GridNet2 Report 

advertisement
GridNet2 Report OGF 22, Cambridge, USA, February 25-28, 2008
[GridNet ID: 128]
Xiaoyu Chen
School of Engineering and Design
Brunel University
[email protected]
This report provides information on my involvement in sessions during OGF
22 held in February 25th – 28th, 2008. Although there are many sessions (at
least two session) i attended daily, the content of this report concentrates
on my own sessions (OGF-RUS and OGF-UR work group) and interesting
sessions (mainly on financial services and data management) from my
personal point of view.
Date:
Session:
Role:
Feb-25, 2008
OGF-Europe
Participant
In this session, David Wallom announced the formation of OGF-Europe
that aims at mainly providing supports for standards adoption in both
academic and commercial areas. Standards defined within OGF are
recognised to be importance on interoperability. Supports however are
required to given in order to minimise complex issues during
implementation development of OGF standards. The OGF-Europe plans to
start from requirement collection across key stakeholders and end users
from Europe. Another important objective of OGF-Europe is to bridge the
gap between eScience and enterprise communities and “mainstream”
grid adoptions. During the next 24 months, a bundle of outputs are
expected including internal events and workshops, reports on OGF
technical & strategic roadmap and best practices, and concluded grid
adoption challenges. The OGF-Europe also brings potential studentship
and project collaboration in Europe. I personally think the formation of
OGF-Europe is a good idea in terms of steering adoption of Grid
technology and development of spec. Implementation “not only for
eScience but for enterprise”. However, the success of OGF-Europe is still a
question depending on the willingness of involvement from enterprise
communities.
1|Page
Date:
Session:
Role:
Feb-25, 2008
OGF-RUS
Member/spec. Contributor
During this session, Morris give a brief background information on Resource
Usage Service (RUS) and focus on UNICORE implementation as an
adopter of RUS specification and future plans. UNICORE has RUS service
implemented mainly for runtime resource usage monitoring as a particular
site. The status of RUS is not as expected in the case of many draft
documents without single one finalised. Besides, the recently
specifications, OGSA-RUS IDL 2.0 and RUS WS-I rendering, recommends
implementations to use WS-Enumeration for query large usage data. This is
unacceptable because the WS-Enumeration is a group submission which
is not a finalised specification. Also the roadmap of RUS is expected to be
convergent to WS-Management, which is therefore recommended to be
removed as well.
Date:
Session:
Role:
Feb-26, 2008
What OGF can Do for
Enterprise
Participant
The Grid, as a new distributed computing technology, has been
accepted as next-generation IT infrastructure for both eScience and
Commercial Usage. The session focuses on the debate about how OGF
contributes to enterprise. As a CIO, the presentator recognised the near
future tasks in University of Chicago still mainly aim at science and
engineering infrastructure, data sharing, high-volume data movement,
secure access, and etc, which means eScience-oriented. In order to bring
enterprise in scope, I think a reasonable strategy would be from
implementation platforms that are mostly attractive from the perspective
of enterprises. Such technologies includes Microsoft .NET for Grid service
implementation, and unbeatable Web 2.0 techniques, such as Adobe AIR
and AJAX. Alternatively, I strongly encourage developers to consider
using or integrate available commercial Web services (e.g. Amazon Web
services). The presentator emphasises the uniquely value about OGF:
“product-neutral experience and perspective” and prospect potential
values that OGF could provide:
• Definition of specifications for legacy system
• Evaluation of various standards and service implementations
• Examples of successful integration of standards and services
2|Page
Date:
Session:
Role:
Feb-27, 2008
UR 2.0: Survey and
Modelling
Co-chair
The Usage Record session is divided into two parts:
Part 1: UR v1 Survey
Usage Record specification (version 1.0) has been published and used in
many product Grid projects (e.g UNICORE and EGEE). After initial
collections of comments on UR 1.0, it is time to collect comments and edit
an experience document in order to advance the standard into next
version. As recongised, the UR 1.0 concentrates on the comptatutional
usage metric representation without concerning storage and network
usage
recording.
The
Survey
form
(available
at
http://forge.ogf.org/short/ur-wg/survey)
is
therefore
interested
in
feedbacks of UR 1.0 metrics only, while advanced feedbacks are also
welcomed as an attachement to be uploaded into response forum
(http://forge.ogf.org/short/ur-wg/experiences)
Part 2: Proposed UR v2 modelling
As proposed in OGF 21, the UR version 2.0 is composed of six record
models extended from an abstract UR core model. Before coming in the
details of each model, there are two main concepts required to be
corrected in UR 2.0: “summarisation” and “aggregation”. The
“aggregation” targets at a single transaction (a non-trivial job), the usage
information of which is modelled as a composite usage record. A
summary usage records presents usage information of multiple
transactions in summaried format, therefore requires extra metric
definition (such as summary algorithm and summarisation timestamp).
The talk also gives an proposed information model of UR 2.0 extended
from CIM metric model (version 2.7). The UR core model is directly
extended from the Unit of Work (UoW) defintion of CIM metric model. The
definition of is generic enough to represent various transactions (batch
jobs, user-initiated interactive operations, and service transactions). The
UoW also contributes to the correlation of multiple usage records. A usage
record is able to represents multiple fine-grained UoWs to be executed in
serial of in parallel. A usage record is also able to represent UoWs refined
into subunits by functionalities (computation and storage access). A
compute usage record extends from the core model with additional
compute usage metrics (mostly from UR 1.0). The metrics of storage and
network usage record requires further comments and investigations. A
composite usage record represents a single UoW with an aggregation of
3|Page
different types of usage information of the UoW. A summary usage record
is then summarise usage information of multiple UoWs and therefore
composite in nature. Considering its complexity, it seems not an option of
UR 2.0 to be recasted from UR 1.0 but to be backforward compatible only.
4|Page
Download