VLAN Over Utilization?

[garciael]

This morning users coming across several our WAN links complained of connection dropping to differnet servers on our hub site. We checked our WAN links and the FastE ports that connect to the Core Switch and found no trouble. But when pinging from our Core Switch to the FastE ports on the routers going to the WAN sites we got @ 850ms response time. We have our servers, switches and routers on VLAN95. We noticed a backup server that was just moved to VLAN95 was backing up server on different VLANs. We stopped the backup and things went back to normal. The Sys Admin reported that he was getting the most throughput ever on that server.

I suggested to create a VLAN just for Backups,but Sys Admin says that is a lot of work.

Question is could the VLAN95 really have been saturated by the backups of the server? I guess so since everything on that VLAN is basically on the same collison domain. If so is there a solution to this problem, maybe QOS or only using 3 of the 9 NIC cards on the server for backup?

Hope I'm making sense and thanks for your attention.

 

[KiscoKid]

A data backup can easily saturate WAN links and your suggestion to use something like QoS or perhaps traffic shaping is a good one.

If the backup server ONLY backups data and does no other function, then your QoS policy need only refer to that server's IP address(es) to effectively manage the bandwidth.

I hope the following URL is of some use:

http://www.cisco.com/en/US/products...figuration_guide_chapter09186a00800bd8f0.html

 

[garciael]

Apparently, the backup server is slowing down everyone else on this vlan99. But strangley enough when I look at the vlan interface I don't see high utilization or errors, except for the throttles keep incrementing. I can't figure out why if the Vlan is not being over-utilized why does everyone else on the Vlan slows down. We will try to run the back with just 3 of the 9 NICs on the server and see what happens.

 

[BuckWeet]

are they using multicast to do the backups?

 

[garciael]

not sure if they are using multicast, i wonder how do i check that? i noticed when people started complaining about the nwtrok being slow the sys admin for the backup server said he never got this much throughput and his job were getting completed faster. could it be that since there is no QOS the Core Switch pulling all its resources to accomondate the backup server and its 9 nics.