How to Measure DSL Speed

dsl speedMeasuring DSL speed is not nearly as simple of a task as it might seem.  In fact, there are companies that go to incredible lengths just to ensure that their measurement systems are accurate and correct before they conduct tests.  These companies then conduct tests on their own tests to ensure that there are no factors being overlooked or counted that should not be, or any factors that are discounted or ignored.  If you want to measure DSL speed at home or in the office, here are some great tips:

Ask Why You Want to Measure DSL Speed

If you are measuring DSL speed simply because you are bored then you might need a hobby or something other to do than just endlessly run DSL tests!  If you want to see how one service compares against another or if one computer or device is faster than another, then you want to look at comparative benchmarks.  If you are looking for application-specific performance, then you probably have a certain application or suite of applications that are mission critical to you.  No matter why you want to know how fast your DSL is, you want to be sure that you understand why because it will really dictate how far you are going to go on some tests.  After all, some tests are not necessarily practical for some environments.  With that said, let’s dive in…

Going Vanilla to Measure DSL Speed

Techs around the world know what vanilla means, even if they do not know what vanilla is!  Simply put, vanilla is a term used to describe removing as many possible factors from the equation; this may also be known as isolating the variables as much as possible.  How does it work?  Well, that depends on the purpose and how serious the test is.  Serious researchers and people writing about technology may have it within their power to test multiple routers, modems, network cards, and install an operating system fresh for each trial.  That might be a little bit too extreme for a small office trying to determine if the upgraded DSL speeds are going to provide results that can be taken to the bank or if that extra $50 a month is better spent elsewhere.

This is why knowing the ‘why’ of the measurement is important because going completely vanilla can take an entire network offline, but if the test is that serious then do that.  Reset everything to factory defaults, download and apply all patches, and reset all network-related settings to the point where they are complete and correct.  Those without the ability to go vanilla have to ask themselves some questions.  For example, let’s take the office that paid $50 a month extra to double their broadband speed.  They cannot take one computer offline and wipe it for a few reasons, but mostly because they need people working on those computers to make money.  Another problem exists with that scenario as well, and it is that isolating all computers but one on the network might not be a valid test.

Consider for a moment a busy office with an overworked router.  Internal file servers and backups keep data flowing across the network, and this traffic alone could be impacting how much data flows beyond the network borders.  After all, any hardware or software firewall may be just as taxed by all this extra traffic, so you could pull the other computers offline and see a huge performance gain on some critical measurement but then that gain would be instantly nullified when you return to the real world environment.  Perhaps best of all, such a scenario may help you pinpoint flaws in network layout and usage, but it may not give you the answers you seek.

Synthetics vs. Real World

So, once the network is troubleshot and everything is working correctly you still have to decide what you want to measure.  Do you want to measure broadband performance from a key application or two, or are you going to measure generic performance with a so-called synthetic benchmark.  Again, we come back to the question of why you want to test.  Some applications like streaming require high reliability and high data rates.  Other applications like low ping times.  Knowing what an application wants is important to knowing what type of test results to run and look for.  You may ultimately want to look at both real-world and synthetic benchmarks to get the best feel for performance.