test which pc has fastest access to server

[RATiO]

I have two PCs plugged into different switches and would like to know which one has the fastest connection to a server at the other side of the building.

A ping test is not very helpful as both are <1ms

Is there software I can run on the two desktops to test this - without having to install or run anything on the server?

[StrikedOut]

I would just use a large file and transfer it from the server to the PC taking it in turns. Run it a few times and look at the trend in the results. That should give you an idea.

[iuerg87yerg879e0rg9erugjer]

if both servers are being pinged at less then 1ms then i doubt its going to make any difference which one you chose.

[+BudMan MVC]

Why would you think one should have faster access than the other?

[Solidstinch]

Use iperf

http://sourceforge.net/projects/iperf/

[RATiO]

Why would you think one should have faster access than the other?

One is plugged into a 'core switch' with CAT 5E cable. This switch has a fibre link to the other switch where the server is plugged into.

The other is plugged into another switch which is plugged into the core switch mentioned above.

Should both have same speed or will the one plugged into the core switch be faster?

[+BudMan MVC]

And is the link to the core switch from the downstream switch to where pc2 is plugged in, is this link saturated? Is it say only a 100mbit connection? And pc1 has a 1000 connection to the core switch?

Just because you have extra switch between does not mean your going to see any difference in speed to the server.. The extra connection is going to add maybe a couple of ns to the latency. And unless there is a bottle neck between the pc and the server that one would see that the other won't there is going to be no real difference.

How many other machines are on that downstream switch connected to the pc2? Is is full, what is its connection to the core switch? What are the speeds of the clients connected to this switch. Your prob not going to want to connected all devices at 1000 if the switch only has a gig pipe to the core switch.. Maybe set the clients down to 100 so 1 client can not saturate the link to the core, etc. Or run fiber or bond a few connections together to the core, etc.

btw Linux ping reports less than ms

64 bytes from i5-w7.local.lan (192.168.1.100): icmp_req=10 ttl=128 time=0.393 ms

64 bytes from i5-w7.local.lan (192.168.1.100): icmp_req=11 ttl=128 time=0.426 ms

64 bytes from i5-w7.local.lan (192.168.1.100): icmp_req=12 ttl=128 time=0.419 ms

^C

--- i5-w7.local.lan ping statistics ---

12 packets transmitted, 12 received, 0% packet loss, time 11020ms

rtt min/avg/max/mdev = 0.350/0.451/0.553/0.072 ms

so you could always just boot a live CD at these machines to see the difference in latency time, lan speeds on decent.. I would guess difference between your connections unless something is hung up to be only a few ns.

[Original Poster]

just ping it like everyone has said .... if you are using windows based then -t it for a constant ping to what ever server to also see if any packets are being lost.... dont forget its not the speed of the connection its the reliability (but obv speed is a plus ;D )

[+BudMan MVC]

his problem is windows ping does not show values below 1ms, so he will not be able to see very small ns differences in latency to the server.

And ping wouldn't be a test of actual speed anyway, since he would see the same ping times be it he was on a 10mbit, 100 or even 1000 gig connection.

The very very very small difference in latency being an extra switch between is not going to make for any real difference in speed. Now line be full between pcX and pcY could make for difference. Connection speeds 10/100/1000 would really make difference in moving of data (speed)..

But if there is no network bottleneck between pc on the downstream switch, and same bandwidth available to the core as to the server off the core - where you have a common path now for both machines. Their speeds is not going to be any different.

If you want to test speed, use something like iperf - which will just test the network and not use the disks of the machien which could put a hit on the test, etc. Now keep in mind that power of the machines, the nics on them, the network settings, etc. could cause different speeds to be shown for these machines that have little to do with the network path.

Your best overall test would prob be to copy a large file down from server on each machine at different times. So that they are not conflicting, and nothing else really going on in the network when you do this test.

But again generally speaking on a lan, all things being equal - adding an extra switch would not cause any sort of slow down in speed. Your only talking a few nanoseconds at worse case.. Unless something is wrong, or downstream switch is overloaded, or connect from that switch to core switch is saturated, etc.

[Original Poster]

his problem is windows ping does not show values below 1ms, so he will not be able to see very small ns differences in latency to the server.

And ping wouldn't be a test of actual speed anyway, since he would see the same ping times be it he was on a 10mbit, 100 or even 1000 gig connection.

The very very very small difference in latency being an extra switch between is not going to make for any real difference in speed. Now line be full between pcX and pcY could make for difference. Connection speeds 10/100/1000 would really make difference in moving of data (speed)..

But if there is no network bottleneck between pc on the downstream switch, and same bandwidth available to the core as to the server off the core - where you have a common path now for both machines. Their speeds is not going to be any different.

If you want to test speed, use something like iperf - which will just test the network and not use the disks of the machien which could put a hit on the test, etc. Now keep in mind that power of the machines, the nics on them, the network settings, etc. could cause different speeds to be shown for these machines that have little to do with the network path.

Your best overall test would prob be to copy a large file down from server on each machine at different times. So that they are not conflicting, and nothing else really going on in the network when you do this test.

But again generally speaking on a lan, all things being equal - adding an extra switch would not cause any sort of slow down in speed. Your only talking a few nanoseconds at worse case.. Unless something is wrong, or downstream switch is overloaded, or connect from that switch to core switch is saturated, etc.

although what you said is a good point he wishes to access the server using the machine as is a believe so removing the disk factor by using a live CD changes the factors completely and the results of speed would not be the same as he is trying to find which machine is better to use you should only test in the set up you are going to use it from ... if he plans to use nothing but live CDs then he should use live CDs or anything like that ...

if u get where i am coming from

[+BudMan MVC]

I think his concern of speed differences is in relation to the network, and there being an extra switch between pc1 and core switch, where pc2 is directly connected to the core switch. After the core switch the path to the server is common.

The way I read his email he was concerned that the extra switch could cause that access to be slower - maybe I misread it? But that is the way it looks to me.

[RATiO]

I think his concern of speed differences is in relation to the network, and there being an extra switch between pc1 and core switch, where pc2 is directly connected to the core switch. After the core switch the path to the server is common.

The way I read his email he was concerned that the extra switch could cause that access to be slower - maybe I misread it? But that is the way it looks to me.

Thats correct.

SERVER ---> SWITCH1 --> CORE SWITCH --> PC1

SERVER ---> SWITCH1 --> CORE SWITCH --> SWITCH 2 --> PC2

SWITCH 1 and Core Switch have 1000 mbps Gigabit speeds all other connections are Cat 5E so 100mbps

SWITCH 2 in above diagram is fully loaded with devices on nearly every port

To those suggesting iperf - doesn't that require the iperf server app to be run on the server machine (going against my original post)

[+BudMan MVC]

"are Cat 5E so 100mbps"

Cat 5e runs gig no problem, so that statement is just meaningless.

Yes you have to run iperf in server mode on the server.. Did you edit your post with the bold section? ;)

What is the connection between core and switch2?

So both your machines are connected at only 100mbit? If line between switch2 and core is busy, then yeah pc1 would be faster. Like I said its all going to come down to any bottlenecks in the network. Has nothing to do really that there is an extra switch. What is going to matter is how much traffic the other devices on switch2 are hitting the line to the core.