WOL Creates network Bottleneck?

[ITGURU]

I have a healthy 10GB backbone network with 1GB to each PC, and ping times are <1ms.

If I run a WOL command to switch all PCs on remotely, monitoring software reports that network components (switches, printers etc) become unreachable.

Running a constant ping against a remote IP address clarifies this as response times are <1 ms, then whilst the WOL task in running times fluctuate upto 48ms and intermittent 'request timed out' responses.

Once all PCs have booted ping times go back to responding every time and back to <1ms.

 

Is this normal??

[Jamman960]

Does this happen if you do it while the pc's are already on? Or just when actually waking a large number of pcs?

[ITGURU]

Does this happen if you do it while the pc's are already on? Or just when actually waking a large number of pcs?

 

If they're already on pc's ping as normal, says they are online and that's it. Usually just when dong a large wake on I get the issue.

[ass17]

I never flood the network with WOL for this reason. I set my computers in to groups and schedule on times 5-10mins apart. And I send two WOL requests per PC just in case the first never reaches. We're on a 1GB backbone.

[Jamman960]

How many PCs are we talking about? I wonder if general boot up traffic is causing issues, if they're all talking to ad/dhcp/file servers at the exact same time at 1gbps you may end up having to re-transmit a lot, having the workstations operate at 100mbit may help(probably not worth it if not causing issues at other times) or staggering the wake up process slightly

 

I also find that ping replies from switches, printers and other devices with basic cpu's tend to be misleading as network performance indicators - they tend to suffer from degraded ping times with traffic spikes fairly easily.

 

I would try a continuous ping to a remote workstation rather than a switch/printer, if you see fluctuations there it may be worth using wire shark to look for anomalies.

[forkies]

Interesting, we use WOL, 1gb network but only WOL 32 PCs at a time. Never seen any issues. I did have an issue before with our Impero server sending WOL packets to 255.255.255.255 about 5 per second continuously for 5 mins with a 5 min break or something before repeating. Turns out one of our rooms does not support WOL (on my list to investigate) and there was a policy in Impero to wake the room at 7am but as the PCs didn't wake it constantly sent them. I spoke to Impero and they put me in the direction of the policy causing the broadcast. However this did not decrease network performance.

[ITGURU]

I'll just have to limit the number I wake at a time then to avoid the issue.

[FN-GM]

Is this 1 flat network you have? Or is it Vlanned off?

[ITGURU]

Is this 1 flat network you have? Or is it Vlanned off?

 

All workstations are on the same vlan. I have other vlans for phones door access cctv and they are not affected during the WOL request.

[FN-GM]

All workstations are on the same vlan. I have other vlans for phones door access cctv and they are not affected during the WOL request.

 

If you chunk up your workstations into different VLANS and send the WOL using a directed broadcast it will help loads. How many workstations do you have? The recommendation is not to span your workstation VLANS over multiple switches / stacks. So have a workstation VLAN for each stack or switch depending on your equipment.

[Sheridan]

We run a wol every morning to wake up all PCs, but with a simple one second delay between each command, never had any issues with bottlenecks.

[afc123]

Sorry to hijack but do you use a script or special software to WOL all your PC's? Having a mare with some 64 bit PC'S at the moment

[Sheridan]

We use a home brew program written in VB, which uses winsock to fire out a wol packet

[disk]

Maybe check for OS/firmware updates for your switches? Check the release notes, maybe something there about improving/fixing this issue?