 |
|
DAILY NEWS AND INFORMATION
FOR THE GLOBAL GRID COMMUNITY / SEPTEMBER 22, 2003: VOL. 2 NO. 38
|
Special Features:
COMMERCIAL GRID DEMONSTRATED BY
IBM AND T-SYSTEMS
By Uwe Harms,
Harms-Supercomputing-Consulting
IBM Research, the Lab in Böblingen and T-Systems signed a partnership to
jointly develop basic technologies for e-business on demand and to gain more
flexibility in IT and telecommunication infrastructures and their
applications. In the T-Systems' high-security computer center in
Frankfurt-Heddernheim, Germany, both partners demonstrated first results. They
used an IBM eServer BladeCenter with a Grid middleware to automatically add
and replace blades, to react immediately in cases of disaster and recover the
situation. The operating system was Linux, the management blades ran a
combination of Tivoli and Globus.
IBM Research Lab and T-Systems agreed in an innovation partnership, IBM
came
from the on demand computing, T-Systems saw the virtual computer center and
managed business flexibility. The virtual center improves the flexibility and
reduces costs and complexity. It contains logical objects as archive, data,
communication and archive with services on top of it. In the middle resides
the resource manager and all is running a Data Center Operating System, an
operating system on top of the different operating systems.
They started with a workshop on using Grid technologies in legacy
environments. Actually, the show case demonstrates the basic principles of the
Grid vision. The nest steps are directed to a commercial applicability. The
first result of this cooperation is a practicability study. It showed how
computers at different locations independently work together -- like a virtual
computer center -- and thus realize Grid computing in a commercial
environment.
In contrary to the scientific Grid computing, computer power is not the
main
issue. In the commercial world, topics like take precedence:
- efficient use of available resources
- optimal and automatic managing the computer center
- managing service level agreements (SLAs) and the quality of service
(QoS)
- recover data in case of disaster
- improvement of the continuity of the businesses.
The new flexibility of computer power and the improved availability allows
new
offerings for the customers. With the demonstrator IBM and T-Systems showed,
how the concepts of Grid computing can change the IT world.
The demonstration took place at T-Systems in Frankfurt-Heddernheim. The
base
was an IBM eServer BladeCenter with two management blades. They ran a
combination of Tivoli Automation and the Globus Grid software. The partners
improved this software layer -- the e-Utility or Grid-Layer. The two blades
worked as a file serve too. Two compute blades waited for tasks, an other
blade was inserted and ready but not in operation. The fourth blade lay on the
table.
The partners started three applications, which demonstrated the broad range
and applicability. The clock served as a stateless application. It was not
allowed to run on the same blade as the Web shop. Then they started the
transactional application, the Web shop. It automatically choose the second
blade. A compute-intensive state-full application ran on both blades and
computed and displayed fractals.
The Grid software, developed by IBM and T-Systems, automatically manages
the
computing resources, allows adding and removing of them, supervises the
dynamic distribution of the compute load, controls the SLAs and QoS and offers
new methods for disaster recovery.
When rising the fractal computation, the higher load reduced the part of
the
Web shop. This violated the defined SLAs and QoS -- the predefined rules. The
Grid-Layer immediately recognized the situation and added -- on demand -- the
third blade as an additional resource. As it was not in an operational state,
it was configured, operating system and the application was installed
automatically. The capacity was added within three minutes. Now the
requirements -- the rules -- of the Web shop have been fulfilled.
They removed the active blade with the Web shop. This simulated a severe
crash. The Grid-Layer immediately copied the applications on the other two
blades. Although the performance system was reduced, the applications are
still running. The customer of the Web shop did not notice the crash. All the
transferred data are kept. This opens new possibilities in the case of a
disaster. Additionally, software automatically recognizes the insertion of a
blade too and starts it without an interrupt.
Now, the partners will commercialize the Grid-Layer. Next steps are
directed
into the managing of heterogeneous computer systems to integrate legacy
systems too. Other issues are the accounting of the computer resources and the
security.
|