Network performance issue....

Hi,

I have asked a question in the past about a network performance issue since moving to our brand new building.  I have been testing from different areas of the building and have found one office that seemed to have a very fast connection (my only way of testing was to run a speed test on www.dslreports.com/stest).  I found that it had about twice the download speeds as some others I have tested.  I swapped computers in that office with an identical computer (both Gateways) in another office.  The speeds were very good with this other computer running in this office too.  So then, I went to one of the known problem computers (which is a Gateway also but is not the identical model) and plugged it into this office and it had a bit higher upload speeds than the other 2 computers, but a much lower download speed than the other 2 computers, almost by half.  So, for whatever reason, this machine is still slow in this office that is generally faster.  On all computers, the link speeds are set to Full Duplex 100Mb.  All of my ports in my switches are also set to Full Duplex 100Mb.  So I know it's not a mismatched link speed or Duplex issue.  Any thoughts on why this computer (which there are others with this problem too) is getting so much slower download speeds than the others in the same office connected to the same port?  

Thanks!

Mark
LVL 13
mark-waAsked:
Who is Participating?
I wear a lot of hats...

"The solutions and answers provided on Experts Exchange have been extremely helpful to me over the last few years. I wear a lot of hats - Developer, Database Administrator, Help Desk, etc., so I know a lot of things but not a lot about one thing. Experts Exchange gives me answers from people who do know a lot about one thing, in a easy to use platform." -Todd S.

pseudocyberCommented:
Hmm.  I was thinking Duplex mismatch until you said they're all 100 Full.  Are you absolutely sure?  What kind of network gear are you running?  Can you get on the switch(es) and see the Ethernet interfaces and check configuration as well as ethernet stats?
Les MooreSr. Systems EngineerCommented:
Ran into almost the same issue not too long ago. Seems like the user (Executive!) dinked around with the QoS settings on his XP desktop.. Disabled the QoS driver and viola' instant speed increase..
mark-waAuthor Commented:
Thank you for your responses.

pseudocyber - I am running Cisco 2950T-48-SI Layer 2 Managed switches and Cisco Catalyst 3548 XL Series switches.  I can access them and view statistics, but I really don't see anything that looks wrong and they definitly say they're running at 100Mb Full.

Irmoore - Are you talking about just disabling the QoS RSVP service?  If so, that service is set to manual and it's not running.  Is there more to it than that?

Thanks!

Mark
Learn Ruby Fundamentals

This course will introduce you to Ruby, as well as teach you about classes, methods, variables, data structures, loops, enumerable methods, and finishing touches.

nexissteveCommented:
Sounds very much like a cable fault to me. I have also seen this issue with inferior cables > is the cable cat5 or cat5e?

A good test to see if it is possibly cabling - lower the speed on the problem port to 10 full - set the computer to 10 full - run your download test again.

try 10 half as well - should be more than ample to test download speeds from DSL.

Have you tried replacing the patch leads? Have you had the cables tested prior to moving in to the new building?

Cheers

S

mark-waAuthor Commented:
Hi nexissteve,  It's cat5e.

We just had the company that wired the building come out and run cable tests again and they all passed with flying colors.  I walked around with them just so I could watch every test.  That was my thinking too, but unfortunately, that wasn't the case.  Here is something else to add to the equation.  One department is complaining the most about the slowness, they all have the exact same model of computer, they were all ordered at the same time.  Even though, they didn't seem to complain as much in our old building, could it be just an issue with this computer or it's network card?

In my original post, I had said that I took a different computer into the office that is normally faster and it tested at about half of the speed of the other computer for downloading.  This computer that I brought in was from that department that is having most of the slowness issues.  My computer is also that same model as that department and it gets the same slow download speeds as the others.  There are other computers that experience slowness, but isn't it strange that this one model of computer (Gateway SB-4100-A) experiences slowness anywhere in the building?  Could it be a cheesy on-board NIC?  I don't see how 2 different NIC's that are 100 full duplex can run at different speeds, that's just strange.  But just going from the newer gateway in that office that had great download speeds to this computer in that same office seems to suggest that.  These machines run almost identical software also.

Any thoughts?  Thanks.

Mark
nexissteveCommented:
Sounds like a nice one -yes it could definitely just be an issue with that particular machine.

Have you considered throwing a second nic in one of the gateway machines and disabling the onboard nic.

That would prove if it was the onboard nic - also there may be a driver / firmware update available from gateway if the problem is the mobo nic.

cheers

S
pseudocyberCommented:
Tried upgrading/changing drivers on the NICs?
mark-waAuthor Commented:
Just looked and I don't find any more current drivers for the NIC.  I think that's a good idea to try a second NIC and disable the on-board.  I'll let you know what happens with that.  Thanks!

Mark
mark-waAuthor Commented:
I have ran some more tests and have gotten really strange results.  Just so we can label these machines, the original machine that was in that office (the newer Gateway) is called HEME, and the machine that I brought in that was a different model of Gateway is called CRUBIN.  HEME gets very fast download speeds (600+KB/sec) while CRUBIN gets slower download speeds of around 300 or less KB/sec.  I can take these machines anywhere in the building and get those results, so I know now that it has nothing to do with location, so that office is not faster, it's these newer pc's that are faster for some reason, even though all computers and ports are set to 100Mb Full Duplex.

