IBM Rounds Out Its Linux Offerings With Power Linux

In the latest evolution of its Linux push, IBM has added to its non-x86 Linux server line with the introduction of new dedicated Power 7 rack and blade servers that only run Linux. “Hah”, you say!
Written by Richard Fichera, Contributor

In the latest evolution of its Linux push, IBM has added to its non-x86 Linux server line with the introduction of new dedicated Power 7 rack and blade servers that only run Linux. “Hah”, you say! “Power already runs Linux, and quite well according to IBM”. This is indeed true, but when you look at the price/performance of Linux on standard Power, the picture is not quite as advantageous, with the higher cost of Power servers compared to x86 servers offsetting much if not all of the performance advantage.

Enter the new Flex System p24L (Linux) Compute Node blade for the new PureFlex system and the IBM PowerLinuxTM 7R2 rack server. Both are dedicated Linux-only systems with 2 Power 7 6/8 core, 4 threads/core processors, and are shipped with unlimited licenses for IBM’s PowerVM hypervisor. Most importantly, these systems, in exchange for the limitation that they will run only Linux, are priced competitively with similarly configured x86 systems from major competitors, and IBM is betting on the  improvement in performance, shown by IBM-supplied benchmarks, to overcome any resistance to running Linux on a non-x86 system. Note that this is a different proposition than Linux running on an IFL in a zSeries, since the mainframe is usually not the entry for the customer – IBM typically sells to customers with existing mainframe, whereas with Power Linux they will also be attempting to sell to net new customers as well as established accounts.

So, other than to be able to say that Linux runs on every server platform, what does IBM expect to accomplish with this offering? IBM contends that Linux on Power has already exceeded their expectations because of the inherent strengths of the Power architecture, and that this offering will be exceptionally competitive, with performance from 60% to well over 100% performance advantages versus similar x86 systems. IBM is initially targeting workloads in three segments – big data and analytics, ISV applications such as SAP, and open source infrastructure services such as PHP, Apache, MySQL, etc. Aside from raw performance, the ability to run more VMs and mixed workloads (IBM’s highest competitive performance claims are for multiple VMs running multiple workloads, far in excess of 100% improvement) will enable software such as SAP to be deployed using a simpler physical infrastructure.

Will IBM succeed with this initiative? Hard to predict the exact shape of things to come, but with the investments they are making in the ecosystem, development tools and partner enablement, it is a serious program, and the price-performance advantages are significant enough to at least force a serious evaluation of this alternative by many Linux users. My guess, and not a particularly difficult conclusion to come to, is that this will a successful program in current IBM accounts, and will also meet with modest success outside of their installed base. Power will not seriously perturb forecasts for x86 Linux, which are probably in the order of a couple of million systems this year, but even if they account for 2% of the market that would represent a $500 million or larger line of hardware revenue at a minimum and will certainly drag a multiple fo that value along in service and software revenues, resulting in a nice tidy billion-dollar or larger incremental franchise for IBM.

If this program shows signs of developing along the lines I outlined above, it is almost a forgone conclusion that they will expand this to larger Power systems, and offer the equivalent of an IFL for Power – a dedicated hardware module that will only run Linus at a preferential price, assuming that they can convince the mainframe crowd that this won’t imperil IFL on Z sales.

All in all, a smart move by IBM, and customers with high-performance Linux requirements would do well to evaluate this platform option.

Editorial standards