***GGF10
by John Latta
Berlin, Germany
March 10 – 12
Grid as a Fundamental Shift in Computing and Digital
Infrastructure
Grid is a computer and network playground for academics? Wrong.
Here at Global Grid Forum, which was only started in 1999, it is
evident that a fundamental shift in computing is taking place. It
will take years to unfold and only now are its impacts beginning
to reach into the enterprise. Certainly IBM is a leader but Sun,
HP, Oracle and others are in rapid chase. The potential to go
into all facets of computing is obvious, even including the home.
This GGF10 is devoted largely to the standards process but there
are many insights and contacts. There are numerous mind shifts
forced by the impact of grid and here is a sample:
A Grid infrastructure looks like an OS,
Software is infrastructure
Grid makes virtualization possible,
Where an application runs, data is kept or a service
originates is transparent
Virtualization enables abstraction that distances physical
entities from the user and the association with the
computer or OS upon which the application runs is much
less important,
An effective Grid infrastructure is one where the
application does not know the Grid is present, and
The notion of a dumb network and smart clients is turned
upside down with Grid, as the network and computation are
inseparable.
The best way to think of GGF is IETF. GGF is largely a standards
body and many of the policies and structures are patterned after
the IETF. This is an open process. It is made up of working
groups where the standards get drafted. In addition, other
documents are published which are informational. The process of
forming a working group is bubble up. It can begin with an ad-hoc
meeting of interested parties then move to Birds of a Feather
(BOF), and when a charter has been approved, it becomes a working
group to draft a standard. Note that there are also less formal
efforts under Research Groups. The IP policy follows that of the
IETF and is based on RAND – reasonable and non-discriminatory
with a strong emphasis on free IP licensing. Once a standard has
been approved there is a period of time that runs from 6 – 24
months for independent implementations to be tested. Here at
GGF10 there is little of the conference activities such as a
plenary sessions and mostly meetings in support of the standards.
As a result, it is technical but also full of insights.
There are two dynamics driving Grid. One is the on going Grid
effort worldwide, mostly concentrated in academia. These networks
are based on resource sharing across large databases or HPC (high
performance computing) centers. In many respects, this is what
might be called the application sector. It is the source of the
experience base from which grid concepts and standards are
emerging. The second dynamic is reflected in the efforts in the
GGF to make grid a part of mainstream computing. This is
beginning to happen with support coming from Intel, HP, IBM and
Oracle. However, for this to directly impact many companies and
individuals, everyone recognizes that grid cannot be outside of
the directions that computing is going. Thus, there are
significant forces to make grid the future of computing.
A review of the working groups and research groups illustrates
how grid is like a distributed OS. These include:
Open Grid Services Architecture (OGSA)
Grid User Services
Grid Security Infrastructure
Data Access and Integration Services
Grid Benchmarking
Grid Protocol Architecture
Grid File System
Grid Resource Allocation Agreement Protocol
Open Grid Service Common Management Model
Open Grid Services Interface
OGSA Resource Utilization
Relational Grid Information Services
Note that this is a short list of the efforts of the GGF, in the
working groups.
Alexander Reinfeld, the local host at Humboldt University gave an
interesting overview of grid. In it he stated:
The network bandwidth is available
The computing resources are available.
Thus, the challenge is to make grids
Reliable – autonomic - and
Provide high level software which supports grid
applications.
The key to value of grid is virtualization.
From this comes the foundation for NGG – next generation grids –
that include the following points:
Grid Nodes are atomic units forming abstractions over
resources,
NGGs will be able to provide services, functions or even new
concepts that are unknown to the clients, and
NGGs will be able to organize on the fly to provide
functionality and behavior that none of the individual
members has.
This has important implications that we will discuss later.
The big news from this event is that GGF and OASIS have combined
their efforts in Web Services. A group of individuals,
independent of but including individuals who have been working on
the specifications for Open Grid Services Infrastructure (OGSI),
began work with OASIS. They are crafting a set of specifications
that will make OASIS more supportive of Grids and will allow OGSI
to use the developing OASIS standards as the foundation web
services. This will put GGF specification efforts at the center
of web services and thus make Grid more mainstream. Steven Tuecke
of Argonne National Labs gave an impressive presentation on these
efforts. A press announcement was posted on the GGF site.
Another highlight of the event was a speech given by the Minister
of the German Federal Government for Science and Education,
Edelgard Bulmahn who was very supportive of Grid. She also
announced the D-Grid initiative in Germany.
Following, Paul Messina, Argonne National Labs, gave a
presentation on "Grid-Enables Computing Environments." He made a
number of references to the report: NSF Advisory Panel on
Cyberinfrastructure that was published a year ago. This is at:
http://www.communitytechnology.org/nsf_ci_report
It is the so-called Allen Report who was the chair of this study
panel. This study is influencing many efforts in grid including
government research priorities.
One of the most important statements made by Paul is that we must
think of
Software as Infrastructure
This has important implications when the network is totally
dependent on the software that underlies it. Such a notion is the
case with the PSTN and the role that SS7 plays. It could also be
said for the OS in the Cisco routers. But the notion of a
software infrastructure in IP networks goes well beyond these.
Virtualization is accomplished when the computing hardware and
network is abstracted by software. Thus, the role of GGF efforts
to set the standards which look like an OS, as seen above,
becomes all the more evident.
The context of the role of grids actually came from an
unannounced session which explored the commercial implications of
grid. Intel gave a talk and IBM gave a presentation. IBM
described a number of grid applications. One was of a large Asia
Pacific telecommunications company that is seeking to use Grid to
handle DSL aggregation. The product foundation includes the IBM
Grid Toolbox that is an enhanced version of the Globus Toolkit.
This runs on the IBM WebSphere platform.
There is enough support for grid in enterprises that a conference
will be held in Philadelphia called:
GT04 Commercial Grid Conference,
May 24 – 26, 2004
Further there is
IEEE CCGrid 2004 "Extreme Clusters, Grids, Networks,"
April 19 – 22, Chicago
18th Annual National HPCC Conference, High-End Computing in
the Wireless World – The Grid and Beyond,
March 30 – April 1, 2004, Newport, RI.
This conference illustrates the significant shifts beginning in
the computing landscape.
Wave Issue 0412 04/02/04 Article 2-01