Freeware to test network performance?

[Leozack]

Hi all - does anyone have a recommended app for testing network infrastructure performance on a network that doesn't involve manual installation on clients?

I'm considering making an app to read in a list of pcnames/locations and visually display them, then you tell it Xmb of data to test with, and it tests sending/receiving that much data to all the clients (somehow) and then maps the results so you can see if there is a problem anywhere on your network. Doesn't sound tooooooo complicated ...

But then I figured hang on - people must've been doing this for years (probably with the expensive network management apps though!). But I thought I'd ask if anyone knows of freeware that will do something like this? I've got networks where I'm dying to know if the infrastrucutre and switches are healthy or not.

_________________________________
Leozack

MakeUniverse($infinity,1,42);

 

[technome]

Aside from the built in OS performance monitors.... QCheck, just for a down and dirty quick idea of whether a change to an OS parameter or network component has increased or decreased network performance. It is not a full scale tester but is good if your trying to tweak out a network, making many OS/network changes, one at a time. Produces instantaneous results, so if your trying 100 different changes, you will not be there for a week. Did this over the last weekend at a client for 2008, took 12 hours but better than days; I had already tweaked it during the intial setup, could not get a dime more speed after 12 hours of play after adding a new NIC setup.
Same link gives access to Chariot, for a full scale performance tester (trial evaluation).

http://www.ixchariot.com/products/datasheets/qcheck.html

........................................
Chernobyl disaster..a must see pictorial

http://www.kiddofspeed.com/default.htm

 

[Leozack]

Ah yes thanks - I think I came accross this before sometime. But for some reason I recall I couldn't get it to work, or had problems putting it to useful use. The description once again sounds like simplicity and heavenly usefulness. I wonder why I never managed to grasp that in reality. I'll have to try it out again I guess (hopefully there is a way to deploy an endpoint or qcheck itself to clients on startup or via gpo).

Of course what this won't do is handle a list of PCs and compare results etc. I'll have to do that side of it myself in something else.

_________________________________
Leozack

MakeUniverse($infinity,1,42);

 

[technome]

Couple of the servers I was testing were 64bit, I had downloaded an older version of QCheck from other sites and it failed to work on them, the version on the site has been revamped, worked fine. This newer version creates a service on client machines so it does not need to be open on them, once testing is done, you can uninstall the software or disable the service on the client machines, install take about 2 minutes/machine.

"Of course what this won't do is handle a list of PCs and compare results etc"
Chariot eval will do this. Have not tried it, as I dislike placing large evals on client's server (code bloat), QCheck is a small footprint, so I am not concerned.
Not sure if you can install QCheck via GPO, did not see anything about unattended installs. Generally I use it between servers, and just a couple "average" machines to get an idea of throughput involving wks and network switches/routers/wiring or if I suspect a particular machine has issues. This last testing at the client also involved new 1 Gig switches... was worried about the wiring (CAT5), which performed flawlessly at both Gig speed, 0 errors on the switches.

http://www.ixiacom.com/popup_filefile=/support/q_check/faqs.php#CommandLine

........................................
Chernobyl disaster..a must see pictorial

http://www.kiddofspeed.com/default.htm

 

[Leozack]

Yeah I'll have to test if I can silent-install the endpoint client. Being able to note down results from various machines around a site you can get a picture of what switches are bottlenecking or what infrastructure may be a damaged cable. Like you I'd rather note install bloaty trials, nor wil anyone ever justify buying software as usual!

Looking at the details of qCheck I think the probelm I probably found was it's old, designed when 100mps networks were new, and for instance the most throughput I can use to measure is 1mb. Nowadays I use files that are 200-2000mb to test a steady network transfer speed if I need to take a measure :| I guess I struggled to do anything more qith qcheck than get a ping result since the other tests are over too quicly? Or maybe you found better use for it in streaming mode and found it dropped 0 packets then?

_________________________________
Leozack

MakeUniverse($infinity,1,42);

 

[technome]

Damn, just lost my last post.

"Nowadays I use files that are 200-2000mb to test a steady network transfer speed if I need to take a measure."

Even though I love to see large file/directories fly, this is not realistic. Windows 2008, network adapters, raid adapters are optimized for the transfer of much smaller data chunks, so even if I do time the transfer of large files/directories occasionally, I reframe from getting to excited. There are just to many variables to get honest result. This form of testing overwhelms the raid array cache/co-processor, motherboard interrupts, CPU, switches, OS optimized parameters etc with unrealistic data. The average data transfer size between wks/server or server to server in a client-server relationship is so much smaller.
Again, QCheck to me is only a method to let me know if a change increases or decrease just the network performance; wish it did have options for large data sizes .

