Issue 11552: GRM::SchedulableResource should not have a property "host:ExecutionHost" (marte-ftf) Source: THALES (Mr. Sebastien Demathieu, sebastien.demathieu(at)thalesgroup.com) Nature: Revision Severity: Minor Summary: In the class diagram, the GRM::SchedulableResource should not have a property "host:ExecutionHost" (defined in QGAM). There is no dependency from GRM to GQAM, therefore the "host" property cannot be typed by "ExecutionHost". Moreover, GRM::SchedulableResource already has a property called "host: Scheduler". In that context, I would suggest just removing the "host: ExecutionHost" property from SchedulableResource in Figure 15.5. Resolution: To add an inheritance from scheduler in figure 15.5 for both executionHost and communicationHost. Then the association between SchedulableResource and executionHost is the same that already exist in GRM to scheduler. The scheduler in communicationHost is easy to justify as the arbitration/protocol that gives to each message access to the media. Revised Text: New visio file for fig 15.5 and 15.9. Add the inheritance from scheduler in the class descriptions this means editing sections: 15.3.2.4, 15.3.2.7 adding the generalization Scheduler (from MARTE::GRM) Also indicate in the constraints the two self pointing associations processingUnits and mainScheduler Actions taken: October 9, 2007: received issue February 17, 2010: closed issue Discussion: End of Annotations:===== m: webmaster@omg.org Date: 09 Oct 2007 07:21:50 -0400 To: Subject: Issue/Bug Report -------------------------------------------------------------------------------- Name: Sébastien Demathieu Company: Thales mailFrom: sebastien.demathieu@thalesgroup.com Notification: Yes Specification: A UML Profile for MARTE Section: 15/2 FormalNumber: realtime/07-08-04 Version: Beta 1 RevisionDate: 08/2007 Page: 267 Nature: Revision Severity: Minor HTTP User Agent: Mozilla/5.0 (Windows; U; Windows NT 5.1; fr; rv:1.8.1.7) Gecko/20070914 Firefox/2.0.0.7 Description In the class diagram, the GRM::SchedulableResource should not have a property "host:ExecutionHost" (defined in QGAM). There is no dependency from GRM to GQAM, therefore the "host" property cannot be typed by "ExecutionHost". Moreover, GRM::SchedulableResource already has a property called "host: Scheduler". In that context, I would suggest just removing the "host: ExecutionHost" property from SchedulableResource in Figure 15.5. Date: Thu, 13 Mar 2008 15:11:58 -0400 (EDT) From: Murray Woodside Reply-To: cmw@sce.carleton.ca To: marte-ftf@omg.org Subject: about issue 11552 and Arlington discussion Thank you Lonnie for minuting the discussion for those of us that were not there (the link is in your message below). This is just a comment on the connecting of SchedulableResource to ExecutionHost. This essentially comes from SPT, not personally from me. However I do prefer it, and suggest that the Scheduler be associated via the host. It is true that the scheduler and host are a unit. For a succinct description, every host has a scheduler so it is more compact to refer only to the host... and then specify the scheduler in one place. By associating the Schedulable Resource (meaning, a process) to the scheduler first and then through it to the host, you make it possible to specify several schedulers on the same host. If this is intentional I would like to understand the context, because it is axiomatic to me that a resource must be managed by a single scheduler. For performance modeling it would be very odd and not understood, to allocate the processes explicitly to a scheduler. Murray Woodside Distinguished Research Professor Dept of Systems and Computer Engineering, Carleton University, 1125 Colonel By Drive, Ottawa K1S 5B6, Canada. (613)-520-5721.....fax (613)-520-5727....cmw@sce.carleton.ca (http://www.sce.carleton.ca/faculty/woodside.html) ---------- Forwarded message ---------- Date: Thu, 13 Mar 2008 17:38:31 -0000 From: "VanZandt, Lonnie" To: marte-ftf@omg.org Cc: "Edwards, Luke" Subject: MARTE FTF Minutes available at our wiki