Collaboration

Helping users optimize their Internet connection speed

With users becoming more reliant on a speedy Internet connection, including file downloads and e-mail attachments, it might be worthwhile to test their connection speed and make upgrades accordingly.

After a recent network infrastructure upgrade (router and switches), I was pleasantly surprised to see an increase in my Internet connection speed -- especially the download speed -- on one of my file servers. According to a test I performed at www.speedtest.net, I increased the download speed almost four times, going from about 2,650 kbps to almost 10,000. I also realized an increase in speed on my own computer as well, and although it wasn't as drastic, it was quite significant and very noticeable.

Realizing that any number of factors could affect a user's Internet connection speed and knowing that the slowest link in the connection chain would ultimately determine the speed, I decided to do some random tests on about a dozen of my computers (well, my users' computers) to compare the results. The range of speed was both very revealing and, in a way, kind of confusing.

I know that components like the network interface card, network cable, and such, would definitely influence the speed; and I suspected that processor speed, memory, and other factors would probably have an effect as well. And, of course, the number of people using up available bandwidth (both inside and outside the office) would have an effect. But my mixed results made me wonder about all the different factors. I access the Internet by way of Comcast Broadband, so any download speed below the 9,000 to 10,000 kbps range would have to be attributed to something inside my own network, at least to the largest degree. I tested this theory by plugging directly in to the modem with a couple of different computers, and both were faster as a result.

The upload speed on these dozen computers was pretty consistent, ranging from 2,182 to 2,717 kbps, but with one exception. A Windows 2000 Server that's dangerously low on available disk space clocked in at only 1,041 kbps. That computer also recorded the slowest download speed of only 4,670 kbps. What I take from this is that hard disk space is definitely a factor. Two other Windows 2000 Servers were measured at speeds of 7,800 and 8,700. That's not too bad, and I would have thought that my Vista machines with Quad Core processors, loads of memory, and lots of available hard drive space would be among the fastest, but that wasn't the case at all. Those recorded speeds of between 5,400 and 8,500 kbps. The fastest computer was a Vista machine with only a P4 3.0 GHz processor. It clocked a download speed of almost 9,400 kbps. And one of my old PIIIs even rated higher than many of my Vista Quad Cores, scoring a download speed of 6,700 kbps.

After I return from my three-day vacation (five days counting the weekend), I might do some more testing, eliminating links along the way. Taking a computer that scored only 5,400 kbps, for example, and plugging it directly in to the modem, should be revealing. Anyway, even after the latest round of upgrading some of my network infrastructure, a few more upgrades might be in order. If I can significantly improve my users' connection speed, I'll post an update.

Do you often -- or have you ever -- tested your users' Internet connection speed? Does it vary significantly between users? Give www.speedtest.net a try, and share your results. Do you have any other tips for checking such things?

9 comments
Joe_R
Joe_R

Re the original piece: http://blogs.techrepublic.com.com/helpdesk/?p=258 Do you often - or have you ever - tested your users' Internet connection speed? Does it vary significantly between users? Give www.speedtest.net a try, and share your results. Do you have any other tips for checking such things?

seanferd
seanferd

Are two things you can check on for the PCs. The settings that Windows chooses aren't always the best. You can get a speedtest and other useful diagnostics and info at Broadband Reports dot com. Smoke Ping is an interesting utility there as well.

brad
brad

I'm a developer of disk to disk backup for AS/400 to pc systems, see www.laservaultbackup.com I've done extensive testing of boards and to a lesser degree, processors. I like the intel 1000 mt and pt ethernet boards, or integration into the motherboard is even better. Different ethernet cards WILL give different results, and some will be better at upload than download. When you are doing 1 gbit ethernet, your processor is moving a LOT of data around. CPU cycle time is much more important than number of cores. FSB and memory speed also make a big difference. The Intel ethernet hardware does some coprocessing, reducing the load on the PC cpu. You can also get a TCP offload engine like Alacritech, but it is much more expensive to do that. The Intel 1000 PT board is under $300 and can improve a heavily loaded server's performance, or for mission critical client PCs. It has two ports, so you can serperate internal traffic from external traffic if they are on different subnets. Registry tweaks can also help, like a larger mtu or larger tcp/ip window size (how much data it accepts before needing a tcp/ip acknowledgement). There is also a completely different setting within ftp for how much data to send at once, that can make a big difference for large ftp transfers.

ehorwitz1
ehorwitz1

Sounds like you're on the right track and I'd be interested to see what you discover. Please continue posting as you get further into your testing and fixes. I use a tiny, free app called TCP Optimizer from speedguide.net that requires no installation and is simple enough for novices to use. It gives a list of existing TCP settings for each installed network adapter. It can run tests and recommend a best MTU setting and measure the average latency. The application recommends optimized settings and can update them with a single click or allow the advanced user to have a field day tweaking individual settings to his or her heart's content. I discovered this tool after having some rogue application mess up my networks' settings when I uninstalled it. After reading lots and trying a variety of settings and tools, I came upon TCP Optimizer - it might even have been in the TechRepublic library (can't remember). It instantly restored my PC's network throughput. Actually, it increased it significantly above anything it had been previously. So, even though it doesn't sound like your issues are specifically at the PC, this tool may help discover the problem and if not, at least optimize the settings on each PC so there's one less area to worry about. The website has a number of other free tools and forums. Have a look!

Joe_R
Joe_R

Those are some good tips. Thanks for posting.

alt1231
alt1231

coudl yyou help me optimize my system using vmware and instructions?

Joe_R
Joe_R

Thanks for posting.

Joe_R
Joe_R

Thanks for posting.

seanferd
seanferd

I've had good results adjusting some of those numbers, when applicable. Some parameters should not be adjusted, depending on the OS. But I do check to see if adjustments are needed whenever anything about the network connection changes.

Editor's Picks