Possible congestion??

[Johnb_2]

I run an 8 building PROCURVE dorm network at a local University. I have been getting some complaints lately about speed and such. I further investigated the situation to see MANY drops on up link interfaces. Now, these uplink interfaces are going from let say floor 2 down to floor 1. Instead of plugging them into gig uplinks they are plugged into fast ethernet due to lack of available open GIG interfaces. I think there maybe congestion due to the massive amount of traffic the students generate. I included an example of one of the dorms errors. From what I can see there is def not a duplex or speed issue on the uplink port.

[cenk sasmaztin]

hi
all upling(between switches) port must be same speed and same media type

for example
one switch on uplink port have 1000 base t full duplex ,other switch uplink port must be 1000 base t full duplex

[Johnb_2]

I agree on port to port connection speed. But if you read above, we ran out of open gig interfaces inside the IDF closet. So the question remains, are the frame drops due to congestion?

I know a possible solution is to add a 24 port gig switch and connect the uplinks into gig interfaces. But with IT and management, need has to be proven to some people. Thanks!

[Ralf Krause]

Hi John,

why don't you install ProCurve Manager plus (PCM+) - at least temporarily, it comes with a 30 dyas evaluation license.
You can download it at http://www.hp.com/rnd/software/network_management.htm


When you enable Traffic Sampling (what the tool does by default for all switch-to-switch links), you will receive very valuable sFlow (see: www.sflow.org) information which not only tells you about the link usage, but also who is talking with whom (communication matrix).

Hope that helps,
Ralf

[Johnb_2]

I would but you cannot sample 2650's or 2626's. From what I read they do not have that capability. The core switch which is a 5308xl has this capability, but I am not experiencing the errors on it's ports. Thank you though for the info.

[cenk sasmaztin]

hi Johnb please send me each sh tech print

[Johnb_2]

Here is one of them with errors

[cenk sasmaztin]

I can't see any warning or info log for int 49
no link mismatch speed duplex or another problem on interface .

drop creation on port(49) mainly cause flow udp protocol on port
udp not use unknowledgement mechanism in this way some ethernet packet drop on network.

[cenk sasmaztin]

this is not congestion
because occur for congestion excessive bandwith limitation please you see utilization rx and tx on port

Status and Counters - Port Counters for port 49


Name :

Link Status : Up

Totals (Since boot or last clear) :

Bytes Rx : 1,465,906,037 Bytes Tx : 129,169,364

Unicast Rx : 2,354,090,833 Unicast Tx : 1,939,163,641

Bcast/Mcast Rx : 54,371,859 Bcast/Mcast Tx : 82,256,277

Errors (Since boot or last clear) :

FCS Rx : 0 Drops Rx : 8629

Alignment Rx : 0 Collisions Tx : 0

Runts Rx : 0 Late Colln Tx : 0

Giants Rx : 0 Excessive Colln : 0

Total Rx Errors : 0 Deferred Tx : 0

Rates (5 minute weighted average) :

Total Rx (bps) : 13763560 Total Tx (bps) : 5732264

Unicast Rx (Pkts/sec) : 812 Unicast Tx (Pkts/sec) : 438

B/Mcast Rx (Pkts/sec) : 6 B/Mcast Tx (Pkts/sec) : 15

Utilization Rx : 01.37 % Utilization Tx : 00.57 %

[cenk sasmaztin]

my advice enable flow control on all uplink interface

for flow control information

http://en.wikipedia.org/wiki/Flow_control