So then, from the HEME pc, I downloaded a 194Mb file from one of our servers (KEYFILE server).  It downloaded it in 1 min. flat (which I thought was a little slow for being inside the network.

Then I downloaded that same file from the same server on CRUBIN and it took 1 min. and 8 sec. (not a big difference from the HEME pc).

Then, I downloaded a 359Mb file from another one of our servers (TAM-APP) and from the HEME pc, it took 1 min. and 3 sec. and from CRUBIN it took 1 min. and 6 sec.  Again, not a huge difference, but strange that I could download almost twice the size file from TAM-APP than KEYFILE and it took about the same amount of time as the smaller size file from KEYFILE.  

Last, I downloaded a 12Mb file from the internet.  On the HEME computer, it took 22 sec. and on CRUBIN it took 59 sec.  So it took almost 3 times as long for CRUBIN to download a file from the internet than it did for the HEME pc, which corresponds with my internet speed tests that I have been running on them to determine download speeds.  But yet they are virtually the same downloading things within our network.

What does this mean?  This just doesn't make sense to me.  All of my network settings are correct.  They have their correct ip addresses, the DNS settings are right, the default gateway is right, everything is correct.

I hope someone can help me make some sense of this.  Thank you!

Mark
pseudocyberCommented:
Same NIC same driver?  Maybe an MTU difference?
mark-waAuthor Commented:
These machines both have different NIC's.  They are both on-board.  I still haven't tried a different NIC in them to see if the result is different.

I didn't even think of MTU!  I'm going to try some tests adjusting MTU settings.  I'll let you know what I find.  Thanks!

Mark
mark-waAuthor Commented:
HEME computer can ping an ip address with a max packet size of 1472 and CRUBIN can ping an ip address with a max packet size of 1272.

I believe, if you never change the MTU settings within Windows, the default is 1500 for the Max MTU, so I will try changing those to their actual max sizes from my tests and see what happens.  Thanks.

Mark
pseudocyberCommented:
No problem ... except for taking a while to think of it! ;)

Correct, default is 1500.
mark-waAuthor Commented:
Well, the HEME computer was able to download the same 12Mb file that I downloaded earlier in 20 sec. and CRUBIN took 1 min. and 25 sec.  This is killing me.  I really don't get it.  I am going to try and get a different NIC to test.  Thanks again for helping.  More ideas are welcome!

Mark
GinEricCommented:
yes, the limitations of Ethernet, as opposed to ATM.

The QoS, Provisioning, the whole scheme of PPPoE and the limit of packet size for Ethernet of about 1500 MTU relegates a lot of switching to about 12,000 bits per packet maximum, regardless of bandwidth.  This increases overhed, hops, switching, Time To Live delays, etc..  It increases packet assembly and disassembly, and thereby increases cpu times and slows any connection to a near convergence, if you understand the mathematics.

The MTU on the internal servers here are about 16,000 or ten times that of the Internet and Ethernet.  That is ATM switching on the loopback.

So, the packets can be ten times the size within a server, a Linux Server anyway.  That means one tenth, at least, of the period of transmission/reception and hence everything else happens about ten times as fast.

It sounds like your servers have provisioned your entire network.  Also, this may be known as load balancing.  There was a move to push MTU to 9,000, which is near the maximum packet size of 65,536 yet this was not accepted nor implemented as far as I know.

That is why I also mentioned fibre optics and parallel data transmission/reception.  Once this happens, speeds and bandwidth will jump tremendously, but only if the packet assembly/disassembly is redefiend so that a parallel buss from all telco's and ISP's are paralleled [which is actually how the phone company grids now do it].

Secondly, this is why USB, SATA, and other serial devices are a step backwards, way backwards, to the days of things like Commodores, dialups at 2400 bps, and other early slowness.

With parallel fibre optics, data and information load so fast you cannot see it.  And that includes downloads.  With 64 parallel lines and better MTU and ATM, the download of even a 1 gigabyte file would be 64X10 or 640 times faster.  That is, less than one second for all of the files you described.  That is how fast modern machines are.

It is not the machines, it is the basic Internet architecture over the phone company wires that is slow, provisioned, and requires Quality of Service controls.

And this architecture is carried forward, by default, into internal networks, especially Ethernet based ones.

NIC's, 1 gigaHertz cards, none of this is really going to matter when everything is basically on one wire.


pseudocyberCommented:
Mark,

Try loading Ethereal on the two computers and downloading the same file, while running a capture.  Compare the two captures to see if you see anything obvious?

Also, I forgot if you mentioned - are these two machines the same?  Same processor, memory, disk, OS, etc?
mark-waAuthor Commented:
I do have Ethereal installed, but haven't tried running it while downloading.  I'll try that.

