New Draft available and next phone meeting
Dear all, We updated the UR 2.0 draft and we uploaded all the files to redmine. The pdf files is available at the usual location. The direct link to latest pdf is the following: https://redmine.ogf.org/dmsf_files/12979?download= If you have some time during the holidays please have a look at it and send your comment. We think that, except for some examples missing we are at a good point with all the blocks. The only one that might need more discussion is the CloudUsageBlock. We are still interacting, especially with Alison, on its definition but also in this case we think we are quite close to the final definition. We could think a bit more if we should include or not a block for the network usage. We can discuss this right after the holidays and take a final decision on next phone meeting which is fixed for Tuesday the 8th of January 2013 at 15:00 CET. Merry Christmas an Happy new Year, Andrea and Jon -- Andrea Cristofori INFN-CNAF Viale Berti Pichat 6/2 40127 Bologna Italy Tel. : +39-051-6092920 Skype: andrea-cnaf
Hi there, I started to read carefully through our document. Gone through the Intro, Compute, Job, Memory and References. I will go on tomorrow. My first comments/correction are below. I also attached a pdf with my changes, as I have not yet checked in the changes into git at redmine. Regards, Ralph General: For every property we sould use the same beginning: Currently we have both: "This property describes the class of memory used." or "The class of memory used." The latter is shorter and doesn't use the boring repetition. I changed it everywhere. Changed everywhere to the common notation: Host MUST be a string. instead of "The Host field type MUST be a string." Both were used in the document. Did some fixes in spelling and grammatical changes. Introduction: What is 1.2 intended for? ComputeUsageBlock: Processors: We don't mention cores. For example a node with two processors and 4 cores each. 8 serial jobs can run there. I added a sentence: "In addition, a processor can consist of several cores which may be used independently. In the example above the use of dual-core processors would thus increase the number to 64." Deleted the second charge block in compute usage and moved the description of the the attributes from the bullets to the text. Charge is now used to describe a currency. But IIRC our discussion it was not meant this way but be more a weight or allocation unit. It is now in the compute usage block, but wasn't it meant to be a job charge? See http://redmine.ogf.org/projects/ur-wg/wiki/ConsideredURFieldsJobs JobUsageBlock: Is not only intended for batch systems, but Grid, Cloud, a local computer or whatever. Therefore we should not only talk about batch systems e.g. in LocalJobId. It can be any local ID, not necessarily the LRMS ID. Otherwise we restrict this blocks usage. The definition of "MachineName" is too restrictive (only batch). I used the one from our Wiki. For "SubmitHost" it is stated: In a grid environment \emph{SubmitHost} MUST be present. But I doubt that a site can always find out from where the job was submitted. I would not see "Middleware" as a MUST. ServiceLevel reports now a normalization factor. I don't understand what the service level has to do with the normalization. In the Wiki we had it defined as "This property identifies the quality of service associated with the resource consumption. For example, service level may represent a priority associated with the usage." The "Host or DestinationResource" is missing, which is the node/host at which the job ran. I would pledge for the name "ExecutionNode" MemoryUsageBlock: The *CapacitiyUsed should be positive integers. References: In the version I checked out from redmine git the reference cristofori2011grid is missing in the bibtex file. Andrea Cristofori wrote on 21.12.2012 15:13:
Dear all,
We updated the UR 2.0 draft and we uploaded all the files to redmine. The pdf files is available at the usual location. The direct link to latest pdf is the following:
https://redmine.ogf.org/dmsf_files/12979?download=
If you have some time during the holidays please have a look at it and send your comment. We think that, except for some examples missing we are at a good point with all the blocks. The only one that might need more discussion is the CloudUsageBlock. We are still interacting, especially with Alison, on its definition but also in this case we think we are quite close to the final definition. We could think a bit more if we should include or not a block for the network usage. We can discuss this right after the holidays and take a final decision on next phone meeting which is fixed for Tuesday the 8th of January 2013 at 15:00 CET.
