Blade servers have been touted for some time as a solution that can improve server management and cut costs, and organizations are increasingly calling on blade technology to deliver on those promises. To meet the growing demand for such products, all of the big players--including IBM, Dell, and Compaq--have stepped into the arena and introduced blade server lines.
Many organizations have held back on making the jump to blade technology, however. Some simply haven’t had the budgets to make the investment or had already invested in different technologies before blades became widely available. Others may still be unconvinced of the potential benefits of blade technology.
I’ll give you some details about what blades are, how they achieve cost savings, and how they streamline management. By examining this solution from the vantage point of someone who has deployed it, I may help you decide if blade servers would be a valuable investment for your organization.
What’s so great about blades?
The term "blade” refers to how thin the servers are. Blades resemble PC add-in cards rather than full-fledged servers. And, in fact, blades really aren’t full servers until they’re plugged into the enclosures.
Organizations are realizing benefits from blade servers in three key areas:
- Reduced space requirements
- Reduced energy consumption
- Improved server management
The basic idea behind blade servers is consolidation. By allowing servers to share resources provided by a single chassis--such as network connections, power, and cooling fans--the individual servers can be made much more compact. Just the CPUs, drives, and memory, for example, can be housed on a single blade. The design essentially allows you to cram a lot more processing power into a smaller space.
In terms of hardware, two components comprise the blade server solution: the blades themselves and the enclosures that house them. The enclosures essentially connect the blades installed to the shared resources, and you can configure the enclosures according to your needs. Because of this consolidation, a chassis can house a dozen or more servers. Vendors make various claims about how many blades can actually fit into a single enclosure--some say as many as 30--but this is highly variable depending on the functionality put into the enclosure.
Quadrix Solutions CTO Tom Palmieri says that data center density is a key benefit blades offer. And the processing power density comes at a lower cost in terms of energy used.
"You can get three times more servers in a cabinet than you’d normally be able to," said Palmieri, "and you can basically use approximately half the power of what the whole cabinet filled with 1U (unit) servers would use.”
Because the enclosure supplies functions that servers would normally house individually in each box, the blades are equipped with the bare necessities.
"The blades don’t have a lot of the circuitry a full-fledged motherboard needs to have,” said Palmieri. "So you don’t have duplicates of the same equipment in cases where it’s a waste to do so.”
The scaling down of what’s installed on the blade is part of the reason they’re more economical in terms of space and power consumed. Instead of having 12 different servers, each with its own power supply, you now have 12 servers housed in one box with one power supply. Palmieri said an important factor to consider is that sufficient power must be supplied to the multiple CPUs. The end result is that an enclosure with multiple blades consumes less power than the same number of standalone servers.
One of the touted benefits of blade servers is streamlined management, but Palmieri points out that it may not be a clear-cut advantage. While the management software released for blades has potential benefits, it also has its limitations. One issue, said Palmieri, is that the software is proprietary. If you buy blades from HP, you’re locked into using HP’s management software whether you like it or not.
Blade vendors are taking either software- or hardware-based approaches to managing the servers, and Palmieri said that KVM (keyboard video mouse) switches are typically used to connect to the servers, but this is just a local solution.
"One solution,” said Palmieri, "is to daisy chain a special port from one machine to another all the way down however many blades you have and connect them all out the back to one KVM system.”
The hardware approach works well, he said, as long as hardware failures don’t cause issues with the switching.
In some cases, a software-based solution might involve using Win2K Terminal Server installed on one blade to gain access to the others.
Mark Palmer, Manager of Technical Services at Merant, a PVCS software development firm, said he uses PC Anywhere to manage the Compaq blades his company recently deployed. For the initial installation, Palmer said he had to use a special adapter that came with the blade to pull up a console to configure it, but after setting things up initially, he said the console was no longer needed.
"Once the OS is installed and up and running,” said Palmer, "you can install VNC or PC Anywhere or a similar product to access the OS remotely.”
Even though this setup works well, Palmer said it’s still a bit of a hassle initially to have to use the adapter to access the console.
Blades may be a good option for saving space and cutting operating costs, but they’re also currently hampered by some limitations. One of them is processing power and speed.
One of the reasons blades consume less energy is that the CPUs and drives installed on the blades aren’t of the same power level you’d find in a typical server.
"It’s essentially laptop technology, if you will,” said Palmer.
The CPU and drive speeds of blades are slower than those of a standard rack-mounted server. The blades from Compaq that Merant recently deployed, for example, are equipped with 800-MHz CPUs and 4200-RPM drives. Palmer said that while 800 MHz is the fastest speed Compaq currently offers in its blade line, standard servers offer CPUs well into the gigahertz range, and the drive contrasts with the 15,000-RPM speed available in rack mounted servers.
While this means that blades are less costly in terms of power consumption, the performance factor also limits what you can do with them. Palmer said that Merant’s blades function as DNS, WINS, and print servers and that their virus scanning software runs on one of the blades.
"If we were deploying actual applications like Web servers,” Palmer said, "with high availability and high performance requirements, we wouldn’t have chosen blades.”
In many cases, however, blades offer an efficient, cost-effective alternative.
"Blades are very good,” said Palmer, "for similar load-balancing or similar redundancy applications where you’re going to keep most, if not all, of the blades pretty much in synch with one another.”
Both Palmer and Palmieri acknowledged that one feature that’s sorely needed on blades is the ability to connect via fiber channel to a SAN. The only network storage option currently available for blades is NAS.
These limitations push blades into filling specific niches. Whether they’re a good option for you depends entirely on how you plan to use them. New blade offerings in 2003 will address some of these issues. Vendors are introducing blades with speedier processors and more memory, for example. As more organizations turn to blades as a solution, the offerings will likely evolve to take on bigger tasks.
For now, blades remain a solution that fills certain needs and can help managers who are looking to cut operating costs and save space.
TechRepublic originally published this article on 6 November 2002