Ofcom’s latest broadband speeds report (released this week) also includes a unique and useful insight into the Latency (Ping) and Packet Loss performance of the UK’s largest ISPs, which can have a serious impact upon the performance of specific internet applications (e.g. online video games, Skype etc.). But which provider is best?
Most ordinary consumers probably aren’t aware of such terms, which is partly why the communications regulator chooses to centre the main thrust of their fixed-line ISP performance report on connection speed. In reality internet provision is about more than connectivity speed and there are other types of performance that can have an equally dramatic effect upon your experience.
Latency and Packet Loss are two related aspects where the general rule of “lower is better” almost always applies. In order to understand how both work it’s best to think of internet data transfers as being a bit like the postal service.
A measure of the time (delay in milliseconds) that it takes for a single packet of data to travel from your computer to a third-party server and back again (aka – ping time). Note: 1000 milliseconds = 1 second.
The term given to a problem that occurs when some of the data packets being transmitted between two or more points (e.g. servers) on the internet effectively go missing or are incorrect. The connection auto-corrects for this but doing so can result in an additional delay (increased latency times) or other problems.
Generally speaking you want a low-latency connection with minimal packet loss. Sadly Packet Loss is a natural fact of life on the internet. Network connections are designed to auto-correct for this problem but doing so can increase the latency time. A highly congested (busy) network or faulty networking hardware can also increase latency, partly by causing a higher degree of packet loss.
Satellite connections are also notorious for their ultra high network latency because the data has to be sent all the way up to space and then back down again. That’s a long round-trip and there’s simply no escaping the laws of physics (transmission delay).
Most fast-paced online multiplayer games need a latency of below 100ms (milliseconds), although gamers usually prefer considerably less than that in order to have the smoothest experience. Every game has a different approach to its netcode but go above 100ms and most soon start to suffer from unusual warping or lag, which occurs because the delay has caused players to become increasingly out-of-sync with the server time. The server software and game netcode can only correct so much for this before it becomes unplayable.
Likewise high latency and packet loss can also cause unusual pauses, breaks and or synchronisation delays in your voice or video calls over Skype (VoIP) or similar services. It can sometimes have a similar affect upon the performance of Virtual Private Networking (VPN) and or remote desktop style services. Essentially any time sensitive real-time apps need to be mindful of latency and packet loss.
Thankfully Ofcom’s report gives us a unique insight into how latency and packet loss affects eight of the UK’s largest broadband ISPs. The regulator splits this up so as to compare the older ADSL2+ (up to 20/24Mbps) based copper broadband services with the latest superfast (BTInfinity’s FTTC and Virgin Media cable) platforms. First let’s take a look at latency.
Clearly KC’s (Karoo) legacy broadband network in Hull (East Yorkshire) is running into some problems and exhibits a typical latency that would be high enough to cause online video gamers concern. Likewise Sky Broadband and TalkTalk need to do better. Interestingly Virgin Media’s results show that being the fastest for raw connectivity speed doesn’t always equal the best for latency, with BT’s FTTC (Infinity) solution appearing to be the strongest and most consistent of the two.
Note: Ofcom only measures the latency between the ISP and your home router/modem connection, thus if you use a wifi instead of wired network connection at home then that can also increase the latency delay (+10-35ms) as you’d be adding an additional path for the data to travel, which will have its own complicated processes. On to packet loss..
As expected the Packet Loss results largely support what we saw above for Latency. Both KC’s legacy broadband and Virgin Media’s superfast broadband platforms appear to have trouble here with a comparatively high number of data packets being prone to error. But at the same time Sky Broadband and TalkTalk do far better, while O2 suddenly shows a problem. This just goes to show that Packet Loss isn’t the only cause of latency but it does play a big part.
At this point it should be said that Ofcom’s sample size of fewer than 2,000 homes makes it difficult to draw any reliable conclusions, although clearly Virgin Media and KC don’t do as well as their competitors when it comes to raw network response time and error rates. The results aren’t high enough to threaten most real-time applications but fans of online video games should certainly take heed.
It’s worth noting that ISPs can sometimes make tweaks (e.g. interleaving and line gain etc.) that might improve your latency performance, thus it’s always worth asking for help if you have a problem. Some router settings (e.g. MTU) can also impact latency but that’s best left to those whom actually understand the technology; such changes can easily become counter-productive if not done correctly.
UPDATE 18th August 2012
The Director of UK ISP Andrews & Arnold (AAISP), Adrian Kennard, has reminded us of another factor in all this. BTWholesale still don’t recognise serious packet loss and latency on a line as a fault, even though they often relate to a more serious problem.
Adrian Kennard told ISPreview.co.uk:
“As an ISP that OFCOM did not monitor, AAISP is one of the few that take loss and latency seriously on a day to day basis. We monitor every line every second, record loss and latency and provide that data to our customers in real time and with history for the lifetime of the line.
We have also tried to get BT Wholesale to recognise these key metrics and have, for many years, tried to get BTW to handle fault reports of high latency of packet loss as a type of fault in their own right. To date BTW still do not recognise these as fault types even though they are easy to measure.”
It should be said that the type of exceptionally high packet loss and latency that Adrian is talking about would generally be well above the real-world examples mentioned in this article and easily enough to impact the general usability of a customers connection. This is not to be confused with Ofcom’s examples, which show fairly normal behaviour.