Superhub : speedtest.net

[Chrysalis]

Brilliant, VM have customers blaming speedtest.net for subpar peering.

[kwikbreaks]

Originally Posted by rmwebs

As an example, I did a test on the maidenhead server. I got a 25mb result.
I did a test less than a minute later on another server (Manchester) and got my full 100mb.

One feature of congestion is wild speed variations over a short timescale. No doubt if you'd seen the same results all on the same speedtest server you would have blamed congestion at the server - the odds are strongly if favour of the congestion being within VM - especially as qas has tried many of them on a university connection which showed them well capable of delivering 100Mbps at any time.

[smallclone]

So, are we stuck with congestion issues? Or are there plans in every area to improve it?

[sorry, I'm sure this has been discussed]

[craigj2k12]

Originally Posted by smallclone

So, are we stuck with congestion issues? Or are there plans in every area to improve it?

Id say were stuck with them, especially when all areas are "doubled"

[Dush]

Originally Posted by qasdfdsaq

Quite the opposite, ISP provided tests are the least accurate.

After all, people get internet connections to access the internet and ISP tests test everything except the internet.

Even if you buy a leased line the bandwidth you are guaranteed is only to your ISP's network. Once it's out on the internet it's all best efforts.

Virgin should offer this as their speeds are so quick that most speed tests do not keep up. Only usenet does for me.

[qasdfdsaq]

On the contrary, when Speedtest.net servers hosted on VM's network deliver 25mbps to VM's own customers but over 200mbps to everyone else that suggests it's VM's own capacity that can't keep up.

[roughbeast]

Originally Posted by smallclone

This site gives my speed at 20mb/s

I have tried 3 others and they give my speed at 57MB/s which is what I pay for.

So which one(s) are right?

Thanks

Have you tried the Speedtest.net NAMESCO server in London? This performs very consistently for me at 100 miles. Not all servers work well despite what the numbers on the speed dial suggest.

[kwikbreaks]

Originally Posted by Dush

Even if you buy a leased line the bandwidth you are guaranteed is only to your ISP's network. Once it's out on the internet it's all best efforts.

That's oversimplification - you are ignoring the peering which will vary from ISP to ISP and VM's is a major bottleneck at times.

Obviously if the specific server you are trying to access is congested or on a poor link it doesn't matter how fast you ISP is but what is increasingly being reported is that many of the speedtest.net servers don't deliver high speeds to VM customers but do when tested from known good connections such as universities.

If you do succeed in getting good speeds from some sources then all that rules out is local congestion at that specific time and the peering to that specific source.

[Dush]

It's not an oversimplification, it's in the SLAs of Leased Line contracts.

[kwikbreaks]

Originally Posted by Dush

It's not an oversimplification, it's in the SLAs of Leased Line contracts.

Sorry but I don't have the faintest idea what you are getting at. Do you?

[qasdfdsaq]

Given his reputation, probably not.

[Dush]

People who buy leased lines for hundreds, sometimes thousands of pounds a month only get bandwidth guaranteed to their ISP's network. So the point (as should be obvious from the post I was replying to!) is that if Virgin hosted their own speed test they could then offer more accurate speed tests.

I know speed test is rubbish for me. I get full bandwidth on torrents and usenet but not on speedtest.net. If I didn't have access to usenet/torrents I'd be moaning that I'm not getting the speed I paid for just like alot of VM users do.

[Chrysalis]

Originally Posted by Dush

People who buy leased lines for hundreds, sometimes thousands of pounds a month only get bandwidth guaranteed to their ISP's network. So the point (as should be obvious from the post I was replying to!) is that if Virgin hosted their own speed test they could then offer more accurate speed tests.

I know speed test is rubbish for me. I get full bandwidth on torrents and usenet but not on speedtest.net. If I didn't have access to usenet/torrents I'd be moaning that I'm not getting the speed I paid for just like alot of VM users do.

however said isp's dont normally let their peering links just clog up. Its already been posted on here how small VM's linx capacity is eg. in comparison to other isp's.

[roughbeast]

Originally Posted by Dush

People who buy leased lines for hundreds, sometimes thousands of pounds a month only get bandwidth guaranteed to their ISP's network. So the point (as should be obvious from the post I was replying to!) is that if Virgin hosted their own speed test they could then offer more accurate speed tests.

I know speed test is rubbish for me. I get full bandwidth on torrents and usenet but not on speedtest.net. If I didn't have access to usenet/torrents I'd be moaning that I'm not getting the speed I paid for just like alot of VM users do.

VM have there own speed test, but this is only available to techs and folk on trials etc. It is at http://www.vmspeed.com/, but this link won't get you anywhere.

Really you don't need that test. This one is just as consistent.

http://www.speedtester.bt.com/beta/speedtest-remote.jsp

[kwikbreaks]

Originally Posted by Dush

People who buy leased lines for hundreds, sometimes thousands of pounds a month only get bandwidth guaranteed to their ISP's network.

Many (most?) leased lines are point to point between different locations of the same big company.

Where the line is primarily for internet access then there will virtually certainly be something in the contract about peering - otherwise what's to stop them only having a 10Mbps VM line to access the internet? It won't happen. That is why I said you can't simply judge by the speed of your connection to your ISP - unless of course all you want to do is look at their billing pages etc.