EnterTheGrid - PrimeurWeekly

EnterTheGrid - PrimeurMagazine is the largest Grid and Supercomputer information source in the world. PrimeurWeekly delivers the news each week in your e-mail box.

PrimeurWeekly 15 June 2009
>Registration still open for "Enabling Desktop Grids for e-Science and Industry" Forum
>EDGeS team calls for participation to joint EDGeS and EGEE Summer School on Grid Application Support
>Bull SAS and Foster Findlay Associates Ltd have entered into a partnership agreement
>Current and future HPDC developments at the International Symposium High Performance Distributed Computing
>Special session on "Cloud Computing & HPC - Synergy or Competition" at ISC'09
>Birth of a star predicted
>Bull equips several French Ministries with globull, the leading mobile security platform
>Computing in the quantum dimension
>The dawn of quantum applications
>Manipulating light on a chip for quantum technologies
>CNRS 2009 Gold Medal: Serge Haroche, physicist and explorer of the quantum world
>Yahoo! raises commitment to Cloud computing with the availability of Yahoo! distribution of Hadoop
>Tech-X selects Terascala RTS 1000 Storage Appliance
>Visualizing the future of scientific discovery
>Korea Meteorological Administration selects Cray to negotiate multi-year supercomputer contract
>Defense Department High-Performance Computing Centers extend their contract with Altair for a third year
>Biomedical Research Institute deploys Force10 Networks to deliver high bandwidth network services
>RedPeak Solutions to partner with 3Tera to provide Cloud computing services for enterprise customers
>HP advances scale-out computing with breakthrough data centre solution
>DreamWorks Animation gets "Extreme" with HP scale-out storage
Special session on "Cloud Computing & HPC - Synergy or Competition" at ISC'09
Hamburg 13 June 2009 On Wednesday, June 24, 2009, the International Supercomputing Conference'09 (ISC09) in Hamburg, Germany will feature a special session on "Cloud Computing & HPC - Synergy or Competition", chaired by Dr. Martin Anthony Walker and Prof. Dr. Dieter Kranzlmüller from the Department of Computer Science at Ludwig-Maximilians-Universität München (LMU). The session features a series of talks by experts from different Cloud providers, e.g. HP, IBM, Google, Amazon, Yahoo, ..., and a panel discussion.
Visit our sponsors
Visit our sponsors

The purpose of this session is to assess current implementations and deployments of Cloud technology, what users are doing with Cloud computing today, and most importantly for the audience of ISC'09, to assess the possibility of high-performance computing in the Cloud. The session will provide key insights for HPC practitioners, observers, analysts, and vendors interested in scalable and distributed computing.

This session consists of two parts, separated by a one-hour break. The session will be opened with the chairmen setting the stage with a brief exposition of the context. The first part will contain presentations from three major IT vendors, HP, IBM and Sun Microsystems, followed by perspectives from a supercomputing centre, namely Jülich Supercomputing Center (JSC).

The second part will start with the presentation from a major software house, Microsoft, and continue with the presentations from three IT vendors driving Cloud computing today, Google, Amazon, and Yahoo!, followed by a round table panel discussion to field questions from the audience.

The eight speakers are outstanding leaders in the fields of high-performance computing and Cloud computing. The session will conclude with a round table to encourage the active participation of attendees.

Recently Prof. Dr. Dieter Kranzlmüller provided an interview on Cloud computing to the German magazineComputer Woche. Here are some extracts from the interview.

Large cloud providers such as Amazon or Google use a multitude of x86 computers in their IT infrastructure, instead of few central servers. Will traditional supercomputers become obsolete in the age of the Cloud?

A classification of applications can provide the answer: supercomputing applications can be appropriately classified in a three-dimensional space with the dimensions computing power, data volume and communication bandwidth. Supercomputing architectures are optimal in one of these three dimensions and offer usually good compromises regarding the remaining two. Clouds offer scalable computing power, and if necessary large memory capacity combined with poor communication performance. In contrast, the highest ranked TOP500 supercomputers are usually equipped with a special interconnect solution that offers high communication capacity. However, the borders between these different systems architectures are becoming increasingly flexible.

In which areas should companies give preference to x86 server farms over traditional supercomputers?

The effective use of a Cloud is dependent on the applications. The Cloud can be used appropriately when dealing with linear processes and independent, relatively small data volumes. For applications with larger storage requirements or closely coupled parallel processes with high I/O requirements, Clouds are often useless.

Which is the most important potential use for Cloud computing in IT companies?

The most important aspects of Cloud computing are on-demand usage of resources, scalability ("scale as you grow") and usage related pricing ("pay per use"). Cloud computing offers a business model that gives the opportunity to "buy" computer resources in a simple way. The infrastructure can be extended arbitrarily (or reduced due to e.g. the financial crisis) to adapt to the current needs.

How interoperable are the diverse Clouds of the main providers?

Currently there is no interoperability between the Clouds. Cloud providers are attempting a "customer lock-in" (similar to the beginning of Grid computing) that makes it very difficult to change the Cloud once it has been chosen. But things are also changing in that area due to pressure from the users who profit from standardized and interoperable Cloud offers. An open cloud API will e.g. be discussed during Open Grid Forum (OGF).

Which are the challenges faced by companies that receive Cloud services from the big players?

Considerable challenges to the use of Clouds are the missing service-level agreements (SLAs) as well as the possible security risk: customers are obliged to trust the Cloud supplier or rather its security infrastructure, and there are scarcely any security guarantees. Frequently, there is additional need for a dedicated "bridge" between data residing in-house and data residing in the cloud to allow an exchange at all.

Which clouds wouldn't you currently recommend?

Until the mentioned challenges are under control, Cloud computing is suited for private use or for areas that are not critical with respect to security. Some of the current problems may be solved in the near future, thus making Clouds usable to a wider range of users.

Suppliers such as IBM propagate a concept of private Clouds, i.e. IT resource pools hosted in-house. Why should organisations build a Cloud infrastructure of their own if IT services can be obtained conveniently from service providers?

In principle, the advantages of external Clouds can be leveraged in an internal Cloud as well, e.g. flexibility in resource management and a higher degree of independence of operating system brand. This flexibility, together with the ability to plan resource capacity and use conferred by Clouds and virtualization often has positive effects on costs.

What do companies need if they would like to introduce a Cloud infrastructure of their own?

Prerequisite for the use of a Cloud is the understanding of the company's own IT processes as well as the knowledge about risks related to IT outsourcing. Provided sufficient network capacity is available, and applications have been identified that can be mapped in the Cloud, there are no other significant obstacles to the introduction of a Cloud infrastructure of their own.

Which technical obstacles stand in the way of the Cloud as a concept?

In many cases the required network capacity that is necessary in order to move data into and out of the Cloud on demand. Given this capacity, a security framework must be available that allows to discriminate between data records that may be stored in the Cloud and those that must not be externalized. It is also worth noting that optimization of application and the detection of bottlenecks is more difficult on an unknown Cloud infrastructure.

More information on the ISC'09 Cloud session is available at http://www.supercomp.de/isc09/itinerary/search&event=24
Source: Ludwig-Maximilians-Universität München

EnterTheGrid - Primeur

James Stewartstraat 248

1325 JN Almere

The Netherlands


mailto:primeur [AT] enterthegrid [DOT] com

© EnterTheGrid - PrimeurWeekly