April 30, 2014
On April 23, IBM rolled out the long-awaited POWER8 CPU, the successor to POWER7+, and given the extensive pre-announcement speculation, the hardware itself was no big surprise (the details are fascinating, but not suitable for this venue), offering an estimated 30 – 50% improvement in application performance over the latest POWER7+, with potential for order of magnitude improvements with selected big data and analytics workloads. While the technology is interesting, we are pretty numb to the “bigger, better, faster” messaging that inevitably accompanies new hardware announcements, and the real impact of this announcement lies in its utility for current AIX users and IBM’s increased focus on Linux and its support of the OpenPOWER initiative.
OK, so we’re numb, but it’s still interesting. POWER8 is an entirely new processor generation implemented in 22 nm CMOS (the same geometry as Intel’s high-end CPUs). The processor features up to 12 cores, each with up to 8 threads, and a focus on not only throughput but high performance per thread and per core for low-thread-count applications. Added to the mix is up to 1 TB of memory per socket, massive PCIe 3 I/O connectivity and Coherent Accelerator Processor Interface (CAPI), IBM’s technology to deliver memory-controller-based access for accelerators and flash memory in POWER systems. CAPI figures prominently in IBM’s positioning of POWER as the ultimate analytics engine, with the announcement profiling the performance of a configuration using 40 TB of CAPI-attached flash for huge in-memory analytics at a fraction of the cost of a non-CAPI configuration.[i]
A Slam-dunk for AIX users and a new play for Linux
The net of this technology for current AIX (and the remaining I-Series) users is a solid bump in performance and price-performance. Adding in nice little fillips such as a very attractive Capacity On Demand (COD) program and improved POWER-VM capabilities, there is nothing not to like in this announcement. If you are an AIX customer and committed to continuing with AIX, there is no reason to hesitate before upgrading – POWER8 will give you years of additional life, along with a roadmap that promises incremental improvements over the next several years while you wait for POWER9.
But the real change comes in the way IBM is targeting the entire Linux ecosystem. Assuming the Lenovo deal goes through, IBM, now minus its own x86 servers, will be even more aggressively targeting Linux as a primary workload for POWER, pushing advantages in price-performance, especially in applications like (surprise) analytics and big data that stress I/O and memory bandwidth, performance per core, and throughput per socket.[ii] At announcement, IBM had an impressive roster of Linux distro partners, including SUSE, Red Hat and Ubuntu, which collectively account for at least 90% of the commercial Linux market. With claims of “nearly 100%” compatibility for interpretive languages (Java, PHP, Python, etc.) and easy recompile of C/C++, IBM is making an aggressive play for Linux workloads. The question of course is whether they will they seriously challenge Intel’s volume hegemony. I doubt that they will ever account for more than a few percent of the potential market for new Linux, and even less in terms of migration of existing footprints, but even 3 – 5% of Intel’s Linux footprint would be a huge volume win for IBM compared to current POWER volumes.
A recent high-profile endorsement of POWER as a Linux platform is the interest from Google in possibly using POWER in their data centers. This week Google showed a prototype server board built around POWER8 at IBM’s Impact conference. If Google actually goes into production with significant volumes of POWER servers, it is unclear if they will purchase POWER processors from IBM or buy a (possibly customized) variant from an OpenPOWER licensee (see below). Of course, this could just be a form of hardball negotiation with Intel – time will tell. But if Google blazes the trail, others will follow.
Perhaps the most challenging aspect of the announcement to parse is IBM’s amped up support for the OpenPOWER initiative, which makes the POWER architecture available as Open Source IP. A comprehensive discussion of the potential ramifications will require multiple posts, but in a nutshell this initiative creates and open and unencumbered alternative to x86 and ARM for multiple purposes, ranging from high-performance enterprise servers to very power-efficient web-scale servers and embedded processors[iii].
Interest in OpenPOWER appears to be growing, especially in APAC, and I am confident that it will have an impact. What I am less certain about, however, is how well IBM can monetize this initiative. Claims that the collective investment and interest in the ecosystem will somehow validate POWER as a mainstream Linux processor are possibly true – what is uncertain is whether IBM can make money on this validation. If POWER succeeds in gaining traction, it will be because an ecosystem of system vendors supplying power systems, most likely Asian ODMs selling lower cost Linux systems, will have developed. Against the background of competitive system alternatives, many of them locally branded in growing Asian markets, IBM will have trouble selling outside of its current base of established enterprise customers. So as insurance and against being marginalized in the perception of growing markets, OpenPOWER is an excellent vehicle. But I remain skeptical about its real monetary value to IBM.
If you are an IBM AIX user, run, don’t walk to this upgrade when you are ready. It offers years of life extension to current systems and will be a solid basis for continued investment in AIX.
For Linux workloads, you now have an additional, and very aggressive, vendor to add to the competitive mix. I would encourage aggressively competing POWER8 Linux against x86 Linux for new workloads as long as you are ot one of the corner cases where all the code is not available on POWER8.
[i] While I tend to take any vendor-supplied cost and performance comparisons with a ton of salt, I think this architecture will indeed deliver substantial performance improvements compared to a conventional distributed memory analytics infrastructure. The long term question is how will other non-POWER alternatives such as UltraDIMM-based and IBM’s own eXFlash compare.
[ii] Sadly IBM has followed a long and depressing vendor tradition of not comparing themselves to the latest and greatest of competitors kit in their benchmark comparisons. I would encourage potential users to ignore comparisons to Intel’s Xeon E5, and focus instead on comparisons against the newest E7 V2, which is a much more appropriate compare to POWER for performance benchmarks.
[iii] One company, Servergy, has based its business plan on power-efficeint POWER-based servers for cale-out web web environments. There are additional companies in the OpenPOWER ecosystem working on server and derivative POWER8 CPU designs, including Tyan, Inspur, ZTE and Chuanghe Telco Tech, and Sozhou PowerCore.