Blade Servers: Everything on One Card

[rtroxel]

Has anyone here installed or configured a blade server? I've read a number of articles on them, but I've never used one myself, so I'd like to talk with someone who's had exeprience with them.

Here's the general picture:

Blade servers look like network cards and contain a hard drive, CPU and memory chips, as well as an operating system. What's missing is the CD-ROM and diskette drive, plus the USB, keyboard and mouse ports. You push the blade server/card into a slot on the motherboard.

These will be used for the most part as front-end systems for ISPs, intranets, and other company networks. In other words, they would be used for operations like Web-hosting, email or firewalls. Motherboards can be built to hold from ten to sixteen servers. If each motherboard board requires only one power supply and fan, that would be a great saving on power costs. The servers would also cut down on cabling and server room space, as well as the labor required to maintain such facilities. Finally,they would also permit dynamic scaling of services and resources for small, medium and large businesses.

Blade servers are currently using the same chips that are deployed on laptops, and their processors are not designed to withstand the all-day, everyday requirements of a typical Internet data center. But Moore's Law states that every 18 months or so, those chips will be able to have twice the processing power, or become twice as small. Perhaps in five more years, blade servers will be handling complex jobs involving SQL servers, load balancing or graphics.Today, however, they are operating on the outer perimeters of ISPs and intranets where processing tasks are less complex. Future servers could also allow companies to spread processing duties across a variety of, say, ten blade servers.

Dynamic Reconfiguration

Because the blade servers would be removable from the motherboard, this would permit dynamic reconfiguration of system components without powering down the entire server "box". This is an example of "server RAID" where the processing would be distributed across a variety of servers - not just drives, but servers. If the Web site runs slowly on one server, use blade server software to transfer operations to the duplicate Web server. If one server is hit by a worm, you can remove it from the motherboard and switch operations to the next server in line.

System management software would enable the administrator to view operations on all servers, the same way the current RAID software monitors activity across four or five drives. It would also enable the systems administrator to partition drives, set up networking connections, IP addresses, etc..

Blade servers combined with network-attached storage (NAS) would enable standard IP networks "to create publicly-accessible volume-level storage". In fact, this is the storage solution that Sun is currently working on.