Does High Performance Computing (HPC) need a “Killer App”?
Recently I read two interesting articles. One was relatively an old article from Intel CTO – 3D Web is going to be the killer application for HPC. Another was a Microsoft article – HPC as the killer application for Microsoft Azure. Both these articles have two things in common. One was HPC, and the other was killer application.
Intel’s article claimed that HPC has a niche market, but it had not percolated to the masses. HPC market can’t survive without a mass puller or a “killer application”. 3D Web was one such application that can give life to HPC. Microsoft article pointed out that it was serious about HPC business, and plans to use its HPC platform over Azure for cloud offering. As it owns both HPC platform (MS HPC Server) and Cloud platform (Azure), it strongly believes both can complement each other in their product portfolios.
Being part of our HPC Lab, these articles were an eye-opener. What did they mean by killer application? Did HPC not have any killer applications before? How was it different from other applications of HPC? What do they mean by killer applications?
These articles forced me to recollect many of my past experiences. To just give a background, I was representing my previous company – a leading mobile manufacturer in the international mobile forum. Many times we went with smart features, at few times with smarter features, and at times with smartest features. However we were being bombarded with questions from mobile operators and the big brothers of the industry!
- What is the burning need for me to accept your features?
- What does it compel me in my life that I should opt your feature?
- Give me a compelling use case? Not just a use case alone!
- Demonstrate me how big is the market for your features?
Often we got amused. In spite of showing many use cases, we were sent back with these standard repetitive questions. These questions were often irritating us! Frustrating us!
Later during my coffee chats, I was educated by my forum mate, about what they mean by “killer application”. A mere use case will just talk about an application for a technology. But the investors don’t want mere use cases, something which would be considered as “nice to have”. The market driven industries always extend their hands to something that “drives or pull the market”. We need features that drive the market or create a new market. The killer use case can be justified as irrevocable marriage between a technology and its application something can’t be resisted.
Let us take a fine example of a killer application – email over web (e.g.: Hotmail, Yahoo Mail). Now let’s ask a question to ourselves. Who has benefited out of this combination, is it email as an application, or web technology. This triggers my thoughts back to the original topic i.e. Killer Application for HPC!
Later I visualized about a 3D Web use case for HPC as suggested by Intel – Social Gaming, a car racing simulation game over conference room, just like how Microsoft Xbox offers interactive games over network. One can map this use case with public chat rooms like Bangalore Chat rooms, Chennai Chat rooms in Yahoo Chat service, where many internet users assemble to enjoy the benefits of social communication. One or more public users join a game conference, choose the car, chooses colour, getting acquainted with Game rules, and the game (of complexity) starts. As more users join, more interactive controls over car increases, more real time complex graphics are included, constant streaming of video and audio streaming throughout the session (not isolated streaming), virtual worlds, different car view point for each users, different video and audio streams, impact observed by two users when two or more cars collide with other, accurate simulations, on-demand scaling as more users join, augmented reality, more similar chat rooms running in parallel, multiple instances of the same chat rooms from all over the world, etc., Let me take a pause here! Enough of complexities!
Having visualized the amount of scaling of objects, complexity in processing graphics, kind of real-time communication, immersive visuals, and complexity involved, I wondered to myself, just like how the big brother asked me on the other day. Does this use case address mass market? How big is this market? By the time, I was asking this question to myself, my eyes were already navigating across the paragraphs in that article as follows:
“The rapid growth of virtual worlds is particularly indicative of the potential for Immersive Computing Experience. In June 2008, the research firm Strategy Analytics predicted that the percentage of registered users that will go on to become active users in virtual worlds will increase from 10 percent now to 27 percent by 2017, creating a market base of almost 1 billion users. This is up from today’s 137 million active users. Pre-teens make up more than half the users, spending 16 hours per week (more than on TV or magazines) and purchasing items in the virtual worlds. Many convert their parents into virtual world users. Even more interesting will be what happens if many of these pre-teens carry their virtual world habits into adulthood. Strategy Analytics puts the 2008 total market for user services at $310 million. The research group predicts that it will grow to $8 billion in 2017”.
I finished reading these statistics, and I was curious to know how well the industry prepared for this leap growth? Industries have been working on Web 3D standards; collaboration with W3C is on recently. Google and Mozilla are equipping their browsers with WebGL to render 3D Web. NVIDIA has come with a Reality Server with interactive 3D applications. Similar initiatives are being carried out by Mobile vendors. Already there have been some talks on Web 3.0 using 3D & virtual worlds.
Now comes final set of questions, how can I achieve these use cases without a right technology? Can these use cases fly without HPC? Do these use cases mandate the requirement for HPC platform? Yes it is clear from now what HPC industry shall do for its own survival!
References:
- Intel CEO article: http://cacm.acm.org/news/52347-3d-web-will-save-hpc-industry-intel-cto-says/fulltext
- Microsoft Article: http://rcpmag.com/articles/2010/09/21/microsoft-hilf-sees-hpc-as-the-killer-azure-app.aspx