about Glue2.1 draft, CLOUD and GPGPU extension
Dear all, recently the CREAM developers at INFN modified and added some attributes in the GLUE2.1 draft regarding GPUs (now called more generally Accelerators). You can find the new version of the document attached at this email for your convenience and also at the following link: https://cernbox.cern.ch/index.php/s/xyFFk09234oRoSL Please let us know if you have any suggestion. I would like also to schedule a teleconference for presenting and discussing the modifications so we can move on this matter for having the extension approved and then implemented. Since in the next weeks there are being some national holidays in the Netherland, we could schedule this meeting around the middle of May (either 2nd or 3rd week), depending also on your availability: let me know what you think. Best regards, Alessandro -------- Forwarded Message -------- Subject: glue2.1 draft including Accelerator specific attributes Date: Thu, 14 Apr 2016 15:14:40 +0200 From: Marco Verlato <Marco.Verlato@pd.infn.it> <Marco.Verlato@pd.infn.it> To: Alessandro Paolini <alessandro.paolini@egi.eu> <alessandro.paolini@cnaf.infn.it> CC: Virtual Team GPGPU <vt-gpgpu@mailman.egi.eu> <vt-gpgpu@mailman.egi.eu>, EGI-Engage Task WP4.4 "Accelerated Computing" <egi-engage-wp4.4@mailman.egi.eu> <egi-engage-wp4.4@mailman.egi.eu> Dear Alessandro and GPGPU Virtual Team, after the discussion we had in Amsterdam, and after having included the IISAS suggestions, we at INFN have produced a new version for the GLUE2.1 draft that adds some new attributes and replaces the GPU specific attributes with more general Accelerator attributes (to include possible support of MIC and FPGA architectures). Both Computing Service and Cloud Compute Service related classes have been modified to accomodate the new attributes (they are all marked with the comment "added for AC support"). Please have a look at the draft inhttps://cernbox.cern.ch/index.php/s/xyFFk09234oRoSL Comments, suggestions, further addition of possibly forgotten relevant attributes are welcome. Thanks and regards, Marco -- _____________________________________________________________________ Marco Verlato Ph.D Istituto Nazionale di Fisica Nucleare - Sez. di Padova Via Marzolo 8 - 35131 Padova - ITALY Office: Room 119, Via Belzoni 7 Phone:+390499677361, Fax:+390498275952, Email:verlato(at)pd.infn.it _____________________________________________________________________ -- Dr. Alessandro Paolini Operations Officer - EGI.eu Science Park 140 1098 XG Amsterdam The Netherlands skype: alessandro.paolini.egi ********************************* "I believe in the power of laughter and tears" "as an antidote to hatred and terror" "A day without laughter" "is a wasted day" >>> Charlie Chaplin
glue-wg [mailto:glue-wg-bounces@ogf.org] On Behalf Of Alessandro Paolini said:
You can find the new version of the document attached at this email for your convenience and also at the following link: https://cernbox.cern.ch/index.php/s/xyFFk09234oRoSL Please let us know if you have any suggestion.
Just a few quick things ... how much has this been prototyped, i.e. how sure are you that it will work in practice? Bear in mind that the timescale for GLUE updates is very long, so if anything is wrong you'll have to live with it! Your new objects are all called "CloudComputeXxx", it would probably be better to have "CloudComputing" to be consistent with the existing objects. The way you have Benchmark won't work, you have two different Benchmark classes. You need to modify the existing one to relate to the cloud objects. The naming of ToCloudStorageService is maybe not ideal since it suggests that it relates to a CloudStorageService object, which isn't true. You have a new ToCloudComputeService object which relates to StorageService and StorageAccessProtocol, so you need to update the definitions of those classes with the inverse relation. Stephen
Hi Stephen, thank you for the remarks, I'll try to modify the draft accordingly. I may need your help for some of the corrections ;) cheers, Alessandro On 21 April 2016 at 14:22, <stephen.burke@stfc.ac.uk> wrote:
You have a new ToCloudComputeService object which relates to StorageService and StorageAccessProtocol, so you need to update the definitions of those classes with the inverse relation.
And the same for ToCloudStorageService of course.
