMP snooping enabled. Although if you have unmanaged switches you will be unable to enable it.

[J_Worth]

Normally we would expect people to post their question once rather than cross posting a you used to do with newsgroups.

 

Ben

 

Have removed the second post!!

[J_Worth]

Sounds like you IGMP snooping enabled. Although if you have unmanaged switches you will be unable to enable it.

 

We have enabled IGMP on all switches barr 2. We have two unmanaged switches that are in a different area to where we are imaging the machines (the link from the workstation back to the server does not pass through any unmanaged switches) - would this still cause us issues?

 

Does anyone know if there are any particular settings on the Netgear switches that need to be altered to get multicasting effectively? - besides enabling IGMP Snooping!

 

We were having issues with multicasting in one part of the school where the workstations go through a switch and then this switch is directly connected back to our core in the server room and it was still painfully slow!!!

 

An interesting thing that we have noticed is that we were imaging some workstations with only 100MB NICs in them and once they had finished the other machines with 1GB NICs began to speed up. This I am really confused - because how can imaging workstations with 100MB NICs in them slow the rest of the imaging process down so much.

 

I am assuming that it is ok for workstations to be connected to a POE switch - I can't see this causing any problems.

 

Thanks again.

[tmcd35]

Been having this issue with WDS myself. This thread prompted me to look further into is and I stumbled upon this...

 

WDS: Multicast kills network

 

Might be of some help.

[Cache]

WDS sets a master machine for the deployment speed, if it picks a 100MB machine as the master, then that is the speed it will send the image out. I spotted in our WDS server options yesterday there is the option to create multiple streams (so a slow/medium/fast or a slow/fast stream) which will group the machines together while sending the image.

 

We still had the issue regradless of whether the computers being imaged went through an unmanaged switch or not (we tested it by plugging the computer being imaged by multicast straight into the core as well, but it still flooded the rest of the network for some reason), the only way it worked properly was when every unmanaged switch was disconnected while the multicasting was taking place.

Edited August 14, 2012 by Cache