[W12.bis.dssnp] 2nd xodx dssnp teleconf - summary (2012-12-17)

[Jonas Smedegaard]

Quoting Natanael Arndt (2013-01-07 11:41:03)
> Am 07.01.2013 10:55, schrieb Jonas Smedegaard:
> > Hi all,
> >
> > Quoting Natanael Arndt (2013-01-07 00:58:04)
> >> We asked our selves how to messure the resource consumption of xodx 
> >> to make sure the software matches the requirements formulated on 
> >> [2] or are those only relevant for the hardware?
> > I'd say those requirements are apply equally well to hardware and 
> > software: We want slowest possible hardware to help keep price and 
> > power consumption low, which means we want the least possible 
> > software activities pushing the demands for yet more powerful 
> > hardware.
> That is, what I have expected as intent of the "Performance 
> requirements": "It must be able to execute all Freedom Box software at 
> less than half of the CPU load".

Yeah, I did wonder why you raised that as a question.  Sorry if I 
misunderstood your email.


> > I was sysadmin at an advertising agency many years ago when internet 
> > connections was done with dialup modems. The agency folks started 
> > engaging in this new "web design" and had read a study saying that 
> > consumers lost interest if a web page took more than 20 seconds to 
> > load. Their resonse was to carefully design web pages so heavy that 
> > with a common modem speed took 19 seconds to load.
> >
> > You see the parallel?  It is craaaazy to aim for "highest possible" 
> > - aim for lowest possible, and the box will have _extra_ breathing 
> > room which _improves_ stability and/or allows our user to activate 
> > more or different functionality than what you'd expected in your 
> > calculations.
> Do you have an idea of, how we can measure the performance? Since it 
> is an application which answers to requests it has some peeks on 
> request and the average load is highly dependent on the frequency of 
> request which can extremely variate. Or do you think measuring the 
> page loading time is a good idea?

If I knew something solid to measure, I would be a millionaire ;-)

The kind of people best familiar with relevant measures are probably 
sysadmins.  Perhaps ask that question to sysadmins at your university - 
and then try (alone or with them) to translate that to the smaller and 
differently equipped ARM devices you play with.

Generally, it is about the various pipelines, and leaving room for 
spikes - RAM/DISK IO, RAM/DISK space, IRQs etc.

Tools like iftop, htop, atop and powertop are useful measurement tools 
to get a feel for a live running system (I especially like htop for 
being interactive, and the rich details of atop).

There's also a slew of statistics monitoring and gathering tools, but 
personally I am less familiar with those (it is still on my TODO to 
setup reasonable surveillance of the networks I am syadmin for).  The 
tool I expect to deploy myself is collectd, so I suggest you try that 
before e.g. nagion, nmon or others.


> > Keep up this great work - I am excitedly following!
> >
> > Oh, and please consider sharing your progress at the FreedomBox 
> > mailinglist and/or blog about it: I am sure others are excited too 
> > to learn what you're exploring (not only the results, but also your 
> > process!).
> Thank you. Do you think we should post status reports to the 
> FreedomBox list or just invite the people there to follow our list?

Well, that depends on your aim ;-)

If we are playing over here, they are playing over there, an you 
are running back and forth playing both places, then would you invite us 
over or tell us about your experiences over there?

Inviting us over would be lighter (less "noisy") on us, but also put the 
burden on us for synergy to happen.

Telling us would be "noisy" at our camp but we wouldn't need to lift our 
lazy asses to gain from your learning.

So if you want more of us to take part in yours, then I see no other way 
than inviting us over.  But if you wanna share, then I see no need for 
us to get excercise over it (and risking us being too lazy for that).

Or to put it differently: Don't mind the noise - we all joined that 
mailinglist in order to be entertained by relevant info - which then 
really isn't "noise"!


 - Jonas

-- 
 * Jonas Smedegaard - idealist & Internet-arkitekt
 * Tlf.: +45 40843136  Website: http://dr.jones.dk/

 [x] quote me freely  [ ] ask before reusing  [ ] keep private
-------------- nächster Teil --------------
Ein Dateianhang mit Binärdaten wurde abgetrennt...
Dateiname   : nicht verfügbar
Dateityp    : application/pgp-signature
Dateigröße  : 836 bytes
Beschreibung: signature
URL         : <http://lists.informatik.uni-leipzig.de/pipermail/w12.bis.dssnp/attachments/20130107/ec794761/attachment.asc>