An anniversary and a memorial
Today I got my 25 Year Service award from CMU. That's a long time, and even though it has nothing to do with the Baronial Election or the death of Bin Laden, I feel moved to comment.
It's actually been more like 26 years, and that corresponds exactly to the lifetime of the Pittsburgh Supercomputing Center. The Center just had its own 25 year celebration, marking the anniversary of the day the grant actually came through. There are a surprising number of people from the early days still working at the PSC, which I think demonstrates that it really is a fairly nice place to work. Since I was the first employee I was asked to write a few comments about the Center's history; an edited version was included in our web page. For posterity, the full text is below, behind a cut.
One of the least pleasant things about working at the PSC is also one of the most predictable and repetitive- periodically we have to come up with proposals for new supercomputers. These are high stress situations; Centers have died for failure to win a proposal, and the PSC was very nearly among the fallen a few years ago. I fortunately am responsible for only tiny bits of proposal, but even so I get pretty stressy. The folks who do the real work get a lot more stress. A couple of months ago, we were part of a team that together and submitted Yet Another Proposal. I'm pretty much a workaholic, so I take a secret pleasure in these high-pressure times. And I'm not alone.
Which brings me to an old friend and boss of mine, Phil Andrews. At the time we were all working on our proposal up here, Phil was down at the National Institute for Computational Sciences working on their proposal, under circumstances which were apparently pretty stressful. NICS is the place that won the supercomputing grant that I ranted about in the notorious blog post that got me mentioned in the NY Times. (Four years later I hope everyone will be calm enough that I can safely un-hide it). Phil was hired as project director for NICS. He did a great job, and yes, NICS continues to be a productive supercomputing resource for the NSF, my rant notwithstanding.
So Phil was working on this proposal late into the night. He was planning a move back out to California to rejoin his family, but he never got to make it- eventually he went to sleep, and somewhere in the night he had a heart attack and died.
I feel very bad for Phil. He died a long way from his family, trying to finish just one more chunk of work. I feel very bad because it's taken me so long to write this, but of course, every evening I've been trying to finish just one more chunk of work. Phil was only a couple of years older than me; it's a reminder that I should slow down a bit.
Anyway, I've been thinking about this post since Phil died. He was a remarkable, impressive, very nice guy, and I am one of the many people who will miss him.
1. What brought you to the PSC?
What brought me to the PSC was the perfect synergy of luck and
laziness. I'd done a Ph.D. in physics at Pitt, taking at least one
class from Ralph in the process. I'd post-doc'd for a while at Pitt,
and spent a few months at the Max Planck Institute for Astrophysics
using their Cray 1S, which was a machine sort of like an iPhone but
with no screen or audio. After that I had to decide between doing
more post-docs while I waited for some tenured general relativist to
retire or changing course. At that precise moment Mike and Ralph
founded the PSC. I went and interviewed with them with the notion
that I'd be an 'on-site physicist'. This was the model of 'staff
scientists' that NCSA was adopting, but it wasn't the way the PSC
ended up working. Fortunately I had good computing skills, and I
picked up software skills as things went along.
2. How has PSC changed or impacted HPC?
I guess another way to ask the question is, 'what has the PSC done that
the other Centers have not?'. We've obviously supported a lot of
unique research. Our support, outreach and teaching programs are
perhaps the best among the Centers. For the past few years I've been
closest to the performance-analysis side, and I do believe that we get
more deeply into the hardware of the machines we propose and build than
the other Centers do. The network trickery we proposed for the Track
2C machine is an example of this; the engineering work we did for the
Track 1 machine was another very impressive case. I'm not sure what I
can say we've brought to HPC as a whole, but within the Centers program
I think we've had less of a 'leave it to the vendors' attitude than the
others.
Oh, one other point would be our habit of pushing for highly parallel
jobs. This has always made more work for us, because it's easier to
come up with a lot of small jobs than one really scalable one. I
think we've benefited the whole community by pushing that boundary,
though.
3. What is your most compelling (funny is ok) memory over the last 25
years?
Oh geez, I have to give you several. I remember the very beginning of
the Center, when we were all sitting around in a big room full of
desks over at Mellon Institute. That was a very, very compelling
time- I think the camaraderie that we built during that time is the
reason so many of us have stuck around for so long. We all sat at
desks and answered each other's phones; I was the User Services
department and Doug Fox was the hardware group. I happened to take a
phone call from a user asking about graphics software, and that earned
me more graphics questions, and eventually I became a Famous Graphics
Guru.
I remember working on the Track 1 proposal. We work well under
stress, but that was too much stress for too many people. That wasn't
a good time.
I remember the NSF cutting us off at the beginning of the Grid era,
and then being annoyed with us for a while because we failed to die.
I also remember the Town Meeting where it was announced that we'd won
the proposal for the TCS machine, with Bob Stock standing up and
shouting "We're Number One!". I think coming through that period
intact was probably the biggest triumph the PSC has had.
