Network utilization and application performance

Unanswered Question
Mar 27th, 2007

Hi Forum,

How do I measure the utilization of my switches and routers. How do I know that they are over-utilized/congested?

My SAP users are claiming that traffic over the WAN link is slow, they use mercury loadrunner to perform the test, How do I cross-check that with my network performance? How do I measure application performance?

Thanks much,

py

I have this problem too.
0 votes
  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 3 (1 ratings)
Loading.

There are several things that you really need to watch to see if its your devices that are causing the slowness.

CPU

Memory

link utilization

physical characteristics of the WAN transport (like error rate, CIR, discard rates, DE packets, input and output queuing strategy/ performance)

You can spend as much or as little as you'd like to monitor these things.

"relative terms"

Free: MRTG

Inexpensive: Solar Winds Orion

Mid range: Ciscoworks LMS 2.6 (sizing option raises cost)

Larger cost: Concord

Pretty durn expensive: lots of choices here.

paulnigel Tue, 03/27/2007 - 19:32

Hi akemp,

I have MRTG, but how do I know it is congested, or the switch is overutilized?

say if the switch can take 10 packets/second, Does it mean that when i hit 4 packets/second, my switch is congested? Has every router/switch a maximum "traffic" rate so that I can measure against? sorry for this stupid question.

Thanks much,

py

paulnigel Tue, 03/27/2007 - 21:32

Hi forum,

When I do this: show buffers on my router, I get the below. From cisco site, it says that there shouldn't have 25 buffer misses occur in an hour.

if i look at the small buffer and divide that by the no. of days(hour), I get 33 buffer misses/hour. Is that consider ok? It seems to me that I do not have problems in the big/large/huge buffers.

Why I have lots of small buffer misses?

-------------------------------------------

Public buffer pools:

Small buffers, 104 bytes (total 64, permanent 50, peak 366 @ 3w6d):

28 in free list (20 min, 150 max allowed)

89662611 hits, 10994 misses, 26380 trims, 26394 created

307 failures (0 no memory)

Middle buffers, 600 bytes (total 25, permanent 25, peak 214 @ 7w0d):

19 in free list (10 min, 150 max allowed)

94384639 hits, 3816 misses, 10975 trims, 10975 created

1 failures (0 no memory)

Big buffers, 1536 bytes (total 50, permanent 50):

49 in free list (5 min, 150 max allowed)

11180863 hits, 0 misses, 0 trims, 0 created

0 failures (0 no memory)

Congestion is when a link or device gets saturated (ie when the limits of what it can do are exceeded by the demands of the clients/applications).

Different types of switches and blades have different maximum sustained throughputs. Some switch blades are designed in such a way that they can actually exceed the ability of the switch to carry all the traffic that can be generated (oversubscription)... a great example was the short lived 16 port gig blade for the 4006 when only the Sup II's were available.

Your switch becomes congested when the CPU and memory get to within 80-90% of theoretical maximum.

Yes every router and switch has a max rate at which they can process traffic. This can be limited by memory, CPU, and attached WAN circuits.

Actions

This Discussion