Clustered Grids Make Their Way Into the Enterprise

Thank advances in the hardware and operating system design, todays computers have the potential to be doubled, as calculated workhorses for distributed computing efforts. And through a series of companies, IT organizations can be expected to use for the first time over-the-counter use of off-the-shelf software and shrink film packaging of the operating system, such as Windows NT 4.0 or Windows 2000 Professional Edition for the development of brain computing power, and their desktop personal computers.

An enterprise customer, as a result of advanced scalability and availability of distributed computing technology, investment banks JP Morgan shares Maikebaiman, Deputy president and global head of interest rate derivatives technology as JP Morgan, traditionally rely on the UltraSPARC-based servers from Sun Microsystems, Inc. to perform such compute-intensive tasks as a long-term derivative calculation.

"We will use sophisticated mathematical model calculated risk, these derivatives, and these mathematical models are very intensive in terms of their computing needs," he explained.

In Bermans account, JP Morgans existing infrastructure - this is based on S unu ltraenterprise servers equipped with up to 2 0 U ltraSPARC processors per box - simply out of control horsepower, and when it came time keep up with the request made by the calculation and its derivatives insurmountable bottleneck efforts.

"With our business growing problem, and as we become more and more complex the model, it enormous strain on the computing power, we have in hand," Berman confirmed.

If it can not find another solution, JP Morgan is not facing the unenviable prospect of a seemingly endless - and costly - a series of hardware upgrade cycle. So, instead of spending hundreds of millions to buy computers from the horsepower Sun, Berman and JP Morgan are determined to take a chance, a totally different approach: "enfuzion," a new distributed computing products from Turbolinux companies. If successful, JP Morgan enfuzion allows use of computing power, hundreds of thousands of distributed workstations - most of whom are running W indowsN T4 .0 or W indows2 000 - perform similar increase in the number of super-computer performance.

"We spent millions of dollars in hardware and has been when you need to double the size of this is very expensive, so we started to look for is a more creative way to address these needs," He confirmed that the.

The technology allows the number of JP Morgans derivatives efforts may actually belong to a more broad headings grid computing. Grid computing refers to the clustering of distributed resources - including super computers, storage subsystems, the traditional desktop computers, and even embedded computing application device - a so-called "computational grid", and then use high Parallel distributed computing efforts. Grid computing itself is often divided into several larger super-peer to peer-to-peer (P2P) basis.

Enterprise vendors are queuing up to support grid computing. In August 2000, Intel, Hewlett-Packard and IBM launched a P2P Working Group champions the interests of grid computing in the enterprise. In addition, in July 2000, Sun purchased the Distributed Computing Project, the so-called gridware company, the development of software, including use of the idle resources in the UNIX workstations similar supercomputer computational tasks.

However, although P2P advocates are still optimistic about the prospects of grid computing, in large enterprises, they also warned that it is not by any stretch imagination of a size-fits-all panacea.

"This is certainly more of a parallel-type things, so this is really on certain types of applications, especially those in the graphics, financial and aerospace industries," commented Stephen Spector , senior marketing manager and Turbolinux. "If you work in these sectors, we can go to and distribution of your computing tasks across 1000 machines, but if you work with the traditional applications, but does not apply." JP Morgan Berman, on the one hand, this shows that although there may be the potential for the development of IT organizations grid computing technology, the traditional applications, these companies will be "There must be more clever when [they ] wrote [its] software " Jiefudaiweishi, Senior Systems Programmer and the Hess oil giant amerada company, has opened up a more traditional distributed computing variation - low-cost Beowulf cluster of L inux . Scalable cluster solutions, such as Beowulf clusters and rs/6000 sp by IBM, after all, blazed a path for todays grid computing tools in the enterprise.

Amerada Hess has been leveraged distributed computing solutions, in one form or another, to help it crunch geophysical information, Davis said, and will therefore Linux and Beowulf cluster operating system of the National 200 workstations, the Dell Computer Corporation is a non-brain-type decision that ultimately help save his company about 2.5 million US dollars.

Amerada Hess is the situation can upgrade their existing IBM rs/6000-based SP2 clusters, Davis allowed, but ultimately determined that a Beowulf cluster is the way to go, from the point of view, not only cost savings, and the performance and flexibility.

Unlike enfuzion and other grid computing solutions, and their ilk - This is usually the use of idle resources across the IT environment - running on high-performance clusters dedicated system, and a dedicated high-speed interconnect, so that they a wide range of computationally intensive applications and implementation.

Course, as Iran roohalamimi, Senior Manager of cluster solutions and Dell seem to indicate that you do not have to deploy a distributed cluster dedicated Linux or Unix system to seek their own interests. After all, Dell has a very successful Windows nt/2000-based clustering program, roohalamimi pointed out, and with customers, such as Cornell University Cornell Theory Center (CTC) to 256 engineers processor, 64 node Windows 2000 Cluster 16, the four-way PowerEdge servers. "When we say, Beowulf, Dell here, we are talking about building high-performance clusters regardless of the use of the Linux or Windows NT/2000, and their industry-standard hardware" , roohalamimi explained.

Distributed Cluster solution is generally dependent on technology - known as "messaging" - that facilitate communication between the nodes. As long as your messaging software installed on your goal, workstations and servers, roohalamimi argue, you can benefit from a Beowulf cluster type, and even skip the traditional 10 - or 1 00 and B t Ethernet.

"You can still use it, even if you want to use a set of workstations are sitting idle in the evening, you only need to ensure that you have messaging software, installation of these workstations," roohalamimi explained. "Physical interconnection play an important role in this regard, but you can support these applications, and even to the standard 10 - or a 00M B Ethernet. If you have a good implementation of messaging, they will do good. " There are already a variety of open-source messaging can be used to achieve the Linux platform, but after manufacturers, such as MPI Software Technology Corporation, the two companies sold messaging solution, called MPI The pro-apply to the Linux and Windows NT/2000 - organizations also could switch to in-house distributed similar Beowulf cluster solution, with its, and their ilk.

"It is possible to use the technology or MPI combine long and construction units or other programs to achieve the same effect [as a Beowulf cluster]," agreed Dell roohalamimi. And if you deploy a group of Windows 2000, he noted, you can use the latters support for the Internet and networking management capabilities, in many cases, not available on Linux. CTC, the use of such a software component - cluster controller, also sold M PI software technology - parallel batch and interactive job scheduling for its Wi ndows20 00 dedicated computer cluster.

"The bottom line is that its easier than ever to deploy scalable Windows or Linux clusters using industry-standard hardware, and said:" roohalamimi. -- Stephen s woyer related article: Microsoft Customer Reference forward SQL clusters high-availability cluster market shrinking ENT guide Microsoft Cluster Technology