IBM Smarter Computing – ‘Big Data In Optimized Systems, Managed As A Cloud’

IBM Smarter Computing Highlights

• Launched at Pulse2011 in March 2011
• Encompasses numerous optimized and integrated server and storage products
• Watson example – beats Jeopardy! champions, demonstrating powerful analytics and deterministic programming
• UPMC Health sector example – Smart Beds, consolidation and cost savings
• STG’s Böblingen lab focuses on cross platform technologies
• IBM believes the definition of ‘system’ is getting wider

We dislike the use of comparative adjectives in campaign and product names: it’s not just that its hard to resist moving from ‘smart’ to ‘smarter’, to ‘smartest’, but someone’s ‘2.0’ implies there are others left behind as ‘1.Xs’. Clearly IBM hasn’t been listening as it infuses its offerings with ‘Smarter’ nomenclature.
In March 2011 it was STG’s turn, launching its Smarter Computing campaign, which is wide, deep and potentially complex. We’ll try to unravel and explain it here.
IBM invited 36 key European IT analysts to its Böblingen research lab to talk about its new approaches. We were outnumbered by our hosts who were mainly from its Systems and Technology Group (STG), mostly from its North West IOT and included many doctorate level engineers.

The meeting was opened by Simon Edward, Marketing Director and lead by Gururaj Rao, IBM Fellow and Chief Systems Engineer, who refers to himself as ‘Guru’ and as we will do in this paper.
For us Smarter Computing is particularly relevant to IBM’s workload optimisation and smart analytics approaches. We will address how it is looking at infrastructure and Cloud Computing in a later research paper.

IBM Böblingen – A Cross Platform IBM Research Laboratory

IBM STG’s Böblingen lab is the third we have visited after seeing its activities at The Watson Research Center in Yorktown, New York State and in La Gaude near Zurich in Switzerland. Böblingen has a cross platform focus, like Beijing, but unlike the specialist labs at Poughkeepsie (System z) and Austin (Power).
Böblingen was founded in 1953 with 7 employees: in 2011 its has 2k and is involved with around 60 hardware and software projects. In particular:

• It designs servers and their chips, as well as server firmware
• It works on Linux on mainframe – I can’t believe it’s really been 10 years since we visited to see its original demo
• It works on virtualization, the z/VSE operating system, workload and storage management
• It also runs customer Proof of Concepts (POCs)

As an area Böblingen is steeped in the computer industry and I’ve visited rival HP’s offices there many times. Stuttgart is the nearest city, which is the home of Mercedes of course.

What Is IBM Smarter Computing?

Guru described Smarter Computing as ‘big data running on optimized systems, managed as a Cloud’. As a campaign it was launched at the Pulse! Industry show in the first week of March, for which internal preparations have been made since October. We see these as evolutionary developments, probably because we spend a lot of time unravelling STG’s approach. For those who have not, they make come as a revelation.
IBM understands that the server market hinges around workloads and infrastructure. It sees the definition of ‘system’ itself stretching far wider than a single machine or even data centre. At the heart of this campaign is the positioning of STG offerings in four different categories, which are:

• General Purpose Systems
• System Optimizers
• Pre-Integrated Solutions
• Integrated Heterogeneous Systems

A picture of where the major STG products fall is shown in Figure 1.
Fewer IBM customers are purchasing general-purpose machines and building their own solutions, while more are buying pre-integrated solutions and workload optimized systems.
Guru indicated that lowering the cost of workloads was the aim of IBM’s more automated and vertically integrated approach. He made a number of claims. For instance:

• The new System z196 and URM deliver cost savings of up to 55% per workload
• Workloads on Power Systems with DB2 are up to 66% cheaper than the equivalent Oracle versions
• Management costs for eX5 System x machines are 50% less than before

We believe that IBM is one of the only server suppliers which have fully researched the needs of its users for pre-integration. We believe it is also one the most likely candidates to avoid the charge of customer lock-in, as it builds its solutions on open standards and is knows the pitfalls of Anti-Trust legislation very well from the late 1970s.