Stephen
-- Dr. Alessandro Paolini Operations Officer - EGI.eu Science Park 140 1098 XG Amsterdam The Netherlands skype: alessandro.paolini.egi ********************************* "I believe in the power of laughter and tears" "as an antidote to hatred and terror" "A day without laughter" "is a wasted day" >>> Charlie Chaplin
Dear all, please find in the attachement a new version of the draft with the modifications suggested by Stephen: - all the new objects called "CloudComputeXxx" were renamed "CloudComputingXxx" to be consistent with the existing objects - Modified the definition of Benchmark class including the relation to cloud objects; deleted the Benchmark paragraph in the cloud section - updated the StorageAccessProtocol definition adding the relation with ToCloudComputingService - after the recent changes, it needs to produce a new version of some of the UML class diagrams With Stephen, we also noticed that there are still some open questions that need to be discussed in the group: - The naming of ToCloudStorageService is maybe not ideal since it suggests that it relates to a CloudStorageService object, which isn't true. We may want to discuss that more in the group. An option could be *CloudComputingServiceToStorageService*, but that's quite long - for the 2.0 schema we decided to make it shorter, but that creates a small problem now. Or maybe it should have a different naming style, since it's somewhat different to the existing object. Other options: *CloudToStorageService* (to keep the name short), ... - in the StorageService definition is not included the relation to ToComputingService class so do we need to include the association to ToCloudComputingService? Same question for StorageService and ToStorageService / ToCloudStorageService classes Best regards, Alessandro -- Dr. Alessandro Paolini Operations Officer - EGI.eu Science Park 140 1098 XG Amsterdam The Netherlands skype: alessandro.paolini.egi ********************************* "I believe in the power of laughter and tears" "as an antidote to hatred and terror" "A day without laughter" "is a wasted day" >>> Charlie Chaplin
Dear GLUE-WG, Would any of the following dates work for a teleconference to discuss the new Cloud specification? May 3 May 17 May 31 at the usual: 15:30-17:00 Central Europe, 8:30-10:00 AM US Central Please reply with hay/nay by COB this Friday. Regards, JP Navarro
On Apr 26, 2016, at 3:55 AM, Alessandro Paolini <alessandro.paolini@egi.eu> wrote:
Dear all,
please find in the attachement a new version of the draft with the modifications suggested by Stephen:
- all the new objects called "CloudComputeXxx" were renamed "CloudComputingXxx" to be consistent with the existing objects
- Modified the definition of Benchmark class including the relation to cloud objects; deleted the Benchmark paragraph in the cloud section
- updated the StorageAccessProtocol definition adding the relation with ToCloudComputingService
- after the recent changes, it needs to produce a new version of some of the UML class diagrams
With Stephen, we also noticed that there are still some open questions that need to be discussed in the group:
- The naming of ToCloudStorageService is maybe not ideal since it suggests that it relates to a CloudStorageService object, which isn't true. We may want to discuss that more in the group. An option could be CloudComputingServiceToStorageService, but that's quite long - for the 2.0 schema we decided to make it shorter, but that creates a small problem now. Or maybe it should have a different naming style, since it's somewhat different to the existing object. Other options: CloudToStorageService (to keep the name short), ...
- in the StorageService definition is not included the relation to ToComputingService class so do we need to include the association to ToCloudComputingService? Same question for StorageService and ToStorageService / ToCloudStorageService classes
Best regards, Alessandro
-- Dr. Alessandro Paolini Operations Officer - EGI.eu Science Park 140 1098 XG Amsterdam The Netherlands skype: alessandro.paolini.egi ********************************* "I believe in the power of laughter and tears" "as an antidote to hatred and terror" "A day without laughter" "is a wasted day" >>> Charlie Chaplin <glue2.1-draft_v0.2.doc>_______________________________________________ glue-wg mailing list glue-wg@ogf.org https://www.ogf.org/mailman/listinfo/glue-wg
Hi John, For me the preferred day would be May 17th because: - today and tomorrow EGI.eu is closed for national holidays in Netherlands (King's Day), and if I have to prepare some slides I won't have enough time for may 3rd (moreover it clashes a bit with another phone meeting scheduled that afternoon, by chance with US people) - From May 31st to June 2nd I'm being busy in a FitSM training event Cheers, Alessandro Il 26/apr/2016 20:14, "Navarro, JP" <navarro@mcs.anl.gov> ha scritto:
Dear GLUE-WG,
Would any of the following dates work for a teleconference to discuss the new Cloud specification?
May 3 May 17 May 31
at the usual: 15:30-17:00 Central Europe, 8:30-10:00 AM US Central
Please reply with hay/nay by COB this Friday.
Regards,
JP Navarro
On Apr 26, 2016, at 3:55 AM, Alessandro Paolini <alessandro.paolini@egi.eu> wrote:
Dear all,
please find in the attachement a new version of the draft with the modifications suggested by Stephen:
- all the new objects called "CloudComputeXxx" were renamed "CloudComputingXxx" to be consistent with the existing objects
- Modified the definition of Benchmark class including the relation to cloud objects; deleted the Benchmark paragraph in the cloud section
- updated the StorageAccessProtocol definition adding the relation with ToCloudComputingService
- after the recent changes, it needs to produce a new version of some of the UML class diagrams
With Stephen, we also noticed that there are still some open questions that need to be discussed in the group:
- The naming of ToCloudStorageService is maybe not ideal since it suggests that it relates to a CloudStorageService object, which isn't true. We may want to discuss that more in the group. An option could be *CloudComputingServiceToStorageService*, but that's quite long - for the 2.0 schema we decided to make it shorter, but that creates a small problem now. Or maybe it should have a different naming style, since it's somewhat different to the existing object. Other options: *CloudToStorageService* (to keep the name short), ...
- in the StorageService definition is not included the relation to ToComputingService class so do we need to include the association to ToCloudComputingService? Same question for StorageService and ToStorageService / ToCloudStorageService classes
Best regards, Alessandro
-- Dr. Alessandro Paolini Operations Officer - EGI.eu <http://egi.eu> Science Park 140 1098 XG Amsterdam The Netherlands skype: alessandro.paolini.egi ********************************* "I believe in the power of laughter and tears" "as an antidote to hatred and terror" "A day without laughter" "is a wasted day" >>> Charlie Chaplin <glue2.1-draft_v0.2.doc>_______________________________________________ glue-wg mailing list glue-wg@ogf.org https://www.ogf.org/mailman/listinfo/glue-wg
participants (3)
-
Alessandro Paolini -
Navarro, JP -
stephen.burke＠stfc.ac.uk