The Mystery that is Speed Measurement of PCs (and Mobile Devices!)
For such a common concept that seems to be so widely understood, speed is actually a far bigger conceptual enigma for PCs (and mobile devices) and even more so for enterprise endpoints. Let’s take a look at some other industries where “speed” is important to service levels, function, or comparison, and how those benchmarks are calculated and reported.
The automotive industry uses “Zero to sixty” times.
- While acceleration and other factors undoubtedly come in to play this is pretty standard.
Football scouting uses “40 yard dash” times.
- There are also other metrics such as closing time for a Cornerback catching up to a Wide Receiver.
The fast food industry uses “mean & median drive-thru wait time”.
- This is simply I/O and doesn’t necessarily take into account proper cooking times (e.g. rushing to get French Fries done in the name of speed, etc.)
This begs a question: what is the de facto metric for speed (or performance) for a computer? Seems straightforward enough of a question but strangely answers remain elusive. Often boot times, shutdown, and application launch speeds are quoted. Other times admins suggest network latency and claim the bottlenecks are out in the cloud or on the network. Some will invariably quote hardware specs such as chip speeds, amount of installed RAM, and disk RPMs. However, as time has shown those numbers are a simply an indicator of initial machine state at best and as we have seen those shiny, new computers with “2.1 GHz chips” can quickly become sluggish and generate the same old support calls where users exclaim “my PC is slow”. Ziften proposes that the fact is simple: in the 30+ years that the PC has been in widespread commercial use there still is no de facto metric for speed/performance.
Now we have thoroughly confused everybody by postulating that there truly is no metric for speed/performance. This begs a 2nd question: if we were to decide which dimension this metric exists in, what is the target value of the metric for my enterprise users or what should a histogram of this metric look like across an enterprise user base?
To Quote Indiana Jones and the Lost Ark perhaps the industry is “digging in the wrong place”?
See next week’s blog that poses an answer for some possible reasons for this disconnect and also offers some suggestions to help solve this dilemma.