ボストン大学の最新スパコン Regatta。
Bigger, and better?
I attended a workshop this morning to learn how to use the latest and greatest supercomputer to arrive at Boston University, the IBM p690, a.k.a the Regatta. The "students" at the workshop were mostly postdocs and professors interested in using the supercomputer for research.
I've been a long-time user of the supercomputers at Boston University, so there was no way I was going to miss out on this event, since I knew that I would be able to sign up as a "friendly user."
The Regatta has just been installed at Boston University, so the university wants to test how well it performs before it opens the computer to the entire community. There may be glitches, so the "friendly users" must be patient as they help the computer settle into production mode. In exchange, the "friendly users" can use the computer as much as they like during this short period of testing.
Because only a limited number of people can be "friendly users," fewer people use the supercomputer at any given time during the trial period. This means that more time is spent computing, and less time is spent waiting for other jobs to finish.
The supercomputer is connected to standard computers via the Internet. I usually work at a computer in my lab, but I can even control my jobs from home. When lots of people send their jobs to the supercomputer, the jobs must wait in a queue. Even after a job starts running, it gets kicked off after a certain number of hours, at which point the job must wait in the queue again. Waiting is an inevitable part of using a supercomputer, but with fewer users, the wait is shorter.
Another benefit of being a "friendly user" is that we can use the computers as much as we want to. Usually, we must write "budget" proposals each year asking the administrators to allocate each proposal a certain amount of computer time. A review board allocates the hours depending on how worthy they deem a project. I've never been denied any computer time that I've asked for, but it's still nice to know that there are no limits on the time I can use the computer.
Even multimillion-dollar supercomputers don't have infinite computing capacity. As the number of users increase, and the computer programs become more complex, a supercomputer becomes more and more obsolete. Thus, the supercomputer must be upgraded regularly.
Boston University has consistently upgraded its supercomputers during my six years here as a graduate student. The Regatta is the fourth supercomputer I've used. The instructors at the workshop seemed to target an audience of similarly experienced supercomputer users. However, I noticed that there were quite a few people who had vague ideas of using the supercomputer but had no clue of where to begin. The instructors suggested that they take some basic, online tutorials, but watching the students, I suspected that many of them will never learn to use the supercomputers.
It's not so hard to learn how to use the latest upgrade if you have used the previous version. It becomes increasingly difficult to jump in midstream, but it's still easier than trying to learn how to use the earlier, simpler supercomputers.
I've found that I don't need to wait in the queue as long as I used to. This may be partly because the supercomputers have greater capacity now than they did five years ago, but I suspect that fewer people have been deciding to learn how to use the supercomputers.
Many former supercomputer users, frustrated with the long queues, have switched to PCs. A good desktop PC can perform as well as a supercomputer of not-so-long ago but costs only a fraction as much and doesn't have to be shared. Supercomputers take a long time to develop and require a team of people to install and maintain them.
My research group relies heavily on computer simulations, but only a handful of my labmates use the supercomputers. This is fine with me, since it gives me more time to run my simulations. I won't be here for much longer, but I intend to squeeze out every minute I can from this new Regatta. I assure the university that I will be a VERY friendly user.
Shukan ST: June 21, 2002
(C) All rights reserved
- latest
- 最新の
- a.k.a
- (=also known as)別名
- postdocs
- 博士課程修了の研究者
- there was no way I was going to miss out on 〜
- 〜を見逃す手はなかった
- sign up as 〜
- 〜として登録する
- has just been installed
- 設置されたばかりだ
- performs
- 機能する
- entire 〜
- 全〜
- community
- 大学のこと
- glitches
- 故障
- settle into production mode
- きちんと作業できる状態になる
- In exchange
- 代わりに
- at any given time during 〜
- 〜の間ずっと
- computing
- コンピューターを使う
- jobs
- ジョブ、作業
- is connected to 〜
- 〜とつながっている
- via 〜
- 〜経由で
- lab
- (=laboratory)研究室
- queue
- 待ち行列
- gets kicked off
- はじかれる
- a certain number of hours
- 一定の時間
- inevitable
- 避けられない
- benefit
- 利点
- "budget" proposals
- (使用時間の)「割当」計画書
- administrators
- 管理者
- allocate 〜
- 〜を割り当てる
- review board
- 審査委員会
- depending on 〜
- 〜によって
- how worthy 〜 deem 〜
- 〜が〜どれだけを価値のあるものとみなしているか
- (have)never been denied 〜
- 〜を拒否されたことはない
- multimillion-dollar
- 数百万ドルの
- infinite
- 無限の
- capacity
- 能力
- complex
- 複雑な
- obsolete
- 廃れた
- Thus
- よって
- be upgraded
- アップグレードされる
- regularly
- 定期的に
- consistently
- 絶えず
- graduate student
- 大学院生
- instructors
- 教官
- target
- を対象にする
- audience
- 受講者
- similarly experienced 〜
- 同じような経験を持つ〜
- vague ideas
- 漠然とした考え
- clue
- 手がかり
- online tutorials
- オンライン上での取扱説明
- previous version
- 前の型
- increasingly
- ますます
- jump in midstream
- 途中から飛び込む
- as long as I used to
- 以前ほど長く
- frustrated with 〜
- 〜にいらいらして
- have switched to PCs
- (=personal computers)パソコンに乗り換えた
- 〜 of not-so-long ago
- そんなに古くない〜
- costs only a fraction as much
- 格段に安い
- be shared
- 譲り合って使う
- develop
- 開発する
- maintain
- 整備する
- relies heavily on 〜
- 〜にだいぶ頼っている
- a handful of 〜
- 一握りの〜
- labmates
- 研究仲間
- intend to 〜
- 〜するつもりだ
- squeeze out
- 絞り出す
- assure
- 保証する
|