Reccently I've been in search of a new host, so i sent out the standard presale questions. One was about a testfile to download.
My query is... Can these testfiles really help in reflecting the server's bandwidth? Also is there a program/website to test all the download speeds at once to get a fair comparism.
I reside in Asia and I tested downloading a file on a Canadian server based in Netnation and I only got roughly 150Kb/s . Then I went on to test another file on a US server based in EV1's datacenter and i got a good speed of 200Kb/s.
I went to test on a LiquidWeb reseller, and I only got around 70kb/s !? (LiquidWeb is known for its good network no?)
On an average, 200Kb/s plus-minus is my maximum download speed. So I'm getting good speed from EV1's DC and terrible speed from a host that resells on LiquidWeb.
On a whole, can these speed test show or prove anything? Someone please enlighten me
They can tell you quite a bit if you know how to interpret them. Lower speeds can be caused by congested networks, routes with heavy packet loss, or even by servers pushing high amounts of traffic.
My suggestion would be to run an mtr (you can get a windows version of the utility at winmtr.sourceforge.net) to the server for about 10-15 minutes, and watch for high ping times/packet loss on your route. Your route to the servers is very important.
The speed test proves exactly what you think it does, the download speed from that server at that particular moment. It can be affected by a poorly performing server, poor network connections to your location, a temporary glitch in the routing. If you constantly get poor speeds, consider another provider if it's important to you.
Try traceroute'ing to the server to follow the route to the server, the ping times can tell you where the connection is slowing down if indeed the network is the problem rather than the server.