Re: testing bandwidth of big internet pipes
I tend to think of T1 and E1 connections as small (STM1s are starting to be big :-). It's easy enough to test them if there's something fast enough to test to. The two kinds of tests your customers are likely to care about are - Is your connection to them really the speed it should be? - Do you have enough upstream bandwidth for your traffic load? There are lots of ways to get and measure a 2Mbps load - - Most Cisco routers have "ttcp" client/server code, http://www.cisco.com/warp/public/471/ttcp.html http://www.netcordia.com/tnm/tnm31/ttcp.htm and you can either talk to another Cisco or use a Windows or Linux PC. - FTP on a Sun 3/60 a decade ago, or a Linux box with FTP or HTTP today, and any medium-fast Windows machine with an FTP or HTTP client that shows you download speed is usually a good demonstration. (You need to ignore initial ramp-up time because of TCP slow-start.) If you've got a spare PC you can hang off one of your routers, run an FTP and HTTP server on it and keep a few "10MB.txt" type files on it. - Qcheck is a nice friendly free client for Windows and other platforms which does speed and latency tests using tcp, udp, and ping, and can also be driven by their non-free testing systems. http://www.ixiacom.com/enterprise/Qcheck.php (Qcheck does use more PC CPU than simpler tools - my 66MHz Pentium doorstop boxes could only do ~300-400 kbps, but anything over about 500 MHz should do just fine.) - A really *bad* method that's popular with some of my customers (:-) is to use really big pings on a Cisco router. There are lots of different reasons this tends to fail. The second question is harder to answer, because you need a site that's not in your network that can give a fast enough response that the customer will feel happy, and I've found that it's often hard to download files from public web sites at faster than 1-2 megabits per second, because either the web site is <= T1, or it's large and shared by many users. Linux distribution sites when there's nothing interesting being released are usually good, but unfortunately mirror sites often mirror lots of things, so you may have trouble finding a quiet one. I don't remember if BitTorrent has a fancy user interface that shows current download speed, but it's usually good at filling up any symmetrical Internet pipe if there's something popular to download.
This is almost comical when I read this, when I sit out here and see people getting whacked at 3Gb/s and from an estimated 2.3 million drones, sitting on various proxies and hacked .edu accounts..... Where are the pages that have solutions for such intense attacks and if you don't want to say publicly then email me direct to this address. Thanks -Henry "Stewart, William C (Bill), RTSLS" <billstewart@att.com> wrote: I tend to think of T1 and E1 connections as small (STM1s are starting to be big :-). It's easy enough to test them if there's something fast enough to test to. The two kinds of tests your customers are likely to care about are - Is your connection to them really the speed it should be? - Do you have enough upstream bandwidth for your traffic load? There are lots of ways to get and measure a 2Mbps load - - Most Cisco routers have "ttcp" client/server code, http://www.cisco.com/warp/public/471/ttcp.html http://www.netcordia.com/tnm/tnm31/ttcp.htm and you can either talk to another Cisco or use a Windows or Linux PC. - FTP on a Sun 3/60 a decade ago, or a Linux box with FTP or HTTP today, and any medium-fast Windows machine with an FTP or HTTP client that shows you download speed is usually a good demonstration. (You need to ignore initial ramp-up time because of TCP slow-start.) If you've got a spare PC you can hang off one of your routers, run an FTP and HTTP server on it and keep a few "10MB.txt" type files on it. - Qcheck is a nice friendly free client for Windows and other platforms which does speed and latency tests using tcp, udp, and ping, and can also be driven by their non-free testing systems. http://www.ixiacom.com/enterprise/Qcheck.php (Qcheck does use more PC CPU than simpler tools - my 66MHz Pentium doorstop boxes could only do ~300-400 kbps, but anything over about 500 MHz should do just fine.) - A really *bad* method that's popular with some of my customers (:-) is to use really big pings on a Cisco router. There are lots of different reasons this tends to fail. The second question is harder to answer, because you need a site that's not in your network that can give a fast enough response that the customer will feel happy, and I've found that it's often hard to download files from public web sites at faster than 1-2 megabits per second, because either the web site is <= T1, or it's large and shared by many users. Linux distribution sites when there's nothing interesting being released are usually good, but unfortunately mirror sites often mirror lots of things, so you may have trouble finding a quiet one. I don't remember if BitTorrent has a fancy user interface that shows current download speed, but it's usually good at filling up any symmetrical Internet pipe if there's something popular to download.
participants (2)
-
Henry Linneweh
-
Stewart, William C (Bill), RTSLS