http://www.omgwiki.org/marte/doku.php?id=marteftfminutes_2008mar13 We spoke much more than I wrote down. Therefore, those of you who attended are welcome to edit the wiki page to add any notes you recorded which I left out. Lonnie VanZandt Field Applications Engineer Denver, CO Artisan Software Tools mobile: 720 201-1349 desk: 303 482-2943 lonnie.vanzandt@artisansw.com Subject: RE: about issue 11552 and Arlington discussion Date: Thu, 13 Mar 2008 20:24:30 -0000 X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: about issue 11552 and Arlington discussion Thread-Index: AciFP1Hp6Zt/2k8BScyiK4XHH4SSxAACOwB/ From: "VanZandt, Lonnie" To: , Murray, You are welcome. I also did not intend to blame you with the comment on the ExecutionHost material. Pursue Julio for the detailed commentary on the issue. Lonnie VanZandt Field Applications Engineer Denver, CO Artisan Software Tools mobile: 720 201-1349 desk: 303 482-2943 lonnie.vanzandt@artisansw.com -------------------------------------------------------------------------------- From: Murray Woodside [mailto:cmw@sce.carleton.ca] Sent: Thu 3/13/2008 1:11 PM To: marte-ftf@omg.org Subject: about issue 11552 and Arlington discussion Thank you Lonnie for minuting the discussion for those of us that were not there (the link is in your message below). This is just a comment on the connecting of SchedulableResource to ExecutionHost. This essentially comes from SPT, not personally from me. However I do prefer it, and suggest that the Scheduler be associated via the host. It is true that the scheduler and host are a unit. For a succinct description, every host has a scheduler so it is more compact to refer only to the host... and then specify the scheduler in one place. By associating the Schedulable Resource (meaning, a process) to the scheduler first and then through it to the host, you make it possible to specify several schedulers on the same host. If this is intentional I would like to understand the context, because it is axiomatic to me that a resource must be managed by a single scheduler. For performance modeling it would be very odd and not understood, to allocate the processes explicitly to a scheduler. Murray Woodside Distinguished Research Professor Dept of Systems and Computer Engineering, Carleton University, 1125 Colonel By Drive, Ottawa K1S 5B6, Canada. (613)-520-5721.....fax (613)-520-5727....cmw@sce.carleton.ca (http://www.sce.carleton.ca/faculty/woodside.html) ---------- Forwarded message ---------- Date: Thu, 13 Mar 2008 17:38:31 -0000 From: "VanZandt, Lonnie" To: marte-ftf@omg.org Cc: "Edwards, Luke" Subject: MARTE FTF Minutes available at our wiki http://www.omgwiki.org/marte/doku.php?id=marteftfminutes_2008mar13 We spoke much more than I wrote down. Therefore, those of you who attended are welcome to edit the wiki page to add any notes you recorded which I left out. Lonnie VanZandt Field Applications Engineer Denver, CO Artisan Software Tools mobile: 720 201-1349 desk: 303 482-2943 lonnie.vanzandt@artisansw.com Subject: RE: about issue 11552 and Arlington discussion Date: Fri, 14 Mar 2008 20:53:14 +0100 X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: about issue 11552 and Arlington discussion Thread-Index: AciFP1Hp6Zt/2k8BScyiK4XHH4SSxAACOwB/ADCvjZU= From: "Medina Pasaje, Julio Luis" To: "VanZandt, Lonnie" , , X-OriginalArrivalTime: 14 Mar 2008 19:53:14.0875 (UTC) FILETIME=[0A7D74B0:01C8860D] X-imss-version: 2.050 X-imss-result: Passed X-imss-scanInfo: M:P L:E SM:0 X-imss-tmaseResult: TT:0 TS:0.0000 TC:00 TRN:0 TV:5.0.1023(15788.000) X-imss-scores: Clean:99.90000 C:2 M:3 S:5 R:5 X-imss-settings: Baseline:1 C:1 M:2 S:2 R:2 (0.0000 0.0000) X-MIME-Autoconverted: from quoted-printable to 8bit by amethyst.omg.org id m2EJsGBq030735 Hi Murray, This issue was assigned to GRM, but I have the impresion that the easiest solution is in GQAM, it is to make executionHost inherit not only from ProcessingResource but also from Scheduler, this way the association host from SchedulableResource is legal and refers to the one already in GRM, and both are a