From raysmile at tom.com Thu Nov 16 10:39:26 2006 From: raysmile at tom.com (=?gb2312?B?zO/A2g==?=) Date: Thu, 16 Nov 2006 23:39:26 +0800 Subject: [Disksim-users] somebody knows how to handle the switch of the disk status:active/idle/standby/sleep? Message-ID: <200611162339241563639@tom.com> Hi, I am working on augmenting the support of the switch of the disk status, such as active, idle, standby, sleep in the disksim. And I have a problem. If the current status of a disk is sleep while an I/O request arrives, the disk can not serve the request immediately and must wait for a period, call wake-up time, to serve it. So, the events in terms of the disk should be delayed till the disk is ready waked-up. considering the RAID configuration, only the corresponding events of the waking-up disks should be delayed and the other events for the other disks which are in the active/idle status should be handled without delay. the question is: how can I do something in the disksim, to differentiate the events' cause by the different disks. If I block all of the events in the event loop once a disk is in the sleep status, it leads to an unfair and wrong result. any suggestion is appreciated. Lei Tian -------------- next part -------------- An HTML attachment was scrubbed... URL: From bucy at gloop.org Tue Nov 21 14:02:22 2006 From: bucy at gloop.org (John Bucy) Date: Tue, 21 Nov 2006 11:02:22 -0800 Subject: [Disksim-users] somebody knows how to handle the switch of the disk status:active/idle/standby/sleep? In-Reply-To: <200611162339241563639@tom.com> References: <200611162339241563639@tom.com> Message-ID: <4fc1e0430611211102w31f8f3e2m549df176b89590f0@mail.gmail.com> So your setup is that you have a bunch of disks and at any given time, some of them are active and some of them are sleeping or whatever and you want the requests that map to disks that are on to issue immediately and the ones that map to disks that are asleep to be delayed for the disk to wake up? I would expect that you'd have to implement that in the logorg code in disksim. Maybe you could introduce a special "wakeup" request that you'd enqueue ahead of any request that accesses a disk that's asleep and arrange for it to block everything else in the queue, etc. john On 11/16/06, ?? wrote: > > > Hi, > > I am working on augmenting the support of the switch of the disk status, > such as active, idle, standby, sleep in the disksim. And I have a problem. > If the current status of a disk is sleep while an I/O request arrives, > the disk can not serve the request immediately and must wait for a period, > call wake-up time, to serve it. So, the events in terms of the disk should > be delayed till the disk is ready waked-up. considering the RAID > configuration, only the corresponding events of the waking-up disks should > be delayed and the other events for the other disks which are in the > active/idle status should be handled without delay. > the question is: how can I do something in the disksim, to differentiate > the events' cause by the different disks. If I block all of the events in > the event loop once a disk is in the sleep status, it leads to an unfair and > wrong result. > any suggestion is appreciated. > > > Lei Tian > _______________________________________________ > Disksim-users mailing list > Disksim-users at ece.cmu.edu > https://sos.ece.cmu.edu/mailman/listinfo/disksim-users > > > From agusgogon at yahoo.com Thu Nov 23 21:46:26 2006 From: agusgogon at yahoo.com (agus =gOn=) Date: Thu, 23 Nov 2006 18:46:26 -0800 (PST) Subject: [Disksim-users] Disksim 3.0 installation in Ubuntu 6.06 Dapper Drake -Successful Message-ID: <20061124024626.61194.qmail@web60319.mail.yahoo.com> Dear All, For your info. After trying and facing the errors during install disksim3.0 in ubuntu 6.06 ,May be I can share the steps : # sudo apt-get install build-essential { this is for installing make command) Then remove the packace : gcc, gcc-4.0, cpp, cpp-4.0. You can use Synaptic (System ? Administration ? Synaptic Package Manager ) to uninstall Then install : cpp-2.95, gcc-2.95, bison, flex-old (do not use flex) and indent. (You can download indent from http://kambing.vlsm.org/ubuntu/pool/main/i/indent/ , and searching for others in this link too) then run #make CC=gcc-2.95 in disksim folder then run # ./runvalid in ../valid folder This is my step. Hopefully help someone who face same problem. thks Agus gogon ____________________________________________________________________________________ Cheap talk? Check out Yahoo! Messenger's low PC-to-Phone call rates. http://voice.yahoo.com -------------- next part -------------- An HTML attachment was scrubbed... URL: From raysmile at tom.com Tue Nov 28 10:07:23 2006 From: raysmile at tom.com (=?GB2312?B?zO/A2g==?=) Date: Tue, 28 Nov 2006 23:07:23 +0800 Subject: [Disksim-users] should the NULL_EVENT and BUS_DELAY_COMPLETION events be delayed during the wake-up period of a disk? Message-ID: <200611282307167654925@tom.com> Hi, I encount a problem on augmenting the support of the switch of the disk status, such as active, idle, standby, sleep in the disksim. I can delay some events during the wake-up period of a disk till the disk is really waken up. But if I delay the NULL_EVENT event and the BUS_DELAY_COMPLETION event, the disksim will output some errors as follows and stops. NULL_EVENT event: Trace event appears out of time order in trace - simtime 34219.413995, time 32309.570000 ioscale 1.000000, tracebasetime 0.000000 devno 0, blkno 61481, bcount 3, flags 0 BUS_DELAY_COMPLETION event: diskno(14):Cannot find ioreq in disk_buffer_update_outbuffer I think BUS_DELAY_COMPLETION's error can be ignored because this event should not be delayed. But how about the NULL event? If I do not delay this event, the disksim can work well. Should it be delayed or not? any hints are appreciated, and thank John Bucy for his answer to my last question. :) Lei Tian -------------- next part -------------- An HTML attachment was scrubbed... URL: