Re: [Pgi-wg] SIENA and OGF PGI : Useful de facto and official standards for grids and clouds
Martin, Morris, Johannes and all, Concerning standards for grids and clouds : I have updated the diagram presenting useful de facto and official standards for grids and clouds. In particular, I have added SNIA CDMI, DMTF OVF, and some OGF HPC profiles. This diagram is open source, and is available at : - http://forge.gridforum.org/sf/go/doc15990 as PNG file, - http://forge.gridforum.org/sf/go/doc15977 as ZARGO file editable with the ArgoUML tool which anybody can download from http://argouml.tigris.org/ I will fix the comment about JSPG as soon as ArgoUML permits it. Then, I will update my slides accordingly. Best regards. ----------------------------------------------------- Etienne URBAH LAL, Univ Paris-Sud, IN2P3/CNRS Bat 200 91898 ORSAY France Tel: +33 1 64 46 84 87 Skype: etienne.urbah Mob: +33 6 22 30 53 27 mailto:urbah@lal.in2p3.fr ----------------------------------------------------- On Mon, 07/03/2011 17:18, Martin Antony Walker wrote:
Thank you, Etienne. I find this material very helpful in arriving at a clear picture. Best regards, - Martin
-----Original Message----- From: Etienne URBAH [mailto:urbah@lal.in2p3.fr] Sent: Monday, March 07, 2011 5:09 PM To: martin.antony.walker@googlemail.com Cc: lodygens@lal.in2p3.fr; edgi-na2@mail.edgi-project.eu Subject: Re: [REB-SIENA] Cloud Federation raised by Reuven Cohen - Standards
Martin,
Concerning Cloud Federation and standards :
DMTF CIM (Common Information Model) ------------------------------------ For information standardization, DMTF CIM may be considered. In particular, it is used by the Japanese NAREGI middleware, as described in chapter 2.6.1 of http://middleware.naregi.org/Download-v1.1.5/Docs/GD-Overview-e.pdf
But most European Grid stakeholders have agreed on OGF GLUE 2.0 instead.
Standards missing from my slides -------------------------------- Yes, my slides at http://www.xtremweb-hep.org/IMG/pdf/Standards-For-Distributed-Data-Processin g-Infrastructures-DCIs.pdf have to be updated.
I know that I have to add references to at least following standards :
- SNIA CDMI : Cloud Data Management Interface
http://www.snia.org/tech_activities/standards/curr_standards/cdmi/CDMI_SNIA_ Architecture_v1.0.pdf
- DMTF OVF : Open Virtualization Format
http://www.dmtf.org/standards/published_documents/DSP0243_1.1.0.pdf
- OGF http://www.ogf.org/documents/ GFD.111 : JSDL HPC Profile Application Extension GFD.115 : JSDL SPMD Application Extension GFD.151 : HPC-BP Advanced Filter Extension GFD.174 : Data Format Description Language (DFDL)
I hope to update the source diagram and the slides very quickly.
Best regards.
----------------------------------------------------- Etienne URBAH LAL, Univ Paris-Sud, IN2P3/CNRS Bat 200 91898 ORSAY France Tel: +33 1 64 46 84 87 Skype: etienne.urbah Mob: +33 6 22 30 53 27 mailto:urbah@lal.in2p3.fr -----------------------------------------------------
On Mon, 07/03/2011 10:21, Martin Antony Walker wrote:
Etienne, thanks again for the slides. They are very clear. I have boiled down the content to one and a half pages for myself. I have a question and a remark. Why is the DMTF CIM schema not considered for information standardization? The remark is that I miss CDMI and OVF among the available standards - and I guess there are others. Best regards, - Martin
-----Original Message----- From: Etienne URBAH [mailto:urbah@lal.in2p3.fr] Sent: Friday, March 04, 2011 4:50 PM To: martin.antony.walker@googlemail.com Cc: Morris RIEDEL; david.wallom@oerc.ox.ac.uk; REB@sienainitiative.eu; lodygens@lal.in2p3.fr; edgi-na2@mail.edgi-project.eu Subject: Re: [REB-SIENA] Cloud Federation raised by Reuven Cohen
Martin,
Concerning Cloud Federation, I fully agree with Reuven Cohen.
The good news is that federations of data processing resources already exist, and run in full production : they are called Computing Grids.
For a presentation about the issues and solutions for a Cloud Federation, you can take a look at my 'Standards for Distributed Data Processing Infrastructures (DCIs)' presentation at
http://www.xtremweb-hep.org/IMG/pdf/Standards-For-Distributed-Data-Processin
g-Infrastructures-DCIs.pdf
Acknowledgments to : - Morris Riedel for the diagram in my slide number 27. - David Wallom for the diagrams showing dependencies between OGF recommendations which he presented at the 6th E-infrastructure Concertation Meeting in Lyon on 24 November 2008.
Professionalism and Engineering pay in the long term.
Best regards.
----------------------------------------------------- Etienne URBAH LAL, Univ Paris-Sud, IN2P3/CNRS Bat 200 91898 ORSAY France Tel: +33 1 64 46 84 87 Skype: etienne.urbah Mob: +33 6 22 30 53 27 mailto:urbah@lal.in2p3.fr -----------------------------------------------------
On Fri, 04/03/2011 15:55, Martin Antony Walker wrote:
FYI – you might have seen SpotCloud getting positive mention in a recent issue of The Economist. Best regards, - Martin
*From:*noreply+feedproxy@google.com [mailto:noreply+feedproxy@google.com] *On Behalf Of *ElasticVapor - Life in the cloud *Sent:* Friday, March 04, 2011 3:08 PM *To:* martin.antony.walker@gmail.com *Subject:* ElasticVapor : Living on the Edge
ElasticVapor : Living on the Edge<http://www.elasticvapor.com/> Image removed by sender.
<http://fusion.google.com/add?source=atgs&feedurl=http://feeds.feedburner.co
m/Elasticvapor>
------------------------------------------------------------------------
*Living on the Edge*
<http://feedproxy.google.com/~r/Elasticvapor/~3/cNO_4CUnSnA/living-on-edge.h
tml?utm_source=feedburner&utm_medium=email>
Posted: 03 Mar 2011 01:04 PM PST
There's somethin' wrong with the cloud today, I think I know what it is. We're seeing things in a different way, but should we be judging a cloud provider by the color of their -- logo?
It's seems that for many, the only basis of comparing cloud providers is based upon superficial aspects. I know of the company, recognize the logo, or read a random review. But the reality is we're moving away from the traditional vendor driven marketing fluff of a single provider world to a multi-cloud, federated ecosystem of capacity providers, where brand recognition is less important than performance and price. I'm talking about living on the edge, the edge of the network.
One of the more interesting recent announcements was *Amazon's Japan availability zone* <http://aws.typepad.com/aws/2011/03/now-open-aws-region-in-tokyo.html>, in describing their launch AWS spoke of latency for users within Tokyo being less than 10ms. Hitting directly at the heart of the opportunity. Yet on the flip side, they also mentioned that the Japanese zone was ideal for other nearby geographies, to which I say they're missing the point. Using Japanese resources in South Korea makes little sense given the rapid advancement and availability of cloud capacity in South Korea. The opportunity going forward isn't to address generalized areas of the world, but to address the specifics, not just on a country basis but on a city or even a neighbourhood basis. A single provider will never be able to get this level of granularity regardless of how much money they have. Economies of scale will always be limited by total market size, the more granular the market the less economies of scale work in favor the large provider. The only way to address this growing movement toward edge based, latency dependent application deployment is to federate many providers with many customers across many diverse geographies.
I admit that this is easier said than done, ask anyone who's attempted to use more than one provider in a federated, intercloud connection global cloud of clouds type of deployment and tell me what you think about your sit-u-a-tion. They'll tell you it's complication, and aggravation.
