Features:

INTERVIEW WITH BOEING HPC SERVICE MANAGER SHUKLA
By Alan Beck, Editor-in-Chief

HPCwire's Editor-in-Chief Alan Beck recently interviewed Suresh Shukla, HPC Service Manager of the Boeing Company. This was pursuant to Suresh’s presentations in US and Europe on CART, the decision making model he proposed for the HPC industry. Interview on CART was published in HPCwire issue of 09.05.03.

The views noted in this interview are Suresh Shukla's own and do not necessarily represent views of others in The Boeing Company.

Benefit Quantification and Tracking (BQAT)

Q: You gave presentations during last year in the US and Europe on the strategic decision-making model CART. How were the presentations received? Was there any common thread you saw in your interactions with leaders in the HPC industry?

A: The concepts in CART were received very well. The HPC leaders expressed many reactions; but I perceived one common thread. Most of them stated that there was not enough information being gathered and tracked about benefits of HPC to justify expansion of it within their organizations. This was the biggest obstacle for deploying HPC for furthering new and innovative designs. They felt that ultimately this could affect their competitiveness.

Q: Why is this situation more severe today than in the past?

A: Previously, some high-level executives believed that they “must” have the latest and greatest HPC equipment in their organization, and used to make gutsy decisions to acquire it with only a general rationale. Today, such gutsy decisions (e.g., the creation of the Earth Simulator) are few and far between, especially in the industry sector. It appears that concrete benefits need to be identified and tracked if the potential of HPC is to be realized today within any organization.

Q: Benefits need to be addressed in developing any business case. What is so different about HPC?

A: No single justification applies to the acquisition and retention of all types of computing equipment. For this purpose, we could classify computing equipment into three types. The first type is old and obsolete equipment that the organization wants to get rid of. The second type is equipment used for known business processes, such as manufacturing or payroll. This type of equipment is upgraded or downgraded because of ups and downs in the business cycle. HPC, though, normally falls into the third category. The computing equipment in this category is used to innovate and revolutionize current product designs or to address competitive new product designs. One can build a business case around the “least-cost” option for the first two categories, but not for the third category. The root cause of the current problem in HPC is that we have become accustomed to developing “least-cost” business cases for HPC equipment, just as for the other two equipment types. Yet the HPC rationale ought to be developed around quantifiable benefits, not cost reduction. An organization that doesn’t develop its HPC business case around quantifiable benefits will become non-competitive in deploying HPC.

Every organization needs to be aware of this problem and work hard to address it.

Q: What are some ways to address the root cause of the problem? Which one needs to be deployed today in HPC environment?

A: One way is to use “gut feel” without enough quantification and tracking of benefits. This is the predominant method today. The problem is that it often compromises product design and innovation for the organization deploying the HPC equipment. Another way is to expand HPC capabilities commensurate with your competitors. But by definition, this method doesn’t get you ahead of the competition. The third way is to rigorously quantify benefits and track them. This process may take one to three years or more; but it would be worth the effort. The process may involve tracking benefits based upon past history, current product design requirements, your ability to deploy the latest application technology for simulation, and currently available HPC options. Estimating benefits in this way may lead to “white space”, meaning excess HPC capacity, for a period of time; but white space is what precisely gives your users headroom to grow, as opposed to being restricted by “least-cost” HPC options that will not accommodate serious innovation.

Q: Can you give an example of the benefits of having “white space”?

A: Let’s say you determine you can save $7 million by converting a portion of experimentation to HPC simulation. The simulation will take six months and will require $1 million in investment in HPC equipment that has a three-year useful lifetime. That leaves you with two-and-a-half years of white space, or extra capability, after the simulation is completed. I would say that this presents a doubly strong business case for the non-least cost, $1 million HPC purchase, because that white space can be deployed for another two-and-a-half years beyond the simulation to create additional innovative designs. Purchasing cycles from an outside vendor just for six months may need to be looked into as an alternative, but it will not provide extra capability for additional innovative designs.

Q: It is difficult to quantify benefits even after the design is complete, let alone before starting it. It is even more difficult to apportion the contribution of benefits due to the deployment of HPC towards the successful completion of the design.

A: Yes, it is difficult. If it was simple, all organizations would have routinely performed it by now; but time has come when we cannot use this high level of difficulty as an excuse not to quantify the benefits of HPC to the organization. The alternative would lead to non-competitiveness.

There is a degree of risk involved in quantifying benefits. Engineers and scientists, who have to play a major role in assuring benefits of HPC, are trained to go from one level of certainty to another. This makes them risk- averse. They need to be assisted by management who are used to address and contain risks. It needs to be a team effort between management, sales, marketing, engineering, finance, and procurement organizations bringing all their individual skills to the table. Over a period of three to five years, such a group can track the benefits from history, and come up with dependable projections.

It may not be necessary to predict benefits very exactly. Large variances should be permissible. In the example above, it would be OK if actual benefits derived exceed $1 million, and are below, say, $14 million.

Q: How do you define this process? How can organizations develop this process?

A: I call this process BQAT (pronounced as “bee-quat”), which stands for Benefit Quantification and Tracking. Organizations need to seriously address many questions while developing this process. Some of them are:

Who identifies the benefits, and how? Who quantifies the benefits, and how? Who approves the quantified benefits? Who tracks that the benefits are achieved, and how? Who establishes the past history and future trends? Who improves the integrity of the process, and how?

Other questions include: How to quantify benefits when faced with an untested environment and product? How to establish permitted variances from standard benefits? How to evaluate the probability of success in the business case? Who should make purchasing decisions around these scenarios?

Q: You raised lots of questions. Where are the answers?

A: Answers will vary based upon the industry segment. The business case has to be created by each organization for itself. It may take as much as three to five years for an organization to develop processes addressing BQAT that everyone can buy into. I cannot overstate significance of non-least-cost approaches such as BQAT to the competitiveness of organizations that deploy HPC as an enabling tool for innovation, rather than simply to increase productivity on known engineering and science. The organizations that deploy BQAT are going to be well ahead in the game more often than not.

I plan to discuss this topic at HPC User Forum in Dearborn, MI, in the next meeting which would take place from April 12th through 14th. People interested in discussing this subject further can contact me at sureshnshukla@aol.com.