All too often, people tend to measure whatever is easiest to measure rather than what matters most.  Examples range from health (body weight, nutrition, etc.) to technical fields such as IT. 

Easy Answers

When I am attempting to “test’ the bandwidth of a system or network connection, I often find myself using on of the common free online tests like Speedtest.net.  It usually runs quickly and requires no configuration.  But what do the results really mean?  Below is an example of a recent test result.

http://www.speedtest.net/result/2115776971.png

But what does this really mean in the real world?  First off, the automatic server selection process favors the server that is “closest” (from a network architecture standpoint) to me.  Generally, the results will give me the best possible speed and path and can be considered a theoretical maximum.  But, I rarely connect to resources on my ISP’s core network.  Rather, almost everything I do requires routing outside of the ISP’s boundaries.  That’s where arrangements like Internet Peering and Content Delivery Networks (CDNs) can make a huge difference.  In this case, the easiest answer is clearly not the best one…

Better Answers

What I really want to know is how well I can connect to “real” online applications and services, ranging from Netflix to Office 365.  I want my Xbox Live connection to have a low latency, and I want to make sure that performance doesn’t vary dramatically during the day.  That’s where more specific tests become important.  Many online content and application providers have their own tests.  You can often find them by doing a basic web search. 

Example: Testing Office 365 Performance

Performance and reliability are among the foremost concerns for most IT professionals that are consider moving some applications and services to the cloud (that is, network infrastructure that they do not completely control).  This often introduces numerous variables, but technical (bandwidth, latency, routing, quality of service) and not-so-technical (quality of support personnel, investments in the network, priority of each customer, etc.)  Even the best implementations can fail if the end-user experience is poor based on limited bandwidth or high latency. 

As an example of a more “Real World” (and therefore relevant) test, I want to highlight Microsoft Online Services’ Performance Test.  This set of online tests takes into account bandwidth, latency, routing, and related parameters to give you a good idea of how well your experience with Microsoft’s Online services will be (from a performance standpoint, at least).  Below is a portion of the “Speed” test result:

image

This clearly shows that I’m not getting my maximum stated bandwidth (~32Mbps down / 3.0Mbps up), but the performance definitely looks good enough for basic usage. 

The tests also measure other important statistics, such as packet loss, round-trip time, packets per second, and related characteristics.  All of this yielded the following summary:

image

Of course, performance is likely to vary at different dates and times (I happened to perform this test on a Sunday afternoon).  If you want some additional detail on the tests, see the blog post titled Moving your customers to BPOS or Office 365? Check their BANDWIDTH!.  And, feel free to try the test yourself if you’re considering moving yourself and/or your  users to Microsoft Office Online.