The problem is the software - Jack Dongarra
June 24, 2006
At the Para 2006 conference in Umea, Sweden, Jack Dongarra from the University of Tennessee and Oak Ridge National Labs gave a “state of the union” talk on parallel computing which I found rather interesting. For those who don’t know, Jack is a leading light in numerical algorithms and parallel computing, the focus of the conference, and is also involved with the TOP500 project. Among topics he touched on
- Multicore CPUs, which he described as a disruptive technology which will change the way programs are written and how easily computing will become available
- Hybrid systems such as cell processors and FPGAs
- A need to rethink the software paradigm, especially in light of new processor types. He even talked about a need for new programming languages
- Heating, which he described as a major problem with all the new efforts to achieve even faster computing speeds by 2010 (speeds that will make the current champ look like a walk in the park
He also spent some time talking about grids and clusters. One of the problems with grids according to him is that the idea got oversold. In his mind there are valuable applications of grids, but they are hardly a panacea and cluster computing might be a better choice, although the latter can have hidden costs.
Jack’s recurring theme was that the current state of hardware was fine, with the challenges lying at the software end.
Scientific computing could use a shot in the arm. Today, there is a lot of academic research in parallel computing/numerical computing, but the software industry is not really involved, except for software coming out of FPGA manufacturers and other hardware companies.
Technorati Tags: Scientific Computing, High Performance Computing, Parallel Computing, Multicore CPUs, Hardware, Software, Computing, Conference
Chris Anderson on the “Wisdom of Crowds”
June 22, 2006
I found this via Nascent. Read all about it at Nature.com. Chris also blogs about the article
I respect Chris Anderson a lot. His concept of the Long Tail is something that I have come to appreciate and embrace. His insight into what constitutes scientific peer review makes for some interesting reading. Chris certainly knows all about the “Wisdom of Crowds”. First of all I like his distinction between pre-filtering and post-filtering. I am still not convinced what constitutes a peer. In more specialized areas, papers might get buried underneath the avalanche of papers on global warming, nanomachines and evolution. There has to be a way to normalize ratings and not just making things a popularity contest. That publishers such as Nature, PLoS experimenting with alternate approaches should be lauded, even if we disagree with the model, because it is at least a start. I have a feeling that things will work out in the not too distant future.
Further Reading:
Nature takes the lead
Chris Anderson
The Omics World
Technorati Tags: Peer Review, Science, Publication, Journals, Nature Magazine, Chris Anderson, Wisdom of Crowds, Long Tail
Microsoft releases its cluster solution
June 9, 2006
Bioinform is reporting that Microsoft has announced the release of Windows CCS, its first real HPC solution. The GA date for Windows CCS is August 2006. For someone who has always been into HPC and Linux clusters, my first reaction to Windows CCS when it was first announced was a combination of “about damn time” and “how bad is it going to be”. Certainly, the usability of Windows CCS (and it does seem usable from the admin standpoint) makes it a viable option for someone intent on getting into small parallel applications, perhaps in a small company or at the departmental level, but Microsoft has a long way to go before the modeling and bioinformatics community accepts Microsoft as a viable provider of HPC solutions. I will say this, Microsoft is not exactly trying to blow Linux out of the water. They are coming into this very carefully, and focussing on smaller applications, and (I think) targeting IT managers, many of whom are comfortable with Windows and to Windows developers who can now parallelize their applications and work in the CCS environment.
In a world where Linux and Windows seem to be the most viable HPC options (I am not counting specialized hardware like Blue Gene), it is important that Microsoft is providing an alternative to Linux HPC. I don’t think this is going to really impact Linux in the _near_ term as much as it will allow people who have stayed away from parallel computing to adopt it. As the BioInform article says, some people just want to get their research done
. Long term. We will have to wait and see. It is always good having options. There any applications that I know off that have Windows and Linux support and offer MPI support only on Linux (I used to manage such products). It would be nice if people with a preference for Windows got some of the same capabilities as their Linux compatriots, and there are always enough people in the modeling and informatics world, who are a little intimidated by Linux and would like to work in a Windows environment (Mac too, but those people are usually very comfortable with Linux clusters)
Further Reading
Scalability.org - A view that shares some of my biases towards Linux
Technorati Tags: Scientific Computing, High Performance Computing, MPI, Cluster Computing, Microsoft, CCS, Windows, Linux
How “I am a scientist” became history
June 9, 2006
Genome Technology’s latest issue has some career counselling advice in the form of a Q&A, and a couple of the questions is about scientists and MBAs. Does a scientist need an MBA to pursue a career in a business track? In my opinion, the answer is usually no.