Jeopardy! Champion Watson Illustrates Smarter Computing

Readers may remember Deep Blue which beat world chess champion Garry Kasparov in 1997. Most recently it has built a systems to emulate a contestant in the American game show Jeopardy!
The novelty of the show is the normal position of the question and answer is reversed, although this usually means little more than having to say ‘what is…’ or ‘who was…’ before answering.
In January Watson challenged ex-champions Ken Jenning and Brad Rutter over three days and won, donating money to charity in the process. This is an illustration of what the company is aiming to achieve with Smarter Computing.
Watson uses a large Power 750 System at its heart housed in 10 racks and delivering 10 TeraFLOPS performance. Each processor within this system can process 32 threads simultaneously, as they have 8 cores per processor and 4 threads per core (please read our earlier description of server design). The speed and parallelism of the system were a sine qua non for answering the quiz show in the allotted short time slots.
The software used for Watson was Linux, Apache UIMA (unstructured information management architecture) and IBM’s DeepQA programme.
Of course Watson is an example of a smart analytics workload optimised system. It is a deterministic system, using semantics to break the question into its various clauses and words before searching and correlating sources such as Wikipedia for matches. It produces a number of candidate answers and the statistical probability of each being correct. Whether or not it puts forward its own answer to the game is dependent on how well it is doing, since it doesn’t need to bet on a 50:50 confidence level if it is already far ahead of its rivals.
Although the application appears relatively trivial, it is a good example of how IBM is trying to demonstrate its deep expertise through challenging a well-known ‘institution’. It makes us think about what other problems could be addressed by these new types of system. As a fan of similar UK quiz shows I’d like to suggest IBM builds ‘Magnus’ for taking on Master Mind and ‘Bamber’ for University Challenge!

UPMC Health Sector Reference – Consolidation And Smarter Beds

Another example of Smarter Computing comes from the Health sector. The University of Pittsburgh Medical Center is a large hospital with annual revenues of around $8 billion, total staff of 50k including 400 teams of doctors. As almost always with consolidation projects, the story starts with a new CIO who managed massive cost savings in the process. In this case the number of Unix servers was cut from 74 to 14 and of x86 servers from 1,300 to 22.
The hospital transformed the way it used IT, creating ‘Smarter Beds’ for inpatients which have become the hub for the exchange of clinical information between doctors, nurses and administrators. Monitoring of the patient’s vital signs are also integrated – clearly these have to be 100% resilient systems as a result. The tie-in stretches to co-ordinating diets from the kitchen to avoid any food allergies the patient may have.
The UPMC reported significant savings, including the dropping of a proposed $80 million investment in a new data centre. It now appears to have plenty of rack space its pre-existing site for new servers, storage and networking.
If Watson demonstrated the deep computing ability of IBM systems, UPMC shows the equally important benefits of consolidation and innovation. The extent to which it is now saving money will no doubt be of interest to thousands of Private sector health units, who need to ‘spend to save’ as part of austerity budgeting.

Some Conclusions – Where Does IBM’s Vision Leads Us?

During IBM’s conference we came across a number of interesting statements, such as:

• Servers are where applications live
• The system becomes the infrastructure
• The IT function moves from operations to service provision

Smarter Computing for us should help STG commercialise the ‘workload’ part of the house, making the most of its links with its two services (GTS and GBS) and software divisions. STG is by no means as skinny as Apple in terms of the number of SKUs it carries and the more appliances and integration of software, hardware and services it puts together in its products, the more the number of grows. We believe it needs to think hard about how it presents its portfolio, although defining its major headings in four groups (see above) is a good start.
IBM’s workload approach makes perfect sense to those of us who have been brought up studying the server market. It will make the competitive landscape interesting as not all vendors are tuned in – Cisco, for instance is perhaps too bust integrating servers into its networking to think too deeply about which applications its customers are running.
We believe that users need open architectures built on open standards before they invest in simplified solutions. The test of the current spate of vertically integrated solutions from system vendors will be the amount of pain suffered when they want to change supplier. IBM has demonstrated its ability to do deep computing with Watson and save money through consolidation for UPMC: if it can protect users from lock-in it will have a good chance of returning to the top of the server market on an annual basis – after all IBM isn’t just for Xmas!