"and found it dropped 0 packets then?"
Forgot where I read it, but it normal for QCheck to drop packets.


PS..
Just remembered to watch Jeopardy. IBM's "off the shelf hardware" computer crushed the top players tonight. Guessing IBM defines "off the shelf hardware" as a shelf the size of a small warehouse, costing as much as the GNP of a small country. Really cool though.
We are close to the point where computers will design themselves at a logarithmic rate... let's pray they always have a power switch.





........................................
Chernobyl disaster..a must see pictorial

http://www.kiddofspeed.com/default.htm

 

[58sniper]

Watching the counters on switch ports can help show issues as well.

Pat Richard MVP
Plan for performance, and capacity takes care of itself. Plan for capacity, and suffer poor performance.

http://www.exchangeblogs.com/

 

[technome]

Yep, thank the IT gods for managed switches.


........................................
Chernobyl disaster..a must see pictorial

http://www.kiddofspeed.com/default.htm

 

[Leozack]

Sadly manaed switches are a rarity for me due to - of course -their extra cost :/

When testing 200-2000mb transfer for instance I was testing speed to/from a NASdrive. Turns out despite being gigabit and sata300, it was faster to/from a 100mps workstation. Oh well :|

I'll see if I can utilise qCheck, else I'll have to try and write something.

_________________________________
Leozack

MakeUniverse($infinity,1,42);

 

[technome]

Sadly managed switches are a rarity for me due to - of course -their extra cost"
The Gig switches I picked up I got from Ebay, 3 Dell 5424s for $825.00, client could not go for new (approx $750.each). Turns out all 3 had a couple small scuffs, setup, but never used.

Leo, I do not know your settings, but one of the biggest boosts is turning off flow control if the wiring is up to standards. Disabling SMB signing on 2008 also helps.

Went through many of these tweaks in the link

http://www.speedguide.net/articles/windows-7-vista-2008-tweaks-2574

........................................
Chernobyl disaster..a must see pictorial

http://www.kiddofspeed.com/default.htm

 

[Leozack]

I've got mainly XP clients and some 2003 servers and some 64bit 2008 servers depending on the site. Occasionally someone has a win7 laptop or something but that's rare. Probably still more 2003 servers than 2008. I had to disable SMB signing on one of the 2008 servers in order to allow MACs to use the fileshares (don't get me started on sharing filespace with MACs ...).
I can't vouch for the varying ages and quality of the wiring so can't lean heavily on it without testing I guess (not sure where I'd disable flow control)

_________________________________
Leozack

MakeUniverse($infinity,1,42);

 

[technome]

not sure where I'd disable flow control"
If it is disabled on either the switch or the NIC, it's disabled. From what I experienced, unmanaged switches have it enabled as a default, not sure of all unmanaged switches.
Generally in the NIC properties, configure, advanced, one of the properties will be flow control. Toggle it off and definitely test. Goes for servers also. I worked with networks under 250 wks, have it disabled on all; on larger networks or if the wiring is sub par likely it would be needed.

As to MAC sharing, never had the pleasure but I had a nightmare when I setup my first 2008 network, SMB signing enabled caused the network to go completely unstable. Had issues with it on previous Win server versions but nothing like with 2008.



........................................
Chernobyl disaster..a must see pictorial

http://www.kiddofspeed.com/default.htm

 

[technome]

Leo,something on flow control..

http://www.smallnetbuilder.com/lanwan/lanwan-features/30212-when-flow-control-is-not-a-good-thing

........................................
Chernobyl disaster..a must see pictorial

http://www.kiddofspeed.com/default.htm

 

[Leozack]

Interesting read - though having got to the end my initial thought was "so you tell the fast transmitter not to wait when it's asked to wait by a slow receiver" and this speeds up network transmissions - great! But err ... wouldn't that mean the slow receiver is left behind and thus you get data corruption or something bad going on? Managed switches would be nice, I don't fancy trying to manually edit flow control on workstation NICs! Perhaps just on the servers ... have to think of a good qCheck test to test with afterwards

_________________________________
Leozack

MakeUniverse($infinity,1,42);