PrimeNet server replacement

PrimeCruncher · 2003-10-24, 01:58

See thread further down: "Replacing Entropia's PrimeNet server"

So are we or aren't we? I see the thread kind of fizzled out, so I take it the answer is the latter, which is too bad because some people had good feature ideas that really should be implemented.

I have a request myself. On the individual stats page it would be nice if you could sort the reserved exponent list by column, i.e not just by exponent but also by the computer that reserved it, bit depth, type of assignment, etc, etc.

Just my $0.02...

eepiccolo · 2003-10-24, 13:02

I purpose that before anyone starts making a bunch of suggestions, we wait for news from Prime95 or Old Man Primenet about server replacement so that if it doesn't happen, we will not have made a bunch of suggestions in vain.

PrimeCruncher · 2003-10-27, 20:42

eepiccolo, I agree completely which is why I waited a few days to see if anyone would respond. Since no one has, I guess the answer to my question is probably one of two possibilities:

1) No
2) Haven't decided yet

Anyone know for sure?

QuintLeo · 2003-10-27, 21:49

I believe the correct answer is "eventually, but it's lower priority than working, eating, sleeping, school, etc."

9-)

Complex33 · 2003-10-29, 00:58

A side question. How difficult would it be to add P-1 factoring as an assignment choice. I ask because I have a friend with dual Xeon 3.06 with 4 gigs of RAM under the account Legget. As mentioned by GP2 I have requested quite a few DC exponents to P-1. I have now released the exponents that do have already been P-1'd. I have been just releasing any exponents that have had P-1 completed.

GP2 · 2003-10-29, 17:37

I have a feeling that there is little chance of modifying the current server code. Perhaps it is even proprietary and owned by Entropia, which Scott Kurowski is no longer associated with. And Scott appears to be too busy with real-life business projects to devote much time to writing a new server.

Has any thought been given to the new BOINC infrastructure for distributed computing projects? SETI@Home will be transitioning to this very soon.

GIMPS meets all the criteria for a BOINC project. Perhaps it's worth evaluating.

Old man PrimeNet · 2003-11-18, 23:09

Entropia retains ownership of PrimeNet's code, however I secured rights to use and modify it as needed to host and support GIMPS indefinitely, subject to a few nominal restrictions. The v4 server is a few years old but is chugging along nicely after some maintenance last summer.

Old man PrimeNet · 2003-11-18, 23:09

At least I'm not as slow as the IETF!

I believe it's important not to rush a new server, particularly an open source design. Some advance work to properly stage this effort has been underway for a while, maybe emerging January '04.

When I wrote PrimeNet v1 in winter '96/'97 and v2 in summer '97 (for v14 clients) I pretty much crashed through it, unconcerned (er, ignorant) as everything was fresh ground. Then came the hard work. There are some subtleties unique to the task longevity and state transitions of GIMPS exponent tests that took a few rewrites (v3-v4 servers, v15-v19 clients) over a few years to converge upon what worked well, lessons hard won.

To illustrate why using another distributed system would probably not be a quick solution, here's a code-by-volume breakdown of server v4:

managing GIMPS exponents 32%
managing accounts, teams 21%
database reporting 15%
operational administration 12%
web site integration 8%
managing machine resources 7%
validation & security 5%

Managing exponents, operational admin, reporting and validation - having requirements distinct to GIMPS - are over half the code. These must be implemented anew for a v5 server regardless of what platform may provide the other functions.

In any case, the results will be worth everyone's patience.

PrimeCruncher · 2003-11-19, 02:08

I see. No rush, I'm sure the wait will be worth every second. But just out of curiosity, when do you think v5 will be ready? End of 2004, sometime in 2005?

Old man PrimeNet · 2003-11-19, 05:31

Once the staging work is done things should pick up fairly quickly. Assuming George and I get this out by mid-January, a wild guess from experience... maybe we'd see a new server in late testing by April/May '04. An initial project outline & timeline will be posted, revised then periodically updated as participation engages and moves forward.

GP2 · 2003-11-19, 06:38

The thing is, SETI has literally 100 times the number of users as GIMPS. This means that BOINC, however limited or even flawed its initial implementation might be (I haven't had the chance to look at it closely yet), is likely to become some kind of de facto standard platform for "volunteer distributed computing".

What I'm referring to "volunteer distributed computing" is really a very different thing than for-pay grid computing where a central organization pays for and fully controls the computing resources. With volunteer computing, things like "customer service" (teams, transferring credit, handling special requests, etc), fancy statistical tables and graphs, accomodating unusual environments (firewalls, proxies, sneakernet), limiting CPU or memory usage at certain times and days, allowing users to split their CPU time easily among different distributed computing projects... all of those things are really not luxury frills, but increasingly essential for participant recruitment and retention.

The point with BOINC is that we're likely to see third-party freeware add-ons or plug-ins, or contributed code improvements, for handling stats and teams and routine admin tasks or routine user tasks (for instance, splitting work around a home farm and centrally monitoring all your home machines from a single display).

Those "nice-to-haves" might soon come to be considered essential, and potential participants might shun projects that don't provide them. We might be out in the cold if we're a PrimeNet Mac in a BOINC Windows world...

Time and again, the history of computing shows that a product with a large user base usually wins against even a technically far superior product with a smaller user base. They get the third-party contributed added value, become the center of an "ecosystem", and gradually improve over time to become the dominant standard.

So what I'm saying -- mind you, without ever having seen the PrimeNet code or looked too closely at the BOINC code, just making the market share argument based on the past history of software -- should we seriously consider BOINC even if it means abandoning work done thus far on PrimeNet 5?

PS,
United Devices delayed BOINC with litigation, since the head of SETI@Home worked for them for a couple of years, alleging misappropiation of trade secrets. While some say this was more for strategic reasons rather than based on a valid claim, and a settlement was reached after a few months, nevertheless... could Entropia pursue the same tactic?

2003-10-24, 01:58	#1
PrimeCruncher Sep 2003 Borg HQ, Delta Quadrant 2·3³·13 Posts	PrimeNet server replacement See thread further down: "Replacing Entropia's PrimeNet server" So are we or aren't we? I see the thread kind of fizzled out, so I take it the answer is the latter, which is too bad because some people had good feature ideas that really should be implemented. I have a request myself. On the individual stats page it would be nice if you could sort the reserved exponent list by column, i.e not just by exponent but also by the computer that reserved it, bit depth, type of assignment, etc, etc. Just my $0.02...

2003-10-29, 00:58	#5
Complex33 Aug 2002 Texas 5·31 Posts	A side question. How difficult would it be to add P-1 factoring as an assignment choice. I ask because I have a friend with dual Xeon 3.06 with 4 gigs of RAM under the account Legget. As mentioned by GP2 I have requested quite a few DC exponents to P-1. I have now released the exponents that do have already been P-1'd. I have been just releasing any exponents that have had P-1 completed. Last fiddled with by Complex33 on 2003-10-29 at 01:00

2003-10-29, 17:37	#6
GP2 Sep 2003 5×11×47 Posts	I have a feeling that there is little chance of modifying the current server code. Perhaps it is even proprietary and owned by Entropia, which Scott Kurowski is no longer associated with. And Scott appears to be too busy with real-life business projects to devote much time to writing a new server. Has any thought been given to the new BOINC infrastructure for distributed computing projects? SETI@Home will be transitioning to this very soon. GIMPS meets all the criteria for a BOINC project. Perhaps it's worth evaluating. Last fiddled with by GP2 on 2003-10-29 at 17:37

2003-11-18, 23:09	#8