[Disksim-users] Re: Disksim-users Digest, Vol 23, Issue 10

[Youngjae Kim]

Dear Yipkei

Thanks for your help. But I used fast writes and combined seq. writes and so on. But the throughputs of writes and reads wouldn't be the similar. The writes are slower than the reads in the throughput. They should be almost the same. Shouldn't they? 

Thanks. 
Youngjae


On Fri, Aug 24, 2007 at 12:00:07PM -0400, disksim-users-request at ece.cmu.edu wrote:
> Send Disksim-users mailing list submissions to
> 	disksim-users at ece.cmu.edu
> 
> To subscribe or unsubscribe via the World Wide Web, visit
> 	https://sos.ece.cmu.edu/mailman/listinfo/disksim-users
> or, via email, send a message with subject or body 'help' to
> 	disksim-users-request at ece.cmu.edu
> 
> You can reach the person managing the list at
> 	disksim-users-owner at ece.cmu.edu
> 
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of Disksim-users digest..."
> 
> 
> Today's Topics:
> 
>    1. Write throughput (Youngjae Kim)
>    2. Re: Write throughput (Yipkei Kwok)
> 
> 
> ----------------------------------------------------------------------
> 
> Message: 1
> Date: Fri, 24 Aug 2007 08:04:10 -0400
> From: "Youngjae Kim" <youkim at cse.psu.edu>
> Subject: [Disksim-users] Write throughput
> To: <disksim-users at ece.cmu.edu>
> Message-ID: <002501c7e648$c96016e0$0202fea9 at genikims>
> Content-Type: text/plain; format=flowed; charset="iso-8859-1";
> 	reply-type=original
> 
> Dear all,
> 
> I tested write/read throughput with Barracuda disk (barracuda.diskspecs, 
> barracuda.parv) for synthetic workloads. The configuration for my test and 
> results are as follows.
> 
> 1. Config.
> 
> I varied sequentiality (0-1). Inter-arrival time is set to be very small 
> enough to fill up the queue all the time. The request size is 8 blocks 
> (8*512=4KB). I tested every case till the program is aborted because of 
> queue filled up.
> 
> 2. Results
> 
> Regardless of sequentiality, write throughputs are almost the same. I don't 
> know why this happens. IOPS is around 100. However, the read throughput 
> varies according to the sequentiality, from 100 to 200.
> 
> Do you have any idea of why write throughput is smaller than read throughput 
> and why write throughput does not change with the sequentiality of the 
> workloads? And in the implementation, is there any additional overhead to 
> the writes different from read ones?
> 
> Thank you.
> YJ,
> 
> 
> 
> 
> ------------------------------
> 
> Message: 2
> Date: Fri, 24 Aug 2007 08:58:58 -0600
> From: "Yipkei Kwok" <yipkeikwok at gmail.com>
> Subject: Re: [Disksim-users] Write throughput
> To: "Youngjae Kim" <youkim at cse.psu.edu>
> Cc: disksim-users at ece.cmu.edu
> Message-ID:
> 	<f82da4e80708240758t2356594er6475635535ec2810 at mail.gmail.com>
> Content-Type: text/plain; charset=ISO-8859-1
> 
> Hi Youngjae,
> 
> There are several write-related parameters in the diskspecs file.
> 
> e.g.
> - Fast write level (i.e. write-back caching)
> - Combine seq writes
> - ... ...
> 
> Do they give some hints? I hope it helps.
> 
> -Yipkei
> 
> On 8/24/07, Youngjae Kim <youkim at cse.psu.edu> wrote:
> > Dear all,
> >
> > I tested write/read throughput with Barracuda disk (barracuda.diskspecs,
> > barracuda.parv) for synthetic workloads. The configuration for my test and
> > results are as follows.
> >
> > 1. Config.
> >
> > I varied sequentiality (0-1). Inter-arrival time is set to be very small
> > enough to fill up the queue all the time. The request size is 8 blocks
> > (8*512=4KB). I tested every case till the program is aborted because of
> > queue filled up.
> >
> > 2. Results
> >
> > Regardless of sequentiality, write throughputs are almost the same. I don't
> > know why this happens. IOPS is around 100. However, the read throughput
> > varies according to the sequentiality, from 100 to 200.
> >
> > Do you have any idea of why write throughput is smaller than read throughput
> > and why write throughput does not change with the sequentiality of the
> > workloads? And in the implementation, is there any additional overhead to
> > the writes different from read ones?
> >
> > Thank you.
> > YJ,
> >
> >
> > _______________________________________________
> > Disksim-users mailing list
> > Disksim-users at ece.cmu.edu
> > https://sos.ece.cmu.edu/mailman/listinfo/disksim-users
> >
> 
> -- 
> Web site:
> http://360.yahoo.com/yipkeikwok
> http://www.mcs.csueastbay.edu/~ykwok/
> email addresses:
> yipkeikwok at gmail dot com
> ykwok2 at miners dot utep dot edu
> ICQ UIN:    2309842
> Google Talk ID:    yipkeikwok at gmail dot com
> Yahoo! Messenger ID:    yipkeikwok
> MSN Messenger ID:    yipkeikwok at hotmail dot com
> 
> 
> ------------------------------
> 
> _______________________________________________
> Disksim-users mailing list
> Disksim-users at ece.cmu.edu
> https://sos.ece.cmu.edu/mailman/listinfo/disksim-users
> 
> 
> End of Disksim-users Digest, Vol 23, Issue 10
> *********************************************