|
| Personal background |
| computers, electronics, |
| Thoughts about SETI and SETI@home |
I am in charge of a number of computers at work. Like most computers, they spend virtually all of their time idle. Many of these computers are kept running 24 hours a day because they keep some connected imaging equipment busy and/or they serve files onto the LAN. Every computer is connected to the LAN, in order to allow outbound sharing of the acquired image collections and inbound sharing of drivers, updates and other data. The LAN automatically gives every computer Internet access via a shared T-1 line.
I install SETI@home on all of "my" computers at work and my personal computers at home. SETI@home makes use of the otherwise idle computing power and contributes it to a shared project.
SEI@home takes a little effort to install and keep up to date. But there is a tangible benefit. SETI@home keeps the computers under continuous test. If a computer latches up or loses LAN connectivity or Internet connectivity, it is more noticeable with SETI@home running. I often notice a problem and correct it sooner than I would without SETI@home.
I don't leave computers powered on just to run SETI@home. There are plenty of computers that can run the search. I usually hibernate my office computer and home computers. SETI soaks up the idle computing power, while the computers have some other reason to be powered on.
I never run the SETI@home screen saver. It wastes a lot of computing to make a display that I don't need. Also, I find the SETI@home screen saver unattractive. The new SETI@home screen saver is even less attractive than the old SETI@home screen saver. I choose a minimal screen saver such as "Mystify". The screen saver only runs for 15 minutes, because then the Power Saver kicks in and turns off the display.
SETI is a big search. I don't have a strong opinion on what it should or should not find.
=====
room for improvement on SETI:
I notice on Windows/Intel computers, the SETI@home task is constantly allocating and de-allocating memory. SETI@home behaved this way, from the earliest days until now (setiathome_4.18). Allocating and de-allocating memroy can be costly and inefficient. I don't see why the SETI@home task cannot use a fixed amount of memory and re-use any odd segments.
I wonder if the WIntel version of SETI@home uses the Intel Math Library / Intel Performance Primitives, to maximize the throughput of SETI@home on the many Pentium computers running it. (I notice that Intel is no longer a sponsor.)
I notice that SETI@home central computers give out each Work Unit to 4 computers. However, only 3 computers are needed to make a "quorum" -- a confirmed result. Many times I find that one of my computers has work units in progress or waiting to start up, for which the central SETI@home computers have already received 3 results. It seems clear that computing a fourth result is a complete waste. When a computer is behind on results, I often abort Work Units that already have 3 results, to allow the needed results to be computed. It would be nice if the client software could occasionally look into the results files and automatically cancel any work units for which 3 results have already been returned. It seems foolish to waste 25% of the donated processing power to compute duplicate results in excess of the required number.
BOINC automatically updates the SETI@home software needed to compute the results. However, BOINC does not know how to update itself. The SETI@home central computers have actually stopped issuing Work Units to my computers because the BOINC software was out of date. It makes a significant amount of unnecessary work, to have to go and update BOINC on each computer manually.
BOINC with SETI@home generally never interferes with any other software running on the computers. There are some exceptions:
(1) SETI is not supposed to run when the computers are not idle. I see no sign that SETI suspends itself. It stays runnig and holds onto a significant amount of memory.
(2) On computers that have Symantec Antivirus, Corporate Edition (tested with v 9 or 10), some kind of scan runs after logon that keeps the "winlogon" task busy for about 1.5 minutes. The scan tries to adapt, and suspends itself if some other task is running. If BOINC is running, the scan never moves forward, it stands still. Until the scan task completes, all actions on the computer run sluggishly. I have a fix in place. I do not let BOINC run at startup, instead I created a startup task that waits 180 seconds before launching BOINC. the system runs much better. One important detail, the old SETI@home (the one without BOINC) DID NOT have this problem.
The SETI@home screen saver uses a lot of computing power AND is unattractive. Worse yet (if true), it might not stop computing after the Power Saver has kicked in and turned off the monitor. Computing power is wasted computing a graphic that is not even displayed. A smart screen saver stops computing after the Power Saver has kicked in.
SETI@home and BOINC seem to provide no place to report on objections and deficiencies or to offer feedback. Very impersonal. So I feel obliged to post them here. |
| Your feedback on this profile |
| Recommend this profile for User of the Day: | Man šis aprašymas patinka |
| Alert administrators to an offensive profile: | Man šis aprašymas nepatinka |
|
|