I know a few of you will point back to the inevitable economies of scale that an AWS or Google bring forth. Yes, they probably spent many millions building their Japanese cloud infrastructure. But did they have to? Economies of scale are important factors mostly for true for commoditized IT aspects such as bandwidth. But for areas such as ultra-localization computing, this is not practical for even the largest web companies. Sure there are many factors that cause a cloud providers average cost per compute unit to fall as the scale of output is increased, purchasing power is probably the most relevant. Essentially the biggest Internet companies can buy the most servers at a massive volume thus getting the lowest cost per unit of compute time and therefore achieving the best margins at the lowest cost. But even this equation has a practical limit when it comes to geography.
When it comes to ultra-localization the boundaries of the so-called provider economies of scale, commoditiziation and volume quickly breakdown. Now it becomes a question of federation and aggregation. Many providers connected through a normalized or structured market interface rather then one provider attempting to address all markets. I'm not just talking about what we're doing with *SpotCloud* <http://www.spotcloud.com/>, but what I believe to be the move toward a market centric economy of federated cloud ecosystems, a move I think is inevitable. Ultra-local capacity or edge based computing, or whatever you chose to call it in a nutshell is the opportunity moving forward.
The choice will quickly become one of choosing a single provider (one to many) or an aggregator (many to many). I believe the choice will quickly become obvious to anyone who has a geographic component to their applications. The opportunity is living on the edge.
------------------------------------------------------------------------
*Announcing The Enomaly Cloud Service Provider Edition* <http://www.enomaly.com> | *Twitter Me*<http://www.twitter.com/ruv> | *Get Linkedin*<http://www.linkedin.com/in/reuvencohen> | *Contact Reuven*<https://cloudcomputing.wufoo.com/forms/contact-reuven/> | *Disclosure Policy* <http://www.elasticvapor.com/2009/05/elasticvapor-disclosure-policy.html>
------------------------------------------------------------------------
Image removed by sender.
*Image removed by sender.*
<http://feeds.feedburner.com/~ff/Elasticvapor?a=cNO_4CUnSnA:_YDXnI4P0Ck:4cEx
4HpKnUU>*Image
removed by sender.*
<http://feeds.feedburner.com/~ff/Elasticvapor?a=cNO_4CUnSnA:_YDXnI4P0Ck:yIl2
AUoC8zA>*Image
removed by sender.*
<http://feeds.feedburner.com/~ff/Elasticvapor?a=cNO_4CUnSnA:_YDXnI4P0Ck:qj6I
DK7rITs>*Image
removed by sender.*
<http://feeds.feedburner.com/~ff/Elasticvapor?a=cNO_4CUnSnA:_YDXnI4P0Ck:V_sG
LiPBpWU>*Image
removed by sender.*
<http://feeds.feedburner.com/~ff/Elasticvapor?a=cNO_4CUnSnA:_YDXnI4P0Ck:F7zB
nMyn0Lo>*Image
removed by sender.*
<http://feeds.feedburner.com/~ff/Elasticvapor?a=cNO_4CUnSnA:_YDXnI4P0Ck:-BTj
WOF_DHI>
Image removed by sender.
You are subscribed to email updates from ElasticVapor - Life in the cloud<http://www.elasticvapor.com/> To stop receiving these emails, you may unsubscribe now
<http://feedburner.google.com/fb/a/mailunsubscribe?k=petIEsZUdsZ3VYU95tUROl7
iW6A>.
Hi Etienne,
I will fix the comment about JSPG as soon as ArgoUML permits it.
I'm not sure what do you mean here, but SPG (formerly known as JSPG) has no direct (not even indirect) influence on standards: it defines operational policies. It does make use of standards, though, in the sense of enforcing them in deployment environments when appropriate, but it is *extremely* careful of not suggesting any specific implementation. Cheers, Oxana, - wearing SPG hat
participants (2)
-
Etienne URBAH -
Oxana Smirnova