These are the old pages from the weblog as they were published at Cornell. Visit www.allthingsdistributed.com for up-to-date entries.

January 22, 2004

I am Evil

Once again I am being accused of being in bed with the Evil Empire, and to manipulate my research results to further the goals of Microsoft. It appears this is never going to stop. This time the discussion is going on at the the ServerSide.com in response to this message by Cameron Purdy in which he hints that the cli-grande benchmarks are not to be trusted, because my department apparently received $60M from the Evil Empire. I wish we had. The old press release he is pointing at is about hardware, network equipment and licenses for one of the Theory Center's clusters. Sigh. We have received funding from many corporations over the years, including all the major Java platform vendors, and actually Microsoft turns out to be one of the companies that poses the least restriction on the use of funding.

I couldn't resist to respond in public, as I had already commented to Cameron about this earlier on accusations (scroll to the "Microsoft Marketing" posting) he had made on his weblog. Yet he continues to use this reasoning to discrete good, impartial research. You can find my response here at the serverside, or below together with my original comments, in the extended version of the post.

My response to the "this is what I call slipping money to the priest .." article

This reply is mainly for Cameron Purdy, but I am doing this publicly so that others may get a better insight in the issues also.

First of all, with respect to the benchmarks: 1) you can download them from http://cli-grande.sscli.net You can run them on your own machine and determine for yourself how the different runtimes stack up. I am not hiding anything, so you can see for yourself. 2) The benchmarks are aimed at the HPC community, and do not test any enterprise computing scenarios. They tell you something about specific, pure computational scenarios, nothing more or less.

But I mainly want to respond to Cameron's suggestion that my academic judgment is not to be trusted because of the fact that I use Microsoft technologies as my platform instead of platforms developed by IBM or Sun. I rebutted this statement by Cameron before when he made it on his weblog (http://www.jroller.com/page/cpurdy/20030722). Unfortunately that page appears no longer available, nor is my rebuttal in teh comment section.

First the story about the $60M & the MS case study: 1) As you can see from the press release this is about the Cornell Theory Center, and has nothing to do with me or my department. 2) the $60M was not in money, but in hardware, networking equipment and software licenses, and not just by MS but also Dell and Intel. 3) The TC used this to build supercomputing facilities to support computational biology, etc.

(Cameron, you aware of that there is no connection between this funding and me because I wrote this in the comments on your weblog, yet you continue to use this argument to discredit technical reports)

Secondly about funding from industry for academic research: I have received corporate funding for many, many years as my research focuses on issues that many corporations appear to see as very relevant. This includes IBM, Sun, Intel, Microsoft, Cisco, Siemens, BBN, AT&T, and others. I also received significant funding from the US government. I can/will only accept funding from these corporations if they do not influence in any way the direction my research is taking.

Third, about MS as a platform for research: I changed from a Unix Guru to a Windows Guru in the mid nineties when Sun decided to pull the plug on my favorite platform; SunOs. The source licenses for Solaris were too restrictive for academic use, so we couldn't use it. We decided to see whether NT, which came with usable academice licenses to the source code, would make a good replacement platform, and stuck with as it turned out to be just as good a platform as any other and sometimes even better. But it turned out that the biggest hurdles too take in research on a windows platform are not technical, but social. Some people (Cameron?) will use the connection to Microsoft to discredit the technical results, regardles of how general applicable they are, I assume based on some religious-style aversion for anything Microsoft.

I normally wouldn't respond to these issues, but for an academic independence is the core of her/his credibility. I challenge Cameron to come up with proof that my academic results are flawed or skewed towards MS, as he hints at. Or that my results are skewed towards Java because I have also received funding from IBM and Sun.

My response to the Microsoft Marketing 101 article:

Hi Cameron, thanks for the nice words. I am always interested in hearing how to run benchmarks better. BTW I am not an anti-java person, I have done my fair share of java coding, and I believe that for any type of problem that is out there you should pick the best tools for it. If that is Java, all the better, if it is the JVM, good, I really have no problem with it. If it is something else I don't care either as I don't own stock or anything like that :-). My interest with respect to VM's is more in whether they in general suffer from the same limitations, from a performance point of view. Whether at system/runtime level there is are differences in architecture that translate into some limitations. This is not about languages anymore, but about language runtimes.

WRT my .Net focus, well I have learned long time ago that if you want to do something very well, you have to limit yourself and focus on a smaller set of problems. Well given that most of my colleagues are looking at Java technologies, I thought that maybe there was value in looking at similar technologies but from a different view.

The $60M for the CTC, weren't really cash dollars. The majority of this comes in 'resources': hardware, software licenses, consulting resources, network equipment, etc. And with an Itanium cluster, these dollars are easily spent. The CTC - Microsoft/Intel/Dell collaboration started earlier, no long after that NSF decided that all but 2 HPC centers in the US should close down. For some time it seemed that the CTC would really close shop built then a deal was struck with MS, Dell, Intel and GigaNet to build an Intel based cluster, and take it from there. The CTC folks themselves are not researchers , they provide an HPC service to the Cornell community (such as the material sciences folks, protein folding research, or the computational finances folks). This way they could continue to operate, so the choice was simple.

Remember that before this deal, they has a large SP2 installation, which came at significant discount from IBM. Lots of 'cases studies' were done on that cluster also, but nobody has unleashed any 'conspiracy theories' on those. If Sun, Oracle, HP, Cisco, SGI or Intel give you money, it is called research funding. If Microsoft does the same, it is called 'buying research'. I have received funds from all of these companies over time, and you will have to take my word for it, that Microsoft's grants come with the least strings attached.

Posted by Werner Vogels at January 22, 2004 10:48 AM
TrackBacks

Comments

This is one thing that I love and hate about the technology business: passionate people. There's a certain set of topics that you just don't bring up for debate because they bring out the irrational, emotional side of folks.

In my CS undergrad the attitude was "Intel is good. Sun is good. Microsoft is evil."

The simple fact is that Microsoft did more for the computer industry than anyone will admit to. Also, good ideas come from all over. Java is a good solution, .NET is a good solution. Pick the right solution for the problem and get on with it.

You & Cornell are being unfairly maligned by Cameron, but arguing with him is just pointless since he's being completely irrational about it. I'd never seen TSS degrade into the mob mentality that you normally find on /.

Posted by: Craig Pfeifer on January 26, 2004 08:54 AM