no problem with iperf

I read the "Proving the Network is Not the Problem With iperf" at http://packetlife.net/blog/2011/feb/28/proving-network-not-problem-iperf/ and I am not sure I understand what he meant by "with an average rate of 939 Mbps... strongly suggest that the network is not at fault." Can you shed some light on this? Thanks
LVL 1
leblancAccountingAsked:
Who is Participating?
 
MiftaulConnect With a Mentor Commented:
Iperf tests the throughput of a network between two end points. It runs on the destination machine as iperf server and source iperf initiates the connection as client.

Here jeremy runs the test on the application server as iperf server mode on tcp port 8042 and later on udp port 8042. This is the port that runs on the application server that users client workstations use to access the application from client machines. With iperf test, jeremy proved that the network throughput to the application servers spcific port is 993Mbps,  and also that the port is open and not blocked by any intermediary network devices that may be present along the path from users client machine and the application server.

The iperf client sends packet load to the iperf server and results with minimum network throughput of 927Mbps and an average of 993Mbps. This confirms the network infrastructure supports the users workstations to access the application server with an average network throughput of 993mbps and as such the slowness in accessing the application is not due to network bandwidth. Rather the slowness could be caused by some other issue on the application software itself, may be the application is responding to the users query delayed for sone other software configuration.

Iperf confirms the slowness is not related to any layer3 - layer4 cause the network administrator is responsible for. It needs to be dealed by the experts dealing layer 7 like the developers.
0
 
gheistConnect With a Mentor Commented:
Use netalyzr and get enlightened there are other problems than bandwidth...
0
 
leblancAccountingAuthor Commented:
...with an average network throughput of 993mbps " If you have 1 gig port and you use more than 90%, isn't there an issue there?
0
Improve Your Query Performance Tuning

In this FREE six-day email course, you'll learn from Janis Griffin, Database Performance Evangelist. She'll teach 12 steps that you can use to optimize your queries as much as possible and see measurable results in your work. Get started today!

 
gheistConnect With a Mentor Commented:
Niope, gigabit port is full duplex. You can have 999Mbps both ways on same wire at same instant.
0
 
leblancAccountingAuthor Commented:
So do you mean a 999Mbps bandwidth on a 2 gigs?
0
 
gheistConnect With a Mentor Commented:
It has 1GBps each direction (say DSL at home has 10/1Mbps independent from eachother)
0
 
leblancAccountingAuthor Commented:
ok. So 1 gbps for each direction. So if my average bandwidth is 939mbps, shouldn't I be worried as it is almost 100% of my bandwidth capacity?
0
 
gheistConnect With a Mentor Commented:
Exactly
Small drops are expected, this is how qos notifies higher layers like TCP to slow down.
(you might want to have 10GbE there, but that comes over time)
0
 
leblancAccountingAuthor Commented:
So back to this statement, ""with an average rate of 939 Mbps... strongly suggest that the network is not at fault." It said that the network is not at fault. But I think it is because it utilized 90% of the link. Don't you agree?
0
 
gheistConnect With a Mentor Commented:
Nope
You are checking network BANDWIDTH only and that reaches the specs.
Since it sustains close to 1GBps, also you may believe cables in the direction 1Gbps was sustained are OK


Try this network quality test (you did not seem to get there without me posting the link):
http://netalyzr.icsi.berkeley.edu/
0
 
leblancAccountingAuthor Commented:
The thing is I cannot install any software on the client nor the server.
0
 
gheistConnect With a Mentor Commented:
Command line Netalyzr client needs just java...

It ananlyses latency, sporadic losses, filtering, PMTU, etc and points out many issues solvable by QoS or simple configuration adjustments (namely PMTU and missing DNS server cause long waits, 1000 packet buffer without QoS will behave similarily etc)
It is hard to model their tests with common network tools.
0
 
leblancAccountingAuthor Commented:
ok I will try it tomorrow
0
Question has a verified solution.

Are you are experiencing a similar issue? Get a personalized answer when you ask a related question.

Have a better answer? Share it in a comment.

All Courses

From novice to tech pro — start learning today.