SI
SI
discoversearch

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.
Strategies & Market Trends : Gorilla and King Portfolio Candidates -- Ignore unavailable to you. Want to Upgrade?


To: saukriver who wrote (51039)4/23/2002 11:41:20 AM
From: paul_philp  Respond to of 54805
 
In a hurry so here is something from Intel. I will pop back later with why I think blades are so important.

Paul

intel.com

server blades emerge

The Internet and Web-based commerce has given rise to a class of economical, front-end servers that serve up Web pages, power searches, and enable transactions. Unlike the powerful multi-processor machines that drive databases and back-end systems, these single- and dual-processor systems are typically clustered in large farms that make it easy to drop in additional servers to meet demand.

It didn't take long for fast-growing Internet enterprises to fill enormous data centers with servers in slim "pizza-box" form factors (known in the industry as 1U configurations). These compact chassis configurations enable firms to stack up to 42 servers in a single, seven-foot high rack. But as demand continues to grow, the challenge is shifting from simply scaling out operations to managing the growing electrical, thermal, and space challenges of such high-density facilities.

Enter server blades. These new servers place the processor, system memory, network connection, and associated electronics on a single motherboard, which slides into an enclosure that can contain dozens of other blades. Unlike traditional chassis-bound systems, server blades share the chassis, power supply, fans and cabling with other blades in the enclosure. The result: simplified physical environments that allow thousands of blades to be deployed at optimal cost.

Keep it Simple
Server blades have a lot of appeal for IT organizations looking for space-efficient solutions with higher levels of serviceability, scalability and manageability. According to research firm International Data Corp, server blades will account for about 27 percent of server unit sales by 2005. While early models are primarily single- and dual-processor configurations, IDC expects a steady migration to higher performance models. Quad-processor blades could account for 35 percent of the server blade market by 2005, according to the IDC report, while dual-processor products could represent 52 percent of units shipped.1

In fact, server blades only continue a trend toward more incremental deployments. Currently, 37 percent of servers shipped come in rack-mounted form factors, according to IDC. That compares to just over five percent at the end of 1999. What's more, slim 1U form factors are the fastest growing segment for rack-mounted servers, accounting for 40 percent of units shipped. Of course, multi-processor pedestal servers hold much higher portions of these sectors by revenue, due to their higher unit costs.

Manageability makes server blades an attractive option. Installing, servicing, and removing blades is much easier than working with chassis-mounted servers. Shared power supply, cabling, fans and storage reduce the number of redundant and "failable" components in the environment. Support for industry standard management tools allows IT managers to readily monitor, configure, and troubleshoot systems. Organizations can also scale out operations in much smaller, more precise increments.

Perhaps most important, blades offer the highest available densities. In a sector where companies often benchmark performance in terms of transactions per watt per unit of floor space consumed, the ability to squeeze more power in less space is critical, especially in today's environment of increased emphasis on reducing overall operational costs.

Server blades may be the future, but they won't replace existing server configurations. Rather, blades will likely gain prominence in specific, high-density computing environments, such as enterprise and service provider data centers that specialize in Web serving and other front-end operations.

Intel Introduces Low-Power Server Processors
Intel has recognized the importance of low-power server processors for some time. For nearly six months, the Intel® Pentium® III processor with 512KB L2 cache has provided a superior mix of high-performance, lower-power consumption, and reduced thermal output for front-end servers of all types. Add to this: an advanced 0.13-micron chip manufacturing process and efficient FC-PGA2 packaging enable efficient operation. The result: low voltages and thermal output even at clock speeds above 1GHz.

In November 2001, Intel introduced a new ultra low voltage version of the Pentium III processor with 512KB L2 cache on uFCBGA packaging. The new processor significantly improves on the power management and thermal output of the earlier processor, making it ideal for server blades and other ultra-dense server environments. Equally important is the low power Intel® 440GX chipset, which is ideal for these new low-power Pentium III processors. This chipset addresses up to 2GB of system memory and features the same error-correcting code (ECC) used to ensure data integrity in powerful, multi-processor servers.

For data centers that measure performance in terms of transactions per watt per unit of floor space, these introductions are a welcome development. The new processor and chipset offer significant reductions in power consumption, lowering the strain on electric and climate control systems even as they allow data centers to deploy more servers in a fixed space.

Server blades are emerging as an intriguing option for IT organizations that must deploy large numbers of front-end servers. Over the months to come, Intel will continue to introduce high-performance processors that offer lower power consumption and thermal output. These advances will play a critical role in enabling dual-processor server blades, for example, and will allow IT organizations to deploy the best possible hardware for each task.

1Source: IDC, ESV, 8-01



To: saukriver who wrote (51039)4/23/2002 3:46:33 PM
From: paul_philp  Read Replies (1) | Respond to of 54805
 
Saukriver,

Let me see if I can answer your questions about blade servers. A blade server is simply a motherboard with at least one CPU, some local disj and some memory. Blade servers have no external chasis. They are mounted into cabinets sized units called racks. Racks can handle 8 - 64 blades. The origical driving force behind rack-based servers was floor space and power. During the boom, space in the data centre became as expensive as NYC real estate and a NYC solution was needed - build up. Thus we entered the wacky world of MIPS/Sq.Ft.

The rack-mounted server as cut down on power usage. By removing the fan and the power supply from the server total power consumption dropped. All the servers in the rack shared a power supply and a centralized cooling system. Think of it as a computer without da noise. A significant number of large web-sites are run by rack-mounted servers today and the role of rack mounted servers is pretty much limited to the front-end web server capabilities.

So far this is little more that +1 marketing for a product on mainstreet. Things get interesting when you add some of the newer technologies being commercialized as you read this:

- Intel Itanium processors (commidity 64-bit server chip)
- Linux (a Unix which is good enough AND cheap)
- Networked Storage (all our blades can share a file system)
- Infiniband (a high-speed external bus to connect the blades)
- Clustering software (share the workload between all the blades)
- 10 GigEthernet, 10 Gig FibreChannel

Add all these ingredients together and something quite magical happens - you bust up a core metaphor which lies at the heart of IT technology. Up until now there has always been a direct connection between the task to be preformed by the 'computer' and physical machine(s) which run that task. No more. With the emerging blade architecture and the newer interconnect technologies you can strap a bunch of (dirt cheap) blades together and have a highly reliable, highly scalable, massively distributed server resource.

Now the computing resources can be doled out on a as needed basis. There is no more direct connection between the user, the app and the hardware. Exactly the same thing is happening in storage. As a matter of fact, the blade server itself is much like the disk drive, a high volume commodity item.

One interesting question is what happens to the software companies when this happens. Almost all software licences depend on counting the number of CPU's or the number of users. In the massively distributed world this may not make any sense. Do all software companies become service providers renting out their capability on a as needed basis? Maybe.

That is a quick overview of my thinking.

Paul