Hi Paul,
thanks for pointing to the EGA reference model. I have briefly browsed through the papers.

Here are some initial thoughts from my side.

Grid components:
I think that virtualization does not add new components. Rather it adds a virtualized flavor in addition to the physical one ... as already pointed out in the EGA reference model.
Accordingly, the DMTF SVPC working group has extended CIM to show virtual and physical systems/resources side by side with associations between both side.

Service Level Management:
System Virtualization definitely has an impact here. Monitoring, metering, workload management needs to be aware of virtualization.
So far applications play a passive role w.r.t. virtualization. (This is surely good to ensure compatibility.) Service level management is usually done from outside. An inside-out view could provide new options, e.g. a job asking for more CPU capacity.

Use cases
Indeed here I see additional use cases. The paper (2nd part) already lists some "Future Use Cases".
By OGF20 we want to describe use cases which are specific to system virtualization.

Wolfgang Reichert
Senior Technical Staff Member
••• SAP on IBM Systems •••
+49-7031-162185 Fax: -163545 (IBM)
+49-6227-764084 Fax: -774084 (SAP) 		IBM Deutschland Entwicklung GmbH
Vorsitzender des Aufsichtsrats: Johann Weihen
Geschäftsführung: Herbert Kircher
Sitz der Gesellschaft: Böblingen
Registergericht: Amtsgericht Stuttgart, HRB 243294




"Travostino, Franco" <ftravostino@ebay.com>

05.04.2007 01:23

To
"Strong, Paul" <pstrong@ebay.com>, <erol.bozak@sap.com>, Wolfgang Reichert/Germany/IBM@IBMDE
cc
<travos@ieee.org>, <delaat@science.uva.nl>, <gridvirt-wg@ogf.org>
Subject
RE: Overlap?





Paul,
thank you for reaching out to our new GridVirt group. I characterize the scope of this WG as a vertical thin slice of the virtualization space. Simply stated, this thin slice has to do with grid infrastructures driving virtual machines resources (whether it's memory quotas or "computons" or infiniband queue pairs). Other worthy aspects of virtualization (e.g., network virtualization) are out of this group's scope (some other OGF groups, like GHPN-RG and NM-WG, have gone or are going down this path).
 
So said, the EGA Ref Model can be quite useful to frame the lifecycle of a VM. I very much like its being crisp and intuitive. Thus far, we have had some loose interaction with the DMTF SVPC group around their modeling work (see their presentation at our DC meeting). I expect that the GridVirt and its nascent constituency will start off by building an appreciation of which model can best serve the GridVirt WG.
 
What we do know today is that the GridVirt will abstain from rolling out yet a new model.
 
-franco
 
---
Franco Travostino
OGF Area Director, Infrastructure
 
 
 
 
 

From: Strong, Paul
Sent: Monday, April 02, 2007 3:22 PM
To: erol.bozak@sap.com; w_reichert@de.ibm.com
Cc: Travostino, Franco
Subject: Overlap?
Hi,
 
I just wanted to touch base with you guys re your scope and the potential to perhaps leverage the work the Reference Model group is doing (see https://forge.gridforum.org/sf/projects/rm-wg).  The EGA Ref Model (which we are in the process of refining in UML) describes virtualization quite a bit, relates it to Grid (see https://forge.gridforum.org/sf/docman/do/listDocuments/projects.rm-wg/docman.root.ega_reference_model_v1_5) and describes the life cycles of all components (virtual and physical) and how they relate to each other.  It would seem that the model may provide you with a useful basis for moving forward.  I think it would be good to discuss via email and perhaps to meet at OGF20 to discuss.  I have cc’ed Franco, now working here at eBay.
 
Regards
Paul
 
Paul Strong
Distinguished Engineer, eBay Research Labs
Tel: +1 408 967 8450
Mob: +1 408 981 8948
Skype/AIM: FractalPaul