Subscribe
About
  • Home
  • /
  • Hardware
  • /
  • How to deliver the RAS factors on Wintel architecture

How to deliver the RAS factors on Wintel architecture

Server vendors continually reinvent the wheel, seeming to learn none of the lessons of the past, says Bernard Donnelly, consultancy services manager at Unisys Africa.

Reliability, availability, serviceability - the well-known RAS factors. They`re what users want. They want their servers to be reliable, available and serviceable: the three attributes are interlinked and non-negotiable from a user`s perspective, yet server vendors, generation after generation, seem to forget these lessons.

So it was that IBM and the Bunch (Burroughs, Univac, NCR, Control Data and Honeywell) learnt what it took to deliver the RAS factors. However, their extreme cost and proprietary nature opened up the potential for open systems-based Unix servers, which by and large failed the RAS test in their first iterations.

As Unix vendors gradually got it right, the relatively high cost of Unix systems opened up a gap for systems based on Intel and Windows architectures, which, as millions of users will attest, failed the RAS test quite abysmally.

The general rule is that the lower you go down the cost ladder, the less reliable, available and serviceable systems become. Companies are forced into a trade-off, swapping one for the other until they find a happy medium. What they would like, of course, is an Intel-based server running Windows, because of its comparatively low cost, with features such as:

* Continuous 80% or greater utilisation with automatic reallocation when necessary;

* The ability to run over 60 applications in the same environment without a service outage in years;

* Clusters of up to 512 processors up to 200km apart;

* On-the-fly reallocation of resources;

* The ability to manage both 32-bit and 64-bit environments as a single entity;

* Clusters that manage themselves automatically and have no single point of failure;

* Support for industry-standard peripherals;

* The processing power of 100 servers consuming less than a tenth of the power of conventional servers; and

* An entire data centre that can be managed from a single console and integrates into all major enterprise management systems.

In an open letter, Mike Chuba, editor-in-chief Hardware Platforms for Gartner, says: "Commodity-based hardware technologies will play a major and defining role in future server architectures. Budget-constrained enterprises are looking for means to cut costs - not just incrementally, but dramatically.

"Driving commodity technology into the data centre are Intel`s new, more powerful processors in the form of its new Itanium 2, nicknamed Madison, its Xeon MP nicknamed Gallatin, and enterprising server architectures from vendors with a history in high-end systems, such as Unisys`s Cellular Multi-Processing (CMP) technology in its ES7000 servers."

"The midrange enterprise server space will continue to be squeezed by the volume server space over time, as volume servers continue to move up the food chain with increased performance and scalability," says Lloyd Cohen, research director, Global Enterprise Server Solutions at IDC. "However, midrange enterprise servers will provide a cost-effective alternative to more expensive high-end enterprise servers because of the midrange enterprise servers` advanced functionality, their high availability, and their partitioning capability to support demanding enterprise workloads."

While these user demands warrant enough attention from the research firms to result in premium reports, South African companies have already garnered the benefits. For instance, integrated health fund manager Medscheme is a demonstration of the scalability and reliability available, having invested in four servers, which it expects to last five years. A spokesperson states:

"We are enjoying mainframe-like levels of reliability, uptime and performance from a system with up-front and running costs that come in at a fraction of the price of the proprietary alternatives."

Other South African clients have shown the consolidation capability of the ES7000, aggregating the load previously run on over 80 servers into a single server cluster. The solution is so reliable that they have not had a single application outage in more than 21 months of operation.

By carefully balancing architectural changes with commodity technologies, a few vendors have already answered users` needs and with the increased attention this activity is now receiving from all the major vendors, and with new and better technologies hitting the shelves, users are in for a reliability, performance, scalability and flexibility festival of choice, at the lowest price point imaginable.

Share

Editorial contacts