Intel invites its industry partners here to experiment with hardware and software configurations and hammer out reference designs that service providers and enterprises can use to equip their data centers to offer new cloud services.
For Intel, the investments in the Cloud Builders program and other, similar initiatives are designed to help the chip maker hold onto its dominant share of the cloud server market in the face of looming competition.
“Intel wants to make sure that as people build these clouds, they are using Intel processors,” said Nathan Brookwood, principal analyst with Insight 64.
This week, Intel invited reporters to tour the Portland data centre, to see some of the reference architectures being demonstrated and hear from software partners who work with Intel in the cloud.
While some companies merely design products and sell them, Intel does much to shape the market into which its processors will be sold, even creating uses that the market may not have discovered yet by itself.
“Intel has always been very interested in trying to increase the size of the market it serves,” Brookwood said. “A lot of chip suppliers just look at chips and say, ‘We’ll make the chip and someone else will figure out how to use it.’ Intel has always taken an expansive view … and tries to make sure that there are places for all the chips they can produce.”
The Cloud Builders program is one example of how Intel does this.
“We have customers who want to do something in the cloud but they want a ‘how-to’ guy,” said Jason Waxman, general manager of high density computing at Intel. “That’s what we’re trying to do with Cloud Builders.”
Working with partners, Intel so far has developed 25 reference architectures and expects to have another 25 by the end of the year. They fall into two broad categories: building the cloud and “enhancing” the cloud. Each is supported by a white paper that describes how to set up the required hardware and software.
One reference architecture Intel showed off this week lets data centre managers move workloads around based on the power consumption of individual servers. The example used Dell servers, VMware’s vCenter management software and a software plug-in developed by Intel.
The plug-in lets users set a maximum power consumption for each server. If the server reaches that threshold, the software shifts part of the workload to a server that is using less power, keeping power consumption balanced across the data centre.
Another reference architecture uses software from Citrix Systems to migrate workloads between a private cloud and a public cloud. It sets up a secure tunnel and retains settings for security and networking and even the IP address for the workload. Citrix demonstrated moving a workload from Intel’s Cloud Builders data centre in California, which Intel treats as a private cloud, into the Hillsboro, Oregon, facility, which acts as the public cloud.
In addition to those data centres, Intel has small labs in Belgium, India and China that are also used to test reference designs. The sites have 150 servers combined and can simulate 5,000 virtual machines and more than 15,000 desktops, said Nikhil Sharma, director of cloud solutions for Intel’s Cloud Builders Factory. Intel also makes about 10 terabytes of storage available for partners to provision servers and test applications.
Intel invited some partners this week to show off a few products for the cloud. Parallels showed how its software helps cloud providers design and offer packages of services to small and medium-size businesses. Oxygen Cloud demonstrated a service that lets end-users share and access corporate documents securely from devices such as smartphones and tablets. HyTrust showed how its appliance enforces policies in a data centre so that administrators don’t accidentally move workloads where they shouldn’t go.
It’s not unusual for large vendors to promote their products using reference architectures. But the variety of products and services on display highlights the wide net Intel has cast to help grow the overall cloud market and retain its dominant share. While Intel leads the x86 server market by far, it has competition “looming in the distance,” Brookwood said.
The industry is shifting to a model in which an increasing amount of computing is done in the cloud and delivered to handheld devices, which are powered predominantly by ARM-based processors. Historically, more computation was done on the client, most often a PC or notebook running an Intel chip, he said.
In addition, some cloud providers are looking at alternative, lower-power ARM chips for certain server workloads. “We will see more and more servers based on ARM because of the power savings users believe they can get,” Brookwood said.
Intel sought to downplay that threat. “There’s been some interest in low-power but it’s been a bit overblown,” said Dylan Larson, director of technology initiatives in Intel’s data center group. He acknowledged that some workloads might benefit from the efficiency of certain very low-power chips.
“We won’t be asleep at the wheel,” he said, implying Intel will act to address the changing demands.
Some vendors already have. Just this week, SeaMicro introduced the second generation of its servers that use 256 low-power, dual-core Intel Atom processors, which Intel intended for use in netbooks. While Intel wasn’t keen to support SeaMicro’s efforts initially, it has changed its tune since the first server was received fairly well by the market, Brookwood said. “Instead of ‘You’re nuts,’ Intel is now saying ‘What can we do to help you,'” he said.
But other companies are working on similar concepts using ARM-based processors.
“Because of Intel’s corporate culture of ‘only the paranoid survive,’ it’s worrying about where the little guys are coming from who can eat up Intel,” Brookwood said.