[cadynce] On being able to pace periodic application strings... (Comparing Linux 2.6.9 and 2.6.20-rt8)

[James Hill]

Hey Gautam, 

> -----Original Message-----
> From: cadynce-bounces at list.isis.vanderbilt.edu 
> [mailto:cadynce-bounces at list.isis.vanderbilt.edu] On Behalf 
> Of Gautam Thaker
> Sent: Thursday, May 10, 2007 12:34 PM
> To: cadynce at list.isis.vanderbilt.edu
> Subject: [cadynce] On being able to pace periodic application 
> strings... (Comparing Linux 2.6.9 and 2.6.20-rt8)
> 
> Hi Adam, Raytheon and James, (and others):
> 
> One thing you may want to be careful of is how you actual 
> subject the application strings to the update rates that you 
> want. 

Maybe I missed something, but who said we are makin changes to the
workloads??

Thanks,

James

> If one looks at message arrival rates in:
> 
> https://escher.isis.vanderbilt.edu/twiki/pub/CADYNCE/OneFiftyF
> iftyCase1/WorkloadDescription.htm
> 
> one observes that there are some non well-rounded values for 
> message rates. In any event, one wants to have good control 
> over rate at which we inject messages into our experiments. 
> In our tests ultimately we rely on 'nanosleep()' to 
> accurately pace our message rates. (this or similar must be 
> what CUTS based models use, James?) Please see the following 
> link that shows if you use non-RT Linux kernel you may get 
> some strange results. It is for this reason, among others, 
> that  we test w/ RT Linux. 
> PoR is also  using RT Linux.
> 
> http://www.atl.external.lmco.com/projects/QoS/RTOS_html/timer_
> gran/linux_2.6.9_and_2.6.20-rt8_compared.html
> 
> Gautam
> _______________________________________________
> Cadynce mailing list
> Cadynce at list.isis.vanderbilt.edu
> http://list.isis.vanderbilt.edu/mailman/listinfo/cadynce
>