These 2 machines I am testing with are different models, but both are Gateways.  The "HEME" machine is only a couple of months old.  We have 14 of this same model and they all get the faster download speeds.  The "CRUBIN" machine is about 2 years old, but it is a decent machine (2Gb Processor and 1Gb RAM).  There are about 12 of this model and they are all getting the slower internet download speeds.  Other machines on the network are older than that even.

The HEME machines are running Windows XP SP2 and everything else on our network right now is Windows 2000 SP4.

I'll check out running Ethereal while downloading a file.

GinEric, I'm not sure I understand everything you were talking about.  All I know is I have these 2 machines that are different models of Gateways that I have been testing with and they both have more than adequate resources.  They both are set to 100Mb Full Duplex, as well as their respective switch ports.  On the internal network, they get very similar speeds.  But downloading from the internet, the newer machine is much much faster.  This doesn't make sense to me.  And if it were a "phone company wiring" problem, wouldn't it still be the same speeds regardless of the machine?

Thanks for everyone's help with this.

Mark
GinEricCommented:
One bit in one trillion can make all the difference in the world.  You can't be sure that any two machines are ever exactly the same, for whatever reasons, bad cable in one machine, nicked and nearly broken wire in a ribbon cable, one faulty RAM gate that is slow, rather than fast, or one corrupt bit on disk that keeps getting one bit correction.

If it's computer or office specific with a certain set of machines, then there is something common to that set; perhaps their computer names are recognised by the server and provisioned, perhaps they are reporting the wrong transfer rate, whatever.  It's usually the transfer rate or they are being provisioned improperly for some reason.  It could be as simple as a bad wall outlet or switch module RJ45 connector that is salted over.  Some of these connectors do grow salt, some even grow algae.

But the focus is that if it is three specific machines, then it is three specific machines and it seems that the error is outside of the normal causes.  Which is why I mentioned things like salt and algae; a question of their environment histories.  Temperatures, any employee growing strange plants, using space heaters, cooking weird foods near the computers, etc..  It just seems to be "abnormal."

You seem to have a classic case of "incompatible space" between perhaps employees and machines.  It seems to be beyoned simple electromechanical failure.

Remember that with Eniac, there were real physical junebugs that got caught up into the computer.

Your problem seems to be leading toward an organic problem, rather than a simple technical problem.
mark-waAuthor Commented:
Thank you for your explaination GinEric, but I kind of disagree.  I stated above that location is not a factor.  I can plug one of these new machines (like the HEME machine, or any of this same model) into any network port on any wall, anywhere in the building, and it will get the fast speeds.  I was using the HEME machine as an example, but we have 14 of this same model machine and they all get the fast internet download speeds, regardless of location.  I think the difference is Windows Xp versus Windows 2000.  I have a machine that is neither of the 2 models I have been testing with, it is about 4 years old, has less RAM, a slower processor, but it has Windows Xp.  I tested it's download speeds and they are very fast, just like the newer machines running Windows Xp.  I then tested another older Windows Xp machine that we have and again, it has faster download speeds.  So, this leads me to believe that either Windows Xp is somehow better, or there is an update or patch or something that has been rolled out on my Windows 2000 machines that is affecting all of them the same.  

Is it possible for Windows Xp to just be faster?

Thanks!

Mark
GinEricCommented:
Yes, XP is faster.  You didn't mention that HEME was XP and all the others were Windows 2000 until this:

"The HEME machines are running Windows XP SP2 and everything else on our network right now is Windows 2000 SP4."

and is was a little hard to pick out, since the original question made no mention of this.


Experts Exchange Solution brought to you by

Your issues matter to us.

Facing a tech roadblock? Get the help and guidance you need from experienced professionals who care. Ask your question anytime, anywhere, with no hassle.

Start your 7-day free trial
mark-waAuthor Commented:
But 2 to 3 times faster?  Thanks!

Mark
GinEricCommented:
Yeah, big difference.  Windows 2000 is about at the end of it's useful life [5 years generally before Microsoft obsoletes a software program].

Developers and driver creators tend to skew toward the leading edge, eventually just abandoning things after a few years.  Old 32-bit drivers, old file systems, dual file system for Windows 2000, and so on, it spends much more time processing than XP does.  And the routing is old hat.
mark-waAuthor Commented:
Thanks for everyone's help, but it looks like this is a case of Window XP being more efficient than older versions of Windows.  I still don't see how the speeds can be THAT different.  But I can find no explaination other than it seems any machine (different models even) running Windows Xp on our network is much faster than any machine running Windows 2000.  If anyone has a mixed environment with some WinXp Sp2 machines and some Win2k Sp4 machines, if they wanted to run some speed tests on the internet and let me know their results, that would be great.  I go to http://www.dslreports.com/stest and then I choose the MegaPath test (second one down).  Thanks!

Mark
It's more than this solution.Get answers and train to solve all your tech problems - anytime, anywhere.Try it for free Edge Out The Competitionfor your dream job with proven skills and certifications.Get started today Stand Outas the employee with proven skills.Start learning today for free Move Your Career Forwardwith certification training in the latest technologies.Start your trial today
Network Analysis

From novice to tech pro — start learning today.