CrossNodes Briefing: Blade Servers
IT managers in every large company and ISP know about space and energy. The multiple servers these companies support demand space, and the servers generate heat that requires air conditioning on top of with the electrical cost of running the equipment.
Blade server technology could provide a solution. Essentially, these units are rack-mounted servers on a card. They feature Intel's Tualatin or Transmeta's low-voltage processors along with necessary components, and vendors include disk storage systems and I/O support, including network interfaces, on the card. Most of the initial systems occupy approximately 1.5 inches of rack space, which is called 1 unit (1U).
Although some vendors already offer server enclosures that will accept additional processor cards, the blade servers represent an innovative implementation. The products implement new chip sets and low-power processors that were originally developed for the portable computing market to create a low-profile and highly reliable server. The racks also provide redundant power supplies, and vendors typically support network diagnostic units to report problems within an individual server.
Managers running large clusters of servers will find the blade server concept interesting. Still, they need to investigate these devices carefully before they decide to implement the processors. Such concerns as network management, storage scalability, reliability, support for current operating environments, and vendor viability become key elements in the decision.
In essence, blade servers can condense a room full of servers into a single enclosure or rack. Just as network management and diagnosis is important with any group of servers, it becomes critical for blade servers. Typically, hardware developments and innovations enter the market, and software and network monitoring capabilities lag behind. Therefore, IT managers must carefully assess the following capabilities:
- Single point of access -- can the IT manager monitor the performance and operating parameters for each processor from a single console? This console also should support remote access to allow managers to reconfigure the blade servers when necessary.
- Performance monitoring -- will the software generate alerts when a component or blade server starts to fail? Some products take proactive steps to identify and isolate the problem or the server. In addition, managers should insist on some method of monitoring data traffic to ensure that the servers' work loads remain balanced.
- Compatibility with other network equipment -- does the blade server software permit managers to consolidate information from other network components and check on the entire network's performance?
More servers and more disk storage do not necessarily translate to more efficiency. IT managers who oversee several servers in a group must guard against overloading the disk drives. This is especially true if the data is not distributed across several servers. Blade servers, like most server technology, do not directly address this problem. Some networks, however, now use Storage Area Networks (SANs) to enhance performance, and IT managers will need to verify that such options exist before they move to blade servers.
Reliability remains a determining factor in any network's success. IT managers considering blade servers need to determine whether the equipment's software supports failover capabilities and redundant components. Many blade servers come with dual power supplies and hot-swappable components. In addition, some allow technicians to isolate a single processor blade and reboot that card without affecting other blade servers. These types of capabilities become critical when managers consolidate the entire computing center.
As an emerging technology implementation, blade server vendors currently limit support to Linux or Unix based operating systems. For most large companies, this may not present a problem. For those companies that run Microsoft Windows on their servers, however, it represents a conversion and a learning curve. Nearly all the vendors plan to support Windows in the near future, but the current lack of support may force some IT managers to postpone a decision.
In addition, the vendors who pioneered blade server technology were new companies. Several of those companies now struggle as Hewlett-Packard and Compaq responded quickly, and IBM announced future plans that, if implemented would take blade server technology to a new level. Most of the other, established vendors will follow and release versions of blade servers over the course of the next year. IT managers, if they choose to implement the technology now, must ensure that the vendor is financially sound and able to continue competing in the market. Such key areas as support and the companies' ability to continue innovating as the market attracts larger and stronger vendors must be considered.
For the IT manager, blade server technology promises to save space, lower energy costs, and simplify the task of maintaining a room full of servers. Gartner Research expects limited growth in the blade servers over the next year, but IDC expects the value of the market to approach $3 billion by 2005. Innovations in storage technology, symmetrical processing, and management software must occur for the company to achieve that type of growth. Only then will the blade servers represent a new way of viewing the way companies establish servers and become a scalable solution to the ever increasing demand for processor cycles.
IT managers may decide to wait awhile to see how blade server technology develops before they revamp their server rooms.
Gerald Williams serves as Director of Quality Assurance for Dolphin Inc., a software development company. Williams has extensive background in technology and testing, previously serving as Editorial Director with National Software Testing Labs (NSTL), Executive Editor with Datapro Research, and Managing Editor of Datapro's PC Communications reference service.