IT development organizations needs to heed Moore’s law
by
Charlie Bess
Now that cooling has become a major problem for computing, the focus for the last few years has been on adding more cores rather than increasing clock speed and adding more cache. At the current rate of doubling the number of cores every year, we should be at 30 cores per desktop by the end of the decade. In the data centers, it this will lead to dynamic grouping of cores (if the current concept of a data center survives). This nebulous grouping of processing to perform a function is why the term cloud is so appropriate.
I was talking with a person from the technology office at a large software provider and they mentioned the internal struggle concerning taking advantage of all the parallel capabilities that exist. They have a significant portion of their organization working on the problem. We'll soon be to the point where every desktop is a little cloud of computational capabilities. Writing software for a multi-processor environment will be the norm and not limited to high-performance computing problems.
Today, writing high-performance parallel software using 3rd generation languages is pretty tough. Moving developers to more model-based approaches that assemble code from existing components will be a pretty hard nut to crack as well. I don't even want to go into the licensing issues for packaged software. We don't have a choice though.
In any case, support organizations need to begin planning now, because making this turn will require significant time and effort. Taking existing code with you may not be much of an option, although SOA should help. I hate to even think out 10 years.
Identity management, security and resource utilization are key hurtles that need to be addressed in this vision, as well as the inertia associated with hardware ownership and control. These areas have not fully come to grips with the cloud computing requirements. Today, there is a great deal of effort focused on virtualization, rightfully so, since it is a tactic that will free up funds to address more strategic issues. We can't confuse the tactic with what we're likely to need to accompany the technology on its journey.
