Editor’s Note: Today we launch the first blog post of many that will make up the Smarter Computing Breakthrough series. These posts, written by many of our Distinguished Engineers and IBM Fellows, will help introduce you to the key technological breakthroughs that are the unsung heroes of the IT infrastructure that enables a Smarter Planet.
Middleware-Optimized Systems: Every Layer Is Designed for Every Other
Business systems today face exceptionally demanding workloads and must nevertheless hit performance and availability goals. That is the essence of the IT conundrum—and a place deep, multi-layer optimization can really make a difference.
To see how, just ponder how hardware and software interact in a modern server. It’s one thing simply to ensure compatibility — that the software will run reliably on the hardware. It’s quite another to ensure optimization — that the software will run much faster than it otherwise would, and thus deliver truly best-in-class performance.
Optimization of this type is very difficult to achieve using commodity hardware, commodity processors and off-the-shelf software because they were never designed for each other in the first place. The software developers and the hardware developers worked in isolation from each other, usually at completely separate companies.
In the case of IBM systems, exactly the opposite is true. IBM is committed to deep, full-stack performance engineering for its systems, in which its software and hardware teams collaborate in the pursuit of holistic performance excellence exactly because that collaboration will translate into significant, quantifiable business gains for our clients and customers.
That means optimization of both IBM’s software stack (including IBM’s operating systems and middleware) as well as optimization of IBM’s processors and other hardware elements on which that stack will ultimately run. All of our groups are aware of what the others are doing, and have full access to the technical specifications they need to deliver the best solutions for the system as a whole.
The business benefits of deep optimization
Let’s consider how that applies to middleware, such as database environments and Java application servers. Both these forms of middleware are essential to IT today and routinely support critical business workloads. Optimizations in these areas will definitely yield impressive performance gains for real-world service levels.
Databases and application servers also share another trait: while they are certainly specific technologies, they are utilized by businesses in innumerable ways. So any deep optimizations IBM can make in these areas are bound to yield not just impressive business benefits, but also many different kinds of business benefits.
How to optimize? One answer: take advantage of the new strengths of IBM’s POWER7 processor, a substantial step up from the previous generation. Where POWER6 had two cores per chip, POWER7 has eight. Where POWER6 supported a maximum of four threads per chip, POWER7 supports a whopping 32 threads per chip. These design improvements mean that one excellent way for software developers to deliver higher performance of their software, when it’s running on POWER7, is to exploit those cores and threading (as well as similar improvements in caching and memory bandwidth).
That, in short, has been the goal of IBM’s Project Mason — an ongoing initiative to optimize IBM software to take ideal advantage of IBM’s own processors on IBM POWER Systems. Project Mason’s scope is nothing less than the total IBM software stack, including not just the operating system (both AIX and Linux — did you know IBM employs 600 Linux developers per year?), but also IBM middleware like DB2 (our industry-leading database offering) and WebSphere (our application server).
These two solutions make up a versatile foundation for business services of all kinds, in all industries, and they have now been extensively optimized for the POWER7 architecture. Because those optimizations are available right out of the box, IBM clients are typically spared the need to struggle with slow and costly code optimizations of their own. Instead, they’ll see enhanced workload execution immediately, manifested in an endless number of ways.
Faster, more available and more scalable than ever
Want an example? In the case of WebSphere Application Server (WAS), Project Mason optimizations included a shorter code path, superior Java Virtual Machine (JVM) performance, advanced caching, and, of course, the fact that POWER 7 processors support up to 32 threads for a single process.
What does that mean in a business context? It means far more business transactions can be processed in a given period of time – not just more than POWER 6, but more than the competition.
Compared to a similarly-equipped Intel Nehalem-driven server, fewer JVMs are required, because each JVM running on POWER 7 delivers 73% more performance than its equivalent running on Intel. As a result, the IBM solution supports up to 3920 transactions per second, where the Intel solution only supports 2260.
Scalability, in fact, is now almost linear — as you increase the number of POWER 7 cores, WebSphere Application Server’s performance will scale in direct proportion. And business services remain responsive and available, even at times of unpredictable workload spikes.
DB2 has experienced similarly impressive gains. In this case, IBM focused heavily not on abstract benchmarks but real-world business performance, by configuring its evaluation setup to mirror a typical customer environment as closely as possible. The results of Project Mason? DB2 9.7 Enterprise Server, when running on AIX (IBM’s flavor of UNIX), now performs up to 19 times faster — all with an out-of-the-box deployment, requiring no special recoding or hardware modifications.
And IBM’s full-stack optimization ambitions don’t stop there. Another more recent example: IBM analytics solutions. On POWER7, ILOG CPLEX now shows up to a 32 percent performance improvement. InfoSphere Warehouse? 40 percent to 8x improvement. And SPSS now shows 2x to 10x improvements across different algorithms and workloads.
One of the great strengths of IBM systems in recent years lies in the way we’ve been able to optimize multiple layers of those systems to benefit each other — and, of course, ultimately benefit our customers.
Because IBM is a true complete systems provider — a single organization that creates much of the hardware and software that comprise its systems — IBM is also exceptionally well positioned to ensure that its hardware and software are well tuned for each other.
And IBM customers will find those optimizations make their IT services run faster, stay available more of the time and, ultimately, contribute more to the business bottom line.
How are YOU transforming your IT efforts for efficiency and more impact? Let us know! Leave a comment on the Smarter Computing blog below or connect with us on Facebook, Google+ or Twitter. If you tweet, be sure to include the #TransformITnow hashtag.
To effectively compete in today’s changing world, it is essential that companies leverage innovative technology to differentiate from competitors. Learn how you can do that and more in the Smarter Computing Analyst Paper from Hurwitz and Associates.