Verizon, which recently expanded its cloud services, is now undertaking a new approach to cloud-based processing. The telecommunications giant is deploying AMD’s SeaMicro servers, which link together hundreds of cores to create a single system that requires significantly less power and space than traditional servers.
In 2012, AMD obtained the technology behind these servers when it purchased low-power server vendor SeaMicro. Verizon and AMD said they have co-developed additional hardware and software to optimize the performance and reliability of the SeaMicro SM15000 servers, which will be available in the public beta of the Verizon Cloud, coming in fourth quarter. The companies said in a statement that the deployment “ushers in a new era of enterprise-class cloud services by enabling a higher level of control over security and performance SLAs.”
The SeaMicro SM15000 system provides 512 compute cores in 10 rack units, with 160 gigabits of I/O networking, more than five petabytes of storage, and a 1.28 terabyte high-performance supercompute fabric. The companies said the system removes the need for top-of-rack switches, terminal servers, hundreds of cables and thousands of unnecessary components. The SeaMicro server product family supports AMD Opterons, Intel Xeon Sandy Bridge and Ivy Bridge, and Intel Atom N570 processors.
‘Programmable Server Hardware’
John Considine, CTO at Verizon Terremark, told news media that this solution gives enterprise clients “control of their infrastructure” by providing the “speed and flexibility of a generic public cloud with the performance and security they expect from an enterprise-grade cloud.”
The companies said that the SM15000 server provides the only available “programmable server hardware” and offers a high-bandwidth, low-latency programmable interconnect fabric, along with a programmable and control plane for network and storage traffic.
New capabilities include virtual machine server provisioning in seconds, fine-grained server configurations that match specific requirements, speeds available in 500 MHz increments and DRAM in .5 GB increments, shared disks across multiple server instances instead of virtual machines having their own dedicated drives, and quality of service definitions up to 5000 IOPS. There’s also traffic isolation and data encryption whose firewalls meet Department of Defense specs, and guaranteed network performance for every virtual machine.
Charles King, an analyst with Pund-IT, said that Verizon’s decision to use AMD’s SeaMicro was “a validation both of the SeaMicro technology and of AMD’s decision to acquire the company,” a deal that originally “took a lot of people by surprise” when the microprocessor maker bought a server company.
He described SeaMicro’s approach as “pretty radical.” The common practice these days, he said, is employing virtualization to “maximize the use of the server,” dividing up the processing capability to support multiple workloads.
But, King added, SeaMicro approaches the problem “from the other direction,” allocating ½ GHz server cards to support applications as needed. “Rather than carving up a larger CPU with virtualization,” he said, smaller CPUs handle specific tasks.
King noted that, among other things, this means the security concerns many companies have about running in multi-tenant environments, where they share processing power, may be alleviated because Verizon can guarantee that each company's applications run only on their processors and storage.
But, he said, at this point it’s still “somewhat cloudy” how many SeaMicros will be deployed initially and what using them will cost.