[Pgi-wg] [EGI Technical Forum 2010] Abstract 74: Standardization of Distributed Data Processing - Useful Standards and Interfaces
Etienne URBAH
urbah at lal.in2p3.fr
Thu Aug 5 13:02:26 CDT 2010
Steven,
Concerning Abstract 74 for my poster proposed at the EGI Technical Forum
2010 :
After logging in INDICO, I can read the content of my abstract, but the
'modify' button in NOT active (both on IE and Firefox).
Therefore, I am providing the abstract in literary English below.
This poster presents two UML diagrams concerning standards which may
be useful for grid computing middleware, and interfaces which should be
taken into account for proper design of grid middleware permitting
operation in production. Selection and then widespread adoption of
adequate standards would permit grid interoperability.
The web page http://forge.gridforum.org/sf/go/doc15990?nav=1
contains a draft version of a first UML diagram presenting official and
de facto standards which may be useful for the design of middleware
permitting remote data processing (cloud computing) and distributed data
processing (grid computing). The main point is that the GFD.147
recommendation of OGF, named GLUE Specification v. 2.0, permits to
describe grid and cloud entities. This allows different GLUE compliant
grids to discover and understand each other, so that this GLUE
specification should be used as foundation for all other grid and cloud
standards and recommendations for interoperability.
The web page http://forge.gridforum.org/sf/go/doc15981?nav=1
contains a draft version of a second UML diagram presenting interfaces
between Clients, Functionalities and Resources existing in remote data
processing infrastructures (clouds) and in distributed data processing
infrastructures (grids).
This second diagram uses color coding, with grid and cloud
functionalities and resources in green, interfaces permitting to access
them in blue. Besides, official and de facto standards are inside
square brackets.
Final Users are scientists, which are interested only in 'payload'
functionalities, like data, instruments, activities, and jobs, but grids
and clouds require also 'support' functionalities, in particular
information publication and discovery, security management, monitoring,
logging and accounting. Therefore, interoperability requires to
standardize all 'client interfaces' shown in blue, at the left.
Grid and cloud functionalities are not independent of each other,
but most functionalities require to access resources managed by other
functionalities, in particular information items, security descriptors,
logging records and accounting records.
So, if we do not standardize 'backend' interfaces shown in blue, at
the right, then different instances implementing the same functionality
will not be able to correctly access required resources, and
interoperability will not be achieved. Therefore, we have to
standardize also these 'backend' interfaces. This task is huge, so we
have assess, for each 'backend' interface, if standardization is really
needed, and we have to prioritize, with following suggestions :
Firstly, cloud and grid computing require information publication
and discovery. For interoperability, we already have the GLUE
specification as foundation, and we urgently need GLUE renderings, in
particular for LDAP, XML, and SQL. Clients performing queries do not
need that the information records are really kept in a relational
database, but that queries may be expressed in a (perhaps limited) SQL
syntax.
Secondly, clouds and grids are valuable targets for criminal
organizations, so that clouds and grids absolutely require high-grade
security management, and interoperability between grids (an later
clouds) require a common set of standards. We should confirm IGTF as
Trust Anchor and SPG (formerly JSPG) for general security policies. For
authentication, we should phase out GSI proxies in favor of
RFC3820-compliant X509 proxies as soon as possible. For authorization,
VOMS extensions of X509 proxies work correctly, but SAML assertions
issued for example by VOMS servers also permit restrictions of
authorization, and perhaps easier delegation.
Thirdly, security audit absolutely requires log records for jobs and
other activities which clients can submit. Interoperability between
grids require that the format of the log records is standardized. We
have no published standard yet, and we urgently need one. The JSDL
working group of OGF is working on such a standard named 'Activity
Instance Document Schema'. We should quickly assess if the draft
available at http://forge.gridforum.org/sf/go/doc15395?nav=1 is
appropriate. If yes, we should help the JSDL working group finalize it
as an OGF recommendation, and implement it.
Fourthly, accounting will become more and more important, as grid
computing evolves from a research topic to a valuable commodity like
cloud computing. The GFD.98 recommendation of OGF, named 'Usage Record
- Format Recommendation', is already published. We should assess if
this recommendation is appropriate, and if necessary, improve it, then
implement it.
By contrast, job management is used by very few other
functionalities. As long as the four functionalities mentioned above
are correctly implemented according to common standards, interoperation
of job management between different clouds and grids could then be
quickly achieved using bridges implementing an adequate adapter for each
job management idiom, though perhaps with reduced scalability compared
to full interoperability using the GDF.108 recommendation named 'OGSA®
Basic Execution Service Version 1.0'.
Now that I have finished, I thank Steven very much : This makes a good
start for an academic paper.
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 at lal.in2p3.fr
-----------------------------------------------------
On Thu, 05/08/2010 15:42, Steven Newhouse wrote:
> Dear Etienne,
>
> We would like to accept your poster but feel the abstract needs to be
> revised to better reflect the proposed content. Can you please submit a
> better abstract - one where the text is easier to read rather than a
> series of bullet points - immediately.
>
> Thanks,
>
> Steven
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 5073 bytes
Desc: S/MIME Cryptographic Signature
Url : http://www.ogf.org/pipermail/pgi-wg/attachments/20100805/ca90f557/attachment-0001.bin
More information about the Pgi-wg
mailing list