Home > Intel Communities > Open Port IT Community > IT@Intel > Blog > 2008 > March > 10
Currently Being Moderated
2

Performance: data to enable comparison

Posted by Tomas Mcinerney on Mar 10, 2008 3:54:12 PM

So we are on the home run of deploying the new pilot cube

environment, in fact I’m on site helping supporting day one move in at our

third US site installation which has certainly been interesting. Flight over

went quickly, though at some points it was rather roller coaster (to the point

coffee was spilt on laps)

 

But I digress…

 

 

 

I wanted to discuss an item I have brought up before;

benchmarking. The project has moved on and worth asking some questions around.

Intel IT has used classic benchmarking applications to compare platforms when

going to RFP (using standard off the shelf applications) but we discovered this

testing wasn’t helping us improve the performance of our software on the client

it was simply giving us faster clients (not a bad thing) We were missing some

critical decision making criteria for evaluating newer versions of

applications, client builds or software tweaks (identifying performance improvement

or impact) As we drive towards more out of the box applications we will also be

using the tool to evaluate impact on the environment.

 

 

 

So we kicked off a project to begin recording certain

productivity metrics to evaluate user perception performance; not necessarily

aimed at just understanding how fast each client is; but more what impact it

has to users

 

 

 

Some of these timing metrics include

 

 

 

  • Time into operating system

  • Time into email application (first email)

  • Time into first instant message conversation

  • Time to first spreadsheet/document application

 

Once changes are made to the client build or application

stack an impact is recorded through the metrics. This means we can start to set

goals and performance targets (10% faster build in 3 months…etc)

 

We hope to publish this data with some fellow travellers to get

some indicators on quantify the overhead an ‘IT’ build compared to an off the

shelf build (we classify it as vanilla OS)

 

 

 

Are you recording productivity metrics to compare

applications and build generations? Any thoughts on if this data would be

useful to you?

 

 



Add a comment Leave a comment on this blog post.
Mar 19, 2008 12:15 AM Guest Phil T  says:

Do you plan to benchmark against "aged" systems as well as newly-built systems? My experience is that these metrics would be relatively quick on a freshly built client, but would take considerably longer a year later or 3 years later. There is little doubt in my mind that colliding security processes and out-of-date "on-load" scripts exacerbate this ongoing performance degradation. Great idea. Suggest that experimental controls be published to make apples and apples comparisons possible.

Apr 2, 2008 8:53 AM Tomas Mcinerney Tomas Mcinerney    says in response to Phil T:

@ Phil T

 

Sort answer Yes - we are exploring how to compare aged systems - a lot more products in the market offer this kind of capability so it's not such a baron world. However many don't run 'user experience' scripts - so that’s were our focus is. The challenge we have, as I'm sure you understand, is once a machine has been with a user for some time they have a lot of different options to slow it down (install new apps, lots of data, toolbars etc) so we are also looking at what we define an 'aged' system to be.