Unless you’re in the business of selling servers or electricity, the announcement by Pivot3 of what it calls “Serverless Computing” probably is good news. The company’s technology promises to eliminate the need for physical application servers in most datacenters. If the technology delivers as promised, users will benefit from lower hardware costs, less energy consumption, and less rack space than a traditional approach would require.
Pivot3, which specializes in high-end storage solutions, gets around the need for physical servers by offloading processing tasks to the storage system. The technology creates virtual servers on the storage array, so the storage node also is a compute node. It’s like consolidation, cranked up a few notches.
Serverless Computing is built around Pivot3’s RAIGE (RAID Across Independent Gigabit Ethernet) platform — a clustered IP SAN made up of x86-based nodes called Databanks, each with its own disk drives, network connection, power and fan, and RAIGE software implementation, which distills it all into a virtual storage pool. The company’s HyperPath technology links the Databanks, allowing multiple Gigabit Ethernet connections and providing for linear scaling of performance.
For its new serverless system, Pivot3 has replaced Databanks with Cloudbanks, which are more powerful with an added CPU, more RAM and more network ports. The other big addition is the Xen hypervisor, which virtualizes the Cloudbank for use as a storage device or a compute device.
“What we’re doing is taking the workload from very busy processors and moving it to the storage node,” says Jeff Greenwald, senior director of marketing at Pivot3. “The storage node isn’t very busy. We found that the processors in a Databank were typically busy only 20 to 30 percent of the time. With this extra compute capability inside the storage device, we virtualize the workload that would run on an application server and move it to the storage controller. We can virtualize the workload and you don’t have to run it on a server. You don’t even have to buy a server. We think it’s actually a cool evolution in virtualization.”
Operating Greener for Less Green
CAPEX and management benefits aside, cutting down on servers also means those servers won’t be taking up rack space, won’t be sucking down electricity and won’t be throwing off heat. Everyone’s mileage would vary, of course, but Pivot3 cites the case of one customer, in the video-surveillance business, that went from 15 conventional servers to zero. Rack space dropped in half, and power and cooling consumption was 44 percent less. Pivot3 combines energy and acquisition costs to figure that this particular customer, over three years, would save about 22 percent. Its energy bill over that period would be about $70,000 instead of about $126,000, the company says.
Why worry about energy? A 2007 report by the EPA stated that energy consumption by datacenters had doubled in the previous five years, and will nearly double again by 2012 to about 100 billion kilowatt-hours and a cost of about $7 billion per year. Clearly, it would be a good thing to bring that bill down a bit.
Energy consumption of IT equipment “most definitely” comes up in conversations about corporate projects where datacenters are involved, says Ann Edminster, architect and owner of Design AVEnues, a leading consulting firm for power-efficient, sustainable building. She says Pivot3’s approach is “refreshing” because its very premise is less equipment. “Virtualization, in general, is going to yield energy savings,” she says. “Pivot3 is bringing virtualization to a specific set of applications, which is particularly important because of the technology trend towards increasingly more rich media usage.”
Pivot3 promotes its technology as one way for buildings to meet the standards of the LEED (Leadership in Energy and Environmental Design) Green Building Rating System, designed by the U.S. Green Building Council to improve energy efficiency for new or existing building projects. LEED ratings, sometimes required for new buildings, require that a project include efficiency improvements 14 percent better than the building code requires.
Surveillance Roots, Broad Future
Pivot3 emerged in 2003 amid swirls of rumors that it was developing a new storage architecture. The company was founded in Houston by former Compaq engineers and some of the original designers of RAID. They took that concept and stretched it into the RAIGE operating system, which runs in parallel on each node in a Pivot3 storage array, distributing RAID parity across all nodes and protecting data from individual node, disk, port or switch failure. The software manages all the storage pools and block data without administrative intervention.
The company’s first products provided “high-definition” storage for applications that had to store lots of video data. Its customers tended to be, in one form or another, in the surveillance business: gaming, public security, education, transportation, and, naturally, the federal government. For example, the 101 Casino in Petaluma, Calif., has a digital video monitoring system with many, many 1-megapixel cameras streaming data to the Pivot3 Databanks. (One camera can generate a terabyte of data in just a few days.) Everything is stored on Pivot3 arrays, allowing the casino to retrieve precise images almost instantly and no longer have to deal with thousands of tapes. When storage demands increase, the casino will be able to grow its cluster by just adding new Databanks. Likewise, Serverless Computing systems can scale quickly by adding new Cloudbanks, which will bring not just more storage capacity, but compute power, as well.
Pivot3 would not disclose any adoptees of Serverless Computing, but Greenwald says they’ll do so in September. “Virtualizing compute on the storage node is an evolutionary transition from today’s virtualization technology,” he says. “When customers hear what we’re doing, those who understand virtualization and consolidation immediately grasp the value of what we’re doing. For others, it’s a longer dialog.”
Pivot3 says most IT installations with “high-capacity, I/O-intensive workloads” will be best able to take advantage of its serverless approach. “We think that many different applications and verticals fall into that category,” explains Greenwald. “Over the next year, we expect rapid adoption in health care and medical applications, and video-serving applications, such as video on demand and entertainment. A bit later, we expect adoption by mainstream IT environments with very high-I/O and high-capacity requirements. Small or medium-size businesses don’t typically have hundreds of terabytes of storage, so we might not be a fit for them. But larger enterprises that do would clearly benefit from incorporating Serverless Computing into their infrastructure.”
By virtualizing servers on top of its storage arrays, Pivot3 is giving datacenter managers one way to meet growing capacity demands without adding as much to the bottom line as they would with old-school IT expansion. Fewer servers to purchase and contend with is certainly an appealing proposition, and reducing power consumption by even a bit has got to be good. If a $7 billion electric bill shows up, Dad is not going to be happy.