First of all, some luck is required. When I left my first job, I had the fortune of moving into a product marketing position that required a lot of technical domain knowledge (and luckily a domain right up my street). That is arguably the most comfortable way of making the transition. When I started, I focused a lot on the scientific aspects of my job, but as time went on, my focus became a healthy balance of the marketing, sales and product development aspects. Some product managers stay closer to their scientific roots, others welcome the challenge of forecasting and product lifecycle management. The latter are the ones who are likely to have a successful business career.
Read as much as you can. I devour as much information as time permits and then some. I have friends who are management consultants and have picked their brains, subscribe to Harvard Business Review and Business Week, and have read books on everything from finance to strategy. Make sure that you actually apply some of the things you read. After all most PhD’s are smart people and if you have the aptitude for the business aspects, you will learn a lot from reading, talking to people and experience. I have noticed that the more analytically inclined scientists often tend to make the transition easily
I am not sure coursework helps. I have attended some short courses, and I don’t think I have really learnt anything more what I might have reading blogs, articles and books, but if you can afford it (or your company provides some training), it doesn’t hurt.
If you want to leave the industry that you are trained in, e.g. if I wanted to move out of the life science business into, e.g. automotive manufacturing, then an MBA probably helps. One of the things that helps a PhD successful in a business environment is that it isn’t always easy for a pure business professional to get a grasp of the subtleties of the market or evaluate market trends without being too dependent on advisors. And early on in your career there are not too many advisors around. If you are a seasoned professional, then I think the argument breaks down in both directions. A seasoned PhD with several years of sales and marketing experience is likely to be as successful in a job in an unrelated industry as an equally competent MBA. Likewise, I think a non-scientist MBA can be successful in a senior position in a scientific company. Obviously, I am biased. I definitely think it is easier for a PhD to learn business than it is for an MBA to learn science, but that’s a bad generalization.
Get your hands dirty as much as you can. Visiting customers and being a part of the sales process is essential. If you do not understand your customers view point from a business perspective, it will be difficult to take the scientific hat off. The beauty of what I do now is that I get to think about business problems within the context of the science and technology that either gives rise to those problems or that can help alleviate them. To me that is the best of both worlds.
In the end it is about attitude, aptitude and what your goals are. Going from research to business purely for monetary reasons will often fail, since it is naive to imagine that you can do something just because you think it’s a good idea. It takes a different type of personality and those that succeed are the ones who are willing to make mistakes, learn from them and think proactively. And yes, and you have to be able to make tough decisions that might tear your scientific heart out.
Technorati Tags: ProductMarketing, Science, Business, CareerChange, Careers
Peer Review: Nature takes the lead
June 6, 2006
Timo Hannay has a post on Nascent about a debate on Peer Review that has been launched by Nature. Congratulations to Nature for taking the lead on this. Many of us in science have had our frustrations with the peer review process, and the fact remains that science and the way it is practised and disseminated has changed a lot. Science is best reviewed by experts in the field, but the process by which science is reported and judged needs a shot in the arm. Science has a tendency of becoming “clique-y”, both in the peer review process as well as the grant process (that’s another story which I will leave to my academic colleagues). The articles in Nature discuss some of the core issues and presents some posssible solutions
The web is a truly disruptive medium as evidenced by the impact that PLoS and Pubmed Central have had. It allows open debate on a number of topics, so why not science? What we need is a process that keeps an open mind, and encourages scientific discussion. I am not sure that the procedure that Nature is testing out will necessarily work, but I certainly like the idea. I have never really trusted the Wisdom of crowds, but I could be wrong. After all, the qualification filter here is quite high and that could be an internal quality control.
Further Reading
Tim O’ Reilly
The Omics World
Nodal Point on PLoS One
Technorati Tags: Nature Magazine, Peer Review, Science, Web 2.0