Merry Christmas an Happy new Year, Andrea and Jon
Hi there, this time some comments to RecordIdentityBlock, StorageUsageBlock, CloudUsageBlock and Abstract. I also added a pdf which colors all changes made. Regards, Ralph General: added references to chapters when an element is the same as in another block (e.g. "(see chapter 5.6") ) RecordIdentityBlock: RecordId - changed the last part of of the description ("of the grid implementation") to a more general "of the accounting context." Infrastructure is set to be a MUST. Why? What about a local record? I would vote for a SHOULD. added "Infrastructure MUST be a string." GlobalGroupAttribute Changed to phrasings used in the rest of the document (e.g. "GlobalGroupAttribute MAY be present multiple times.") StorageUsageBlock: StorageClass - we give examples ("replicated" ...). But do we want to define values? (E.g. as for MemoryClass) StorageResourceCapacityUsed: Comment to the implementation note: we say one should use 64-bit integers as it will overflow at 8 Exabytes. 8 Exabyte is not that long the way (in some years, currently we are talking about several PB) - Would it not be better to propose 128-bit integers? CloudUsageBlock LocalVirtualMachineId - why MUST it be the ID as assigned by the Cloud system? I would better see here the same phrasing as for the LocalJobId in the JobUsageBlock. SuspendDuration: Changed the description to "The amount of time in which the Virtual Machine status was ``suspended''." moved Status before SuspendDuration, as the latter depends on it ImageId: Added type and example Abstract completed accounting resources (added computing and memory) completed author list Acknowledgements added some sentences other projects to thank? commented out glossary as we have no entries (yet?) Ralph Mueller-Pfefferkorn wrote on 03.01.2013 16:59:
Hi there,
I started to read carefully through our document. Gone through the Intro, Compute, Job, Memory and References. I will go on tomorrow.
My first comments/correction are below. I also attached a pdf with my changes, as I have not yet checked in the changes into git at redmine.
Regards, Ralph
General: For every property we sould use the same beginning: Currently we have both: "This property describes the class of memory used." or "The class of memory used." The latter is shorter and doesn't use the boring repetition. I changed it everywhere.
Changed everywhere to the common notation: Host MUST be a string. instead of "The Host field type MUST be a string." Both were used in the document.
Did some fixes in spelling and grammatical changes.
Introduction: What is 1.2 intended for?
ComputeUsageBlock:
Processors: We don't mention cores. For example a node with two processors and 4 cores each. 8 serial jobs can run there. I added a sentence: "In addition, a processor can consist of several cores which may be used independently. In the example above the use of dual-core processors would thus increase the number to 64."
Deleted the second charge block in compute usage and moved the description of the the attributes from the bullets to the text.
Charge is now used to describe a currency. But IIRC our discussion it was not meant this way but be more a weight or allocation unit. It is now in the compute usage block, but wasn't it meant to be a job charge? See http://redmine.ogf.org/projects/ur-wg/wiki/ConsideredURFieldsJobs
JobUsageBlock:
Is not only intended for batch systems, but Grid, Cloud, a local computer or whatever. Therefore we should not only talk about batch systems e.g. in LocalJobId. It can be any local ID, not necessarily the LRMS ID. Otherwise we restrict this blocks usage.
The definition of "MachineName" is too restrictive (only batch). I used the one from our Wiki.
For "SubmitHost" it is stated: In a grid environment \emph{SubmitHost} MUST be present. But I doubt that a site can always find out from where the job was submitted.
I would not see "Middleware" as a MUST.
ServiceLevel reports now a normalization factor. I don't understand what the service level has to do with the normalization. In the Wiki we had it defined as "This property identifies the quality of service associated with the resource consumption. For example, service level may represent a priority associated with the usage."
The "Host or DestinationResource" is missing, which is the node/host at which the job ran. I would pledge for the name "ExecutionNode"
MemoryUsageBlock: The *CapacitiyUsed should be positive integers.
References: In the version I checked out from redmine git the reference cristofori2011grid is missing in the bibtex file.
Andrea Cristofori wrote on 21.12.2012 15:13:
Dear all,
We updated the UR 2.0 draft and we uploaded all the files to redmine. The pdf files is available at the usual location. The direct link to latest pdf is the following:
https://redmine.ogf.org/dmsf_files/12979?download=
If you have some time during the holidays please have a look at it and send your comment. We think that, except for some examples missing we are at a good point with all the blocks. The only one that might need more discussion is the CloudUsageBlock. We are still interacting, especially with Alison, on its definition but also in this case we think we are quite close to the final definition. We could think a bit more if we should include or not a block for the network usage. We can discuss this right after the holidays and take a final decision on next phone meeting which is fixed for Tuesday the 8th of January 2013 at 15:00 CET.
Merry Christmas an Happy new Year, Andrea and Jon
-- ur-wg mailing list ur-wg@ogf.org https://www.ogf.org/mailman/listinfo/ur-wg
participants (2)
-
Andrea Cristofori
-
Ralph Mueller-Pfefferkorn