[hcs-d] Seat Computers at the Cool Table

James Cuff james_cuff at harvard.edu
Sat Dec 4 18:35:09 EST 2010


Hi all,

Fascinating article!  I as did Greg and Harry each noticed this one :)
 Are there particular courses already that cover more advanced topics
such as web scaling, massively parallel file systems, MPI, compiler
optimization and HPC in general?  I look to current successful
companies and architectures - where graduates would be getting jobs in
the future.  Each are huge scale out systems (we do some of this here
on campus, albeit on a much smaller scale).

Azure, GFS, Hadoop, Amazon EC2 systems and heck even facebook's neat
tricks to make php scale (non trivial) are important as we continue to
ask more and more complex social and scientific questions.  Along side
that, some of the operational challenges that I'm faced with on a day
to day basis of power and cooling technologies to make sure such large
scale systems can actually work.  Even physical building systems can
no longer cope with such massive scale.  The big boys are having to
engineer way beyond just the tin and silicon these days:

http://www.channelregister.co.uk/2010/09/23/yahoo_compute_coop/
http://www.google.com/corporate/green/datacenters/

I know Hanspeter runs a GPGPU course, and we in RC help out with a
bunch of courses that are in a way related to "using HPC", certainly
in statistics, economics, phylogenetics and some molecular biology
courses.  No courses that I know of actually explain the deep inner
workings of HPC and how to get towards solid algorithmic scaling that
challenge my team and scientists across the globe daily.  I could be
wrong, I'm not all that up on current course offerings.

Maybe I'll take up Adam's suggestion and reach out to some of the
computer science faculty to see if there is interest, at the very
least I would be able to do a guest spot - just need to work out which
existing CS platform would be a good fit.

Happy to hear any suggestions from the student body if this would be
interesting, feel free to contact me, and I can summarize to the list.

Yes - computers are certainly not just for nerds any more - that is for sure :)

Best,

j.

--
dr. james cuff, director of research computing & chief technology architect
harvard university | faculty of arts and sciences | division of science
rm 210, thirty eight oxford street, cambridge. ma. 02138
tel: +1 617 384 7647 | www: http://rc.fas.harvard.edu





On Sat, Dec 4, 2010 at 5:09 PM, Adam Gold <agold at fas.harvard.edu> wrote:
> Hey everyone!
>
> Thanks so much for reading my article. It seems like you all got a lot
> out of it.
>
> I submitted a request to my editor for a correction on the HCS name and
> the managed languages thing yesterday. Hopefully they'll get around to
> it soon. My column usually runs on the other Friday, but due to a last
> minute schedule change, they needed to run my evergreen piece that I
> wrote sophomore year (gasp!) back when the world was young and CS51 used
> to use Scheme. I tried to update what I could, but I didn't have enough
> time to catch everything on such short notice.
>
> My point about Windows/MacOS versus Unix was really more about
> development tools. You can't make a Windows or an iPhone app using pure
> C and gcc, which probably comes as a shock to most people who just come
> out of CS50.
>
> Just because MacOS uses a Unix kernel doesn't mean that if you already
> know how to write a Linux app, you don't need to learn anything new
> about making a Mac application in the real world. I didn't want to go
> into too much detail because I'm writing for a general audience (i.e.
> the Crimson readership).
>
> For those of you wondering what the "next step was," I think I spelled
> it out pretty clearly when I said that "More funding should be allocated
> towards speakers, seminars and programs that educate Harvard students
> about computers and reinforce the notion that it’s not geeky to
> understand them." I know it was pretty close to the end of the article,
> and the piece was a whopping 800 words, so I can see why some of you
> maybe were too busy to read all the way to the end before forming an
> opinion.
>
> If anyone else is thinking of checking out MIT's 6.839 (Adv. Computer
> Graphics), because Gortler's on leave this year, you should totally do
> it! Given how little I know about computers, I'm sure I'll bring down
> the curve for everyone on this list!
>
> xoxo,
> Adam
>
>
> On 12/4/2010 3:33 PM, Jim Danz wrote:
>> John,
>>
>>> Of course, the name itself is pretty obviously plastered all over
>>> official HCS material, and it's incumbent on the Crimson to get it
>>> right.
>> +1
>>
>> I greatly enjoyed your take on this, so thanks for writing that up.  I
>> do think it's worth noting that many groups/activities don't have such
>> a choice to make.  For instance, "consulting" is the name of an
>> activity but also the name of an industry; "Harvard (College)
>> Consulting Group" has got to be the only reasonable choice.
>>
>> For what it's worth (and to be safe, with this I'm only speaking for
>> me personally), I think that part of the reason for the naming as we
>> do it is that "computing," in terms of connotation, generally seems to
>> describe using software (ie, "personal computing") rather than
>> developing software | configuring machines | building machines.  But
>> yeah, it's all just words.
>> _______________________________________________
>> hcs-discuss mailing list
>> hcs-discuss at lists.hcs.harvard.edu
>> https://lists.hcs.harvard.edu/mailman/listinfo/hcs-discuss
>
> _______________________________________________
> hcs-discuss mailing list
> hcs-discuss at lists.hcs.harvard.edu
> https://lists.hcs.harvard.edu/mailman/listinfo/hcs-discuss
>


More information about the hcs-discuss mailing list