Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : EMC How high can it go? -- Ignore unavailable to you. Want to Upgrade?

To: JDN who wrote (11123)	9/15/2000 1:18:36 PM
From: Gus	Read Replies (2) \| Respond to of 17183

Boom Forbes, October 2, 2000 issue In surveys of customer satisfaction, "EMC has the highest ratings we've ever seen--in any customer survey, for any company, *ever," says Carl Howe, a researcher at Forrester Research........ .........To shore up EMC's seemingly invincible position, Mike Ruettgers will spend $10 billion on research in the next five years -- more than the total sales of some of his rivals. Isn't that enough for him to sleep at night? Not quite. On Route 128, anything can happen. forbes.com Here are some interrelated pieces of the puzzle: The Symmetrix has had five major overhauls in ten years. The latest model arrived in April of this year. With as many as 432 microprocessors and as much as 32 gigabytes of RAM,* a Symmetrix system has the power of the largest supercomputers and a price tag to match. This is an excerpt from the McDATA website: Future of SAN - Embedded and distributed file system; - intelligent SAN-smart file system where portion of file system is in SAN; - storage network management; *- concurrent processing and manipulation of intelligent data streams;* - data routing; - server independent storage tasks (peer to peer copying, peer to peer backup, automatic back up using Fibre Channel); - data sharing, data formatting; - security-authorization, authentication, access control. mcdata.com Here's one implementation of NUMA-Q in an EMC fibre-channel based SAN deployed at a proof-of-concept data warehousing project at British Telephone: 3.1 NUMA-Q Server The selection of a NUMA (non-uniform memory addressing) architecture for the server is a significant aspect of the solution. *NUMA architecture is the most recently developed of the major parallel architectures in widespread commercial use and has some major strenghts for database applications.* A NUMA server is constructed from separate processing units or building blocks, where each building block has its own complement of processors, memory and input/output facilities. In the NUMA-Q architecture employed by IMB, the building blocks are Intel Quads, each of which has four processors. Because the quads have dedicated, local memory and I/O paths, their operation is very fast and efficient. A large percentage of all operations is local to a quad and therefore provides no opportunity for quads to interfere with one another. This is in contrast to the SMP architecture where processors are linked to all system resources by a shared bus; as the number of processors in an SMP increases, there is frequent interference in the use of the shared bus, shared memory or shared devices. A key innovation in the NUMA architecture, however, is that the separate and largely independently operating blocks do operate at the hardware level so as to implement a unified memory model. Thus, the software sees a single, large shared memory address space and is shielded from the complexities of distributed memory that are characteristic of MPP and cluster architectures. In contrast to these systems that employ some form of distributed memory visible to the software, NUMA systems can operate with much simpler and more efficient software. NUMA thus promises to offer the efficiency and software simplicity of the SMP architecture combined with an element of the scalability of the MPP architecture. Scalability of the server platform is a key concern of users involved with very large scale data warehousing. The reason is that data warehouses of all sizes continue to grow at a rapid pace. Thus, even users moving beyond the frontier to create the largest database in the world by some measure must anticipate that their data warehouse -- in data volume, workload volume and complexity -- is likely to impose rapidly growing capacity requirements on its server over the next 2-5 years. The IBM NUMA-Q Server offers considerable capacity beyond what was configured for the Proof of Concept and is expected to deliver major capacity increases beyond the capabilities of the current product line, over the next several years. Data Warehousing beyond the frontier Proof-of-Concept with an 80-Terabyte System at British Telecommunications wintercorp.com EMC stock dropped close to 10% when they announced the DG acquisition last year because the Data General server business was widely considered to be a dog even though the Clariion mid-range storage business and ccNUMA patent portfio were widely considered to be top drawer. While the server business continues to look for a bottom, it did manage to post some interesting market share numbers last year. Data General Aviion 1999 Market Share (IDC) Wintel Intel Unix $ 25,000 - $ 50,000 9%(3rd) - $ 50,000 - $100,000 42%(1st) 37% (1st) $100,000 - $250,000 84%(1st) 11% (3rd) Lastly, some of you will remember that EMC inherited a patent infringement lawsuit between DG and IBM over NUMA technology. It was settled earlier this year with IBM paying a small amount to EMC and with both companies agreeing not to sue each other for the 5 year term of the original cross-licensing agreement. It's just as well because despite the many anecdotal reports about IBM slashing prices on the Shark by as much as 50%, IBM was only able to muster the following: Estimated Shark sales - 2000 1Q2000 -- ~$150-160 million 2Q2000 -- <$200 million Ouch!