What is the future of enterprise processing, and will the x86 platform have a role in it?
The short answers are “very promising,” and “yes,” but the real interesting possibilities only emerge when you delve into the long answers.
The catalyst for this discussion is ARM Holding’s development of a new networking architecture for its low-power processors. ARM has made no secret of the fact that it wants to see its design (remember, ARM doesn’t actually make the chips, it only designs them) infiltrate the enterprise server farm. The company has repeatedly touted its low-power advantages, which make it such an attractive option for mobile devices, as the key to lowering data center power consumption and operating costs.
The problem is that an ARM-based server architecture will only be as effective as current x86 systems if you pack a lot of cores into it. And for that you need rapid communication and coordination between them. With the new CoreLink CCN-504 cache system and DMC-520 memory controller, ARM says it can up the number of cores in a typical configuration to 16, and then better manage the parallel workloads that multicore environments are supposed to excel at.
So in a nutshell, ARM is not looking to displace x86es in the enterprise simply by increasing the performance level in the traditional fashion. Rather, it wants to create an entirely new server platform — ideally one more suited to the cloud-based, collaborative workflows that are quickly gaining ground in enterprise development circles. As the company’s Ian Ferguson pointed out to Chip Design magazine, with software coming to dominate network infrastructure and higher-level programming languages like Java and C++ gaining ground, the long-standing ties to x86 are starting to weaken. What better time for a radically new way of looking at the way data is processed?
For its part, Intel is still confident of x86’s future in the enterprise, even as it cautiously opens the door to new server options for its low-power Atom processor. Ironically, one of the top Atom designers is SeaMicro, which, despite being acquired by AMD earlier this year, still offers systems like the SM15000 with either Xeon or Atom processors. Expect to see more Atom machines in the coming year as Intel rolls out the new generation of quad-core “Bay Trail” SoCs, rumored to sport such server-friendly attributes as 1333 MHz DDR3 memory support, built-in virtualization and 64-bit processing.
Intel is also targeting the Atom at the storage market, coming out with solutions based on the D2500 and D2550 processors, designed mainly for low-power support for personal cloud services to businesses and consumers. The idea is to provide readily available storage that can be accessed remotely from multiple devices, while at the same time supporting collaboration and file sharing through various data transfer protocols. The system even features built-in hardware acceleration and up to 4 GB of main memory, allowing it to be connected to HDMI-compatible video displays and other high I/O devices.
So, to get back to our original question, will x86 continue to play a role in the enterprise, the answer remains yes because the installed bases of hardware, middleware and applications that have grown up around x86 aren’t going anywhere soon. But the fact remains that there is more than one way to process data and applications, and an entirely new architecture that is arguably more in tune with the collaborative data environment that is emerging around us is ready to go.
x86 will continue to play a key role in this world, but not necessarily the dominant one.