It's actually been more like 26 years, and that corresponds exactly to the lifetime of the Pittsburgh Supercomputing Center. The Center just had its own 25 year celebration, marking the anniversary of the day the grant actually came through. There are a surprising number of people from the early days still working at the PSC, which I think demonstrates that it really is a fairly nice place to work. Since I was the first employee I was asked to write a few comments about the Center's history; an edited version was included in our web page. For posterity, the full text is below, behind a cut.
One of the least pleasant things about working at the PSC is also one of the most predictable and repetitive- periodically we have to come up with proposals for new supercomputers. These are high stress situations; Centers have died for failure to win a proposal, and the PSC was very nearly among the fallen a few years ago. I fortunately am responsible for only tiny bits of proposal, but even so I get pretty stressy. The folks who do the real work get a lot more stress. A couple of months ago, we were part of a team that together and submitted Yet Another Proposal. I'm pretty much a workaholic, so I take a secret pleasure in these high-pressure times. And I'm not alone.
Which brings me to an old friend and boss of mine, Phil Andrews. At the time we were all working on our proposal up here, Phil was down at the National Institute for Computational Sciences working on their proposal, under circumstances which were apparently pretty stressful. NICS is the place that won the supercomputing grant that I ranted about in the notorious blog post that got me mentioned in the NY Times. (Four years later I hope everyone will be calm enough that I can safely un-hide it). Phil was hired as project director for NICS. He did a great job, and yes, NICS continues to be a productive supercomputing resource for the NSF, my rant notwithstanding.
So Phil was working on this proposal late into the night. He was planning a move back out to California to rejoin his family, but he never got to make it- eventually he went to sleep, and somewhere in the night he had a heart attack and died.
I feel very bad for Phil. He died a long way from his family, trying to finish just one more chunk of work. I feel very bad because it's taken me so long to write this, but of course, every evening I've been trying to finish just one more chunk of work. Phil was only a couple of years older than me; it's a reminder that I should slow down a bit.
Anyway, I've been thinking about this post since Phil died. He was a remarkable, impressive, very nice guy, and I am one of the many people who will miss him.
1. What brought you to the PSC?
What brought me to the PSC was the perfect synergy of luck and
laziness. I'd done a Ph.D. in physics at Pitt, taking at least one
class from Ralph in the process. I'd post-doc'd for a while at Pitt,
and spent a few months at the Max Planck Institute for Astrophysics
using their Cray 1S, which was a machine sort of like an iPhone but
with no screen or audio. After that I had to decide between doing
more post-docs while I waited for some tenured general relativist to
retire or changing course. At that precise moment Mike and Ralph
founded the PSC. I went and interviewed with them with the notion
that I'd be an 'on-site physicist'. This was the model of 'staff
scientists' that NCSA was adopting, but it wasn't the way the PSC
ended up working. Fortunately I had good computing skills, and I
picked up software skills as things went along.
2. How has PSC changed or impacted HPC?
I guess another way to ask the question is, 'what has the PSC done that
the other Centers have not?'. We've obviously supported a lot of
unique research. Our support, outreach and teaching programs are
perhaps the best among the Centers. For the past few years I've been
closest to the performance-analysis side, and I do believe that we get
more deeply into the hardware of the machines we propose and build than
the other Centers do. The network trickery we proposed for the Track
2C machine is an example of this; the engineering work we did for the
Track 1 machine was another very impressive case. I'm not sure what I
can say we've brought to HPC as a whole, but within the Centers program
I think we've had less of a 'leave it to the vendors' attitude than the
others.
Oh, one other point would be our habit of pushing for highly parallel
jobs. This has always made more work for us, because it's easier to
come up with a lot of small jobs than one really scalable one. I
think we've benefited the whole community by pushing that boundary,
though.
3. What is your most compelling (funny is ok) memory over the last 25
years?
Oh geez, I have to give you several. I remember the very beginning of
the Center, when we were all sitting around in a big room full of
desks over at Mellon Institute. That was a very, very compelling
time- I think the camaraderie that we built during that time is the
reason so many of us have stuck around for so long. We all sat at
desks and answered each other's phones; I was the User Services
department and Doug Fox was the hardware group. I happened to take a
phone call from a user asking about graphics software, and that earned
me more graphics questions, and eventually I became a Famous Graphics
Guru.
I remember working on the Track 1 proposal. We work well under
stress, but that was too much stress for too many people. That wasn't
a good time.
I remember the NSF cutting us off at the beginning of the Grid era,
and then being annoyed with us for a while because we failed to die.
I also remember the Town Meeting where it was announced that we'd won
the proposal for the TCS machine, with Bob Stock standing up and
shouting "We're Number One!". I think coming through that period
intact was probably the biggest triumph the PSC has had.