Comments on: Help me graph lab utilization
http://ask.metafilter.com/120251/Help-me-graph-lab-utilization/
Comments on Ask MetaFilter post Help me graph lab utilizationWed, 22 Apr 2009 17:30:58 -0800Wed, 22 Apr 2009 17:30:58 -0800en-ushttp://blogs.law.harvard.edu/tech/rss60Question: Help me graph lab utilization
http://ask.metafilter.com/120251/Help-me-graph-lab-utilization
Please help me make more meaningful graphs for my boss! Our department runs computer labs on campus, and being the end of the school year we're looking at utilization rates. I can give him: number of users served per hour; number of machines used per hour; and average session time per hour. What he also wants is some sort of utilization, or fill factor graph. But none of us can really figure out what formula to use. <br /><br /> What do I mean by utilization/fill factor? Consider this example: Suppose a lab has 60 seats. Just by chance, 30 users come in sequentially, each using a different machine, for 2 minutes. That is:<br>
<br>
* 00:00 - 00:02: user A uses machine01.<br>
* 00:02 - 00:04: user B uses machine02.<br>
* 00:04 - 00:06: user C uses machine03.<br>
<br>
etc. Now, obviously the lab wasn't very utilized. But the graph will show we serviced 30 people in that hour, using 30 machines! We could have happily serviced all 30 people with one machine, however, the you can't see that from the graph.<br>
<br>
That's a pathological case, but I think it illustrates what I mean. How do we measure this in a meaningful way? It seems to me that with the variables I have (login_count, login_average, host_count) I should be able to somehow derive a utilization but I'm stuck.<br>
<br>
Note: I also have complete control over the data source and how these variables are built. So if I need to get a different number, or generate numbers a different way, please let me know!post:ask.metafilter.com,2009:site.120251Wed, 22 Apr 2009 17:14:31 -0800sbutlerstatisticsutilizationfillfactorgraphBy: polyglot
http://ask.metafilter.com/120251/Help-me-graph-lab-utilization#1720915
utilisation = (average session time) * (sessions per hour) / (computer count)comment:ask.metafilter.com,2009:site.120251-1720915Wed, 22 Apr 2009 17:30:58 -0800polyglotBy: kcm
http://ask.metafilter.com/120251/Help-me-graph-lab-utilization#1720920
You used one computer hour of a total of thirty hours available. polyglot's formula then works.comment:ask.metafilter.com,2009:site.120251-1720920Wed, 22 Apr 2009 17:38:46 -0800kcmBy: jeoc
http://ask.metafilter.com/120251/Help-me-graph-lab-utilization#1720927
Have you thought about making your denominator the total number of user-minutes available? So if you have 60 seats, and your lab is open 8 hours per day, you would have a total of 28,800 user minutes available. <br>
<br>
Or: Number of seats in the lab x number of minutes the lab is open per day<br>
<br>
In your example of 30 people using machines for 2 minutes each, they took up 60 user-minutes. So your fill graph would show that day had a "fill factor" of 0.2%. <br>
<br>
Or: (Sum of each users time on each computer/Total number of user-minutes) x 100<br>
<br>
This can be adapted for whichever length of time your supervisor wants it reported (weekly, monthly, etc. ) by calculating the number of minutes in that timeframe.comment:ask.metafilter.com,2009:site.120251-1720927Wed, 22 Apr 2009 17:42:22 -0800jeocBy: Ironmouth
http://ask.metafilter.com/120251/Help-me-graph-lab-utilization#1720956
What are the figures for? If they are for convincing outside funders the importance of what your office does, use a formula likely to provide data supporting your case.comment:ask.metafilter.com,2009:site.120251-1720956Wed, 22 Apr 2009 17:59:41 -0800IronmouthBy: polyglot
http://ask.metafilter.com/120251/Help-me-graph-lab-utilization#1720957
jeoc's formula is the same as mine just rearranged. You can rearrange it as much as you want and in terms of absolute utilisation or in terms of utilisation rates.<br>
<br>
utilisation = (total session time) / (computer-time available)<br>
(sessions per hour) = (session count) / hours<br>
(average session time) = (total session time) / (session count)<br>
(total session time) = (average session time) * (session count)<br>
(computer-time available) = (computer count) * hours<br>
utilisation = (average session time) * (session count) / ((computer count) * hours) = <br>
= (average session time) * (sessions per hour) / (computer count)<br>
<br>
You can cut it any which way you want.comment:ask.metafilter.com,2009:site.120251-1720957Wed, 22 Apr 2009 18:00:58 -0800polyglotBy: nyc_consultant
http://ask.metafilter.com/120251/Help-me-graph-lab-utilization#1721014
Your manager cares about costs. The drivers of cost are<br>
a) number of computers (capital expense)<br>
b) hours that they are on (electricity)<br>
c) hours that the lab is open (electricity, lab worker costs)<br>
<br>
Does he need to buy more computers? Should he shift computers from lab A to lab B? Can he get rid of some computers? Should he develop some way to turn off computers to save on power, since 90% of them are not being used at any given time? Should he close the lab at some times?<br>
<br>
Y-axis: users at time t<br>
X-axis: time (daily, weekly, monthly, quarterly, yearly - take your pics as to what looks best)<br>
<br>
line 1: simultaneous users at time t<br>
line 2: capacity of that lab (horizontal line)<br>
<br>
Continue pulling on the thread - there's lots of ways to push deeper on this.comment:ask.metafilter.com,2009:site.120251-1721014Wed, 22 Apr 2009 18:59:40 -0800nyc_consultantBy: Jeanne
http://ask.metafilter.com/120251/Help-me-graph-lab-utilization#1721044
(Minutes used/Total number of available minutes) is a good statistic, and probably the main one that you're going to want to use, but I think it might be useful to break it down by hour as well as by a larger increment (that is, usage from 4:00 to 5:00, usage from 5:00 to 6:00, etc.) If your computer lab is utterly overflowing at 4:30 in the afternoon, it doesn't matter that there's a stretch from 8:00 to 10:00 in the morning when nobody was using the computers. My experience with public library computer labs is that, in terms of customer service, the problem isn't the average wait time to get a computer; the problem is the <i>worst</i> wait time to get a computer.comment:ask.metafilter.com,2009:site.120251-1721044Wed, 22 Apr 2009 19:30:38 -0800JeanneBy: Gyan
http://ask.metafilter.com/120251/Help-me-graph-lab-utilization#1721050
Besides occupied/available user-minutes, you also want to look at trends in simultaneous use. If a lab is hosting classes with <i>x</i> students, it doesn't matter if only 3 students, on average, use it the rest of the time, you still need <i>x</i> computers in that lab.comment:ask.metafilter.com,2009:site.120251-1721050Wed, 22 Apr 2009 19:35:19 -0800GyanBy: sbutler
http://ask.metafilter.com/120251/Help-me-graph-lab-utilization#1721126
Ahh... thanks guys! Total computer time available is the part I was missing. I'll probably be graphing it on a per hour basis since that's the kind of breakdown we're interested in for the other parts.<br>
<br>
To answer some other questions:<br>
<br>
- I'm already graphing users served per hour and computers in use per hour.<br>
<br>
- We're using the graphs for multiple purposes/multiple audiences. Mostly to justify lab expenses, decide lab hours, and staffing level (some labs are double staffed).comment:ask.metafilter.com,2009:site.120251-1721126Wed, 22 Apr 2009 21:12:55 -0800sbutler