unit as expected for most analysis purposes (including the majotity of RMA based scheduling analysis). The rational for having them separated in GRM is that there are not only platforms, and conceptual designs but also analysis techniques that enable hierarquical schedulling and a first strict temporal partitioning of the CPU (like in ARINC 653 for example) before scheduling threads with another policy. So my proposal to solve the issue is just to add an inheritance from scheduler in figure 15.5 for both executionHost and communicationHost. Then the association between SchedulableResource and executionHost is the same that already exist in GRM to scheduler. The scheduler in communicationHost is easy to justify as the arbitration/protocol that gives to each message access to the media. We discussed alternatives with Bran and he agrees on this solution. But we all saw the necessity of discussing it with you to prevent disarrangements. If you considere it useful I may call you to discuss this before going to the next teleconf on tuesday. Cheers, Julio ________________________________ De: VanZandt, Lonnie [mailto:lonnie.vanzandt@artisansw.com] Enviado el: jue 13/03/2008 21:24 Para: cmw@sce.carleton.ca; marte-ftf@omg.org Asunto: RE: about issue 11552 and Arlington discussion Murray, You are welcome. I also did not intend to blame you with the comment on the ExecutionHost material. Pursue Julio for the detailed commentary on the issue. Lonnie VanZandt Field Applications Engineer Denver, CO Artisan Software Tools mobile: 720 201-1349 desk: 303 482-2943 lonnie.vanzandt@artisansw.com ________________________________ From: Murray Woodside [mailto:cmw@sce.carleton.ca] Sent: Thu 3/13/2008 1:11 PM To: marte-ftf@omg.org Subject: about issue 11552 and Arlington discussion Thank you Lonnie for minuting the discussion for those of us that were not there (the link is in your message below). This is just a comment on the connecting of SchedulableResource to ExecutionHost. This essentially comes from SPT, not personally from me. However I do prefer it, and suggest that the Scheduler be associated via the host. It is true that the scheduler and host are a unit. For a succinct description, every host has a scheduler so it is more compact to refer only to the host... and then specify the scheduler in one place. By associating the Schedulable Resource (meaning, a process) to the scheduler first and then through it to the host, you make it possible to specify several schedulers on the same host. If this is intentional I would like to understand the context, because it is axiomatic to me that a resource must be managed by a single scheduler. For performance modeling it would be very odd and not understood, to allocate the processes explicitly to a scheduler. Murray Woodside Distinguished Research Professor Dept of Systems and Computer Engineering, Carleton University, 1125 Colonel By Drive, Ottawa K1S 5B6, Canada. (613)-520-5721.....fax (613)-520-5727....cmw@sce.carleton.ca (http://www.sce.carleton.ca/faculty/woodside.html) ---------- Forwarded message ---------- Date: Thu, 13 Mar 2008 17:38:31 -0000 From: "VanZandt, Lonnie" To: marte-ftf@omg.org Cc: "Edwards, Luke" Subject: MARTE FTF Minutes available at our wiki http://www.omgwiki.org/marte/doku.php?id=marteftfminutes_2008mar13 We spoke much more than I wrote down. Therefore, those of you who attended are welcome to edit the wiki page to add any notes you recorded which I left out. Lonnie VanZandt Field Applications Engineer Denver, CO Artisan Software Tools mobile: 720 201-1349 desk: 303 482-2943 lonnie.vanzandt@artisansw.com Date: Fri, 14 Mar 2008 15:59:54 -0400 (EDT) From: Murray Woodside Reply-To: cmw@sce.carleton.ca To: "Medina Pasaje, Julio Luis" Cc: "VanZandt, Lonnie" , marte-ftf@omg.org Subject: RE: about issue 11552 and Arlington discussion This solution is fine with me. Murray Woodside Distinguished Research Professor Dept of Systems and Computer Engineering, Carleton University, 1125 Colonel By Drive, Ottawa K1S 5B6, Canada. (613)-520-5721.....fax (613)-520-5727....cmw@sce.carleton.ca (http://www.sce.carleton.ca/faculty/woodside.html) On Fri, 14 Mar 2008, Medina Pasaje, Julio Luis wrote: Hi Murray, This issue was assigned to GRM, but I have the impresion that the easiest solution is in GQAM, it is to make executionHost inherit not only from ProcessingResource but also from Scheduler, this way the association host from SchedulableResource is legal and refers to the one already in GRM, and both are a unit as expected for most analysis purposes (including the majotity of RMA based scheduling analysis). The rational for having them separated in GRM is that there are not only platforms, and conceptual designs but also analysis techniques that enable hierarquical schedulling and a first strict temporal partitioning of the CPU (like in ARINC 653 for example) before scheduling threads with another policy. So my proposal to solve the issue is just to add an inheritance from scheduler in figure 15.5 for both executionHost and communicationHost. Then the association between SchedulableResource and executionHost is the same that already exist in GRM to scheduler. The scheduler in communicationHost is easy to justify as the arbitration/protocol that gives to each message access to the media. We discussed alternatives with Bran and he agrees on this solution. But we all saw the necessity of discussing it with you to prevent disarrangements. If you considere it useful I may call you to discuss this before going to the next teleconf on tuesday. Cheers, Julio ________________________________ De: VanZandt, Lonnie [mailto:lonnie.vanzandt@artisansw.com] Enviado el: jue 13/03/2008 21:24 Para: cmw@sce.carleton.ca; marte-ftf@omg.org Asunto: RE: about issue 11552 and Arlington discussion Murray, You are welcome. I also did not intend to blame you with the comment on the ExecutionHost material. Pursue Julio for the detailed commentary on the issue. Lonnie VanZandt Field Applications Engineer Denver, CO Artisan Software Tools mobile: 720 201-1349 desk: 303 482-2943 lonnie.vanzandt@artisansw.com ________________________________ From: Murray Woodside [mailto:cmw@sce.carleton.ca] Sent: Thu 3/13/2008 1:11 PM To: marte-ftf@omg.org Subject: about issue 11552 and Arlington discussion Thank you Lonnie for minuting the discussion for those of us that were not there (the link is in your message below). This is just a comment on the connecting of SchedulableResource to ExecutionHost. This essentially comes from SPT, not personally from me. However I do prefer it, and suggest that the Scheduler be associated via the host. It is true that the scheduler and host are a unit. For a succinct description, every host has a scheduler so it is more compact to refer only to the host... and then specify the scheduler in one place. By associating the Schedulable Resource (meaning, a process) to the scheduler first and then through it to the host, you make it possible to specify several schedulers on the same host. If this is intentional I would like to understand the context, because it is axiomatic to me that a resource must be managed by a single scheduler. For performance modeling it would be very odd and not understood, to allocate the processes explicitly to a scheduler. Murray Woodside Distinguished Research Professor Dept of Systems and Computer Engineering, Carleton University, 1125 Colonel By Drive, Ottawa K1S 5B6, Canada. (613)-520-5721.....fax (613)-520-5727....cmw@sce.carleton.ca (http://www.sce.carleton.ca/faculty/woodside.html) ---------- Forwarded message ---------- Date: Thu, 13 Mar 2008 17:38:31 -0000 From: "VanZandt, Lonnie" To: marte-ftf@omg.org Cc: "Edwards, Luke" Subject: MARTE FTF Minutes available at our wiki http://www.omgwiki.org/marte/doku.php?id=marteftfminutes_2008mar13 We spoke much more than I wrote down. Therefore, those of you who attended are welcome to edit the wiki page to add any notes you recorded which I left out. Lonnie VanZandt Field Applications Engineer Denver, CO Artisan Software Tools mobile: 720 201-1349 desk: 303 482-2943 lonnie.vanzandt@artisansw.com