SETI@HOME has obviously grown so large as to be unmanageable in its present form. To fix this impasse, I offer the following suggestion:

--------------------

Clone the systems as follows:

1) Make an identical copy (or two or three) and independent installation(s) of SETI@HOME, each containing all the elements necessary for splitting, download and upload, etc. They could be named SETI@HOME_A, SETI@HOME_B, etc.

2) Each copy would process a chosen portion of the available data and then contribute the results to a common science/results database.

3) Any user would be eligible to subscribe to one or more of the sites at their choice.

4) Redistribution of the existing equipment may be possible thus avoiding acquisition of any more.

--------------------

The benefit is that if the two or more sites were indeed independent of one another, the likelihood of at least one being available at any one moment is near certain (except in the present (10/15/2007) situation when A/C is out or if there were a significant and lengthy area-wide power outage).

The user benefits by having a continuing stream of work units to process and avoids becoming frustrated due to many frequent and lengthy outages.

The project benefits by being able to take one site down for maintenance or upgrade without impacting a users ability to acquire and process work units from one or more of the others.


Please consider my suggestion in the hopes that this very important project may once again demonstrate its true potential for the furtherance of our scientific goals.

Thanks,
Jim Wright

---

Jim, your idea has merit, but the real problem is money. Seti@home is running on a shoestring budget. The major support dried out years ago. Much of the equiptment being used today was either donated or sold to the project at a higly discounted price. If you know someone who could come up with the money to do as you suggest, then problems could be fixed. But for now the project relies on donations from us the participants and of course some money from the University and major donors.

---

This has been proposed and discussed before, and the major stumbling block is that the communications between each site to maintain synchronization of data would far outweigh any advantages, for a project running on a shoe string.

Plus there would be additional costs in maintaining a second site. personnel, rent, etc etc.

This has been proposed and discussed before, and the major stumbling block is that the communications between each site to maintain synchronization of data would far outweigh any advantages, for a project running on a shoe string.

Plus there would be additional costs in maintaining a second site. personnel, rent, etc etc.

But.........................

don't let it discourage anyone from attempting to find an "unthought of" solution. Perhaps someone will.........

Here is a nasty idea: use the clients (all of us) for backup and even routine tasks. I'm too tired to even think about what all would be involved in turning boinc/seti into a living and breathing entity living in dynamic P2P-land.

SETI@HOME has obviously grown so large as to be unmanageable in its present form. To fix this impasse, I offer the following suggestion:

--------------------

Clone the systems as follows:

(much removed)

Thanks,
Jim Wright

I'm not so sure that it has grown so large as to be unmanageable, certainly many of us here were crunching just fine during the "local heat wave" this weekend.

Your comment reminds me of The Twelve Networking Truths, RFC-1925

Specifically, #6.

I understand your suggestion. Take half of the servers and make SETI-A. Half the splitters, half the transitioners, half the validators, and feed half of the tapes into SETI-A. Do the same with the rest of the servers, making SETI-B. Let everyone connect as they will -- everyone could even connect to both if they wished.

Put 'em all in the same closet (where they are now), and on the same bandwidth, so no site issues.

It'd mean the folks in Berkeley would have two projects to monitor, not one.

But what I'd like to point out is that through most of the outages, most of us have never run out of work. With 5.10.20 and the right cache settings, the odds of actually running out are vanishingly small.

We have the bad habit of thinking that the SETI servers need the same level of reliability as a big E-Commerce site like Amazon.

Trouble is, when you do that you're measuring an internal point in the BOINC system, and not measuring BOINC overall.

The BOINC client is part of the system. If you keep 4 "extra days" of work in the BOINC client, then you won't run out of work for at least three days.

If you crunch two projects, BOINC can run out of work for one, and crunch the other until the first project comes back -- and then balance out the time "loaned" during the outage.

Either way, the BOINC servers can be down a significant amount of time while the BOINC system keeps running at 100%.

---

give up $25 bucks a year and donate! plain and simple if u like and want this project to exist donate!

Month Amount Number of Donations Mean Donation
Oct 2006 $2101.00 31 $67.77
Nov 2006 $13394.00 272 $49.24
Dec 2006 $71793.82 1417 $50.67
Jan 2007 $26329.10 166 $158.61
Feb 2007 $56490.15 148 $381.69
Mar 2007 $5712.12 78 $73.23
Apr 2007 $12927.13 73 $177.08
May 2007 $8282.20 151 $54.85
Jun 2007 $1649.00 42 $39.26
Jul 2007 $3893.63 39 $99.84
Aug 2007 $1805.45 41 $44.04
Sep 2007 $909.00 20 $45.45
Oct 2007 $1470.93 20 $73.55

---

SETI@HOME has obviously grown so large as to be unmanageable in its present form. To fix this impasse, I offer the following suggestion:

"obviously" is a very dangerous word to use in the world of Science...

--------------------

Clone the systems as follows:

1) Make an identical copy (or two or three) and independent installation(s) of SETI@HOME, each containing all the elements necessary for splitting, download and upload, etc. They could be named SETI@HOME_A, SETI@HOME_B, etc.

2) Each copy would process a chosen portion of the available data and then contribute the results to a common science/results database.

3) Any user would be eligible to subscribe to one or more of the sites at their choice.

4) Redistribution of the existing equipment may be possible thus avoiding acquisition of any more.

--------------------

That is in effect already done within the Boinc server framework. Multiple servers can be brought online to share the workload.

There is one common bottleneck with using a central database to coordinate all operations and manage the Boinc system state. However, there are various implimentations for running a database in parallel... I don't think s@h have quite hit that limit yet.

The main limit for s@h is available hardware on which to host the Boinc system, and more recently, cooling capacity for their server closet!

[...]
The project benefits by being able to take one site down for maintenance or upgrade without impacting a users ability to acquire and process work units from one or more of the others.
[...]

There is much scope already provided within Boinc already to have the upload and download servers spread across the world. I believe such as Einstein@home and CPDN already make use of this.

Even if some of the 'Boinc backend systems' are down for maintenance, the users still get to upload and download data.


Good ideas,

Happy crunchin',
Martin

---

See new freedom: Mageia Linux
Take a look for yourself: Linux Format
The Future is what We all make IT (GPLv3)

Here is a nasty idea: use the clients (all of us) for backup and even routine tasks. I'm too tired to even think about what all would be involved in turning boinc/seti into a living and breathing entity living in dynamic P2P-land.

I'm working on a graduate school project to research "Decentralized Distributed Computing". I'll let you know the findings when the semester is over (December). By decentralizing SETI, in theory we could accomplish more with less overhead at Berkeley.

Here is a nasty idea: use the clients (all of us) for backup and even routine tasks. I'm too tired to even think about what all would be involved in turning boinc/seti into a living and breathing entity living in dynamic P2P-land.

I'm working on a graduate school project to research "Decentralized Distributed Computing". I'll let you know the findings when the semester is over (December). By decentralizing SETI, in theory we could accomplish more with less overhead at Berkeley.

I think it would be a bad idea for several reasons.

1) What if a person decides they no longer want to crunch SETI? How would any backup related data be restored if the person simply removes BOINC without contacting the servers.

2) What if a person's hard drive crashes? What happens to all the lost data on the individual's machine that may have been very important to SETI?

3) This would reduce security of the scientific data as it would be too easy to manipulate the data going out to other users in a P2P environment.

4) All data would have to be taken from SETI's servers at some point (as they are the ones creating all the data) and returned at yet another point (as they have to store the results to be looked at further). Basically, this is not reducing the network load at all but pushing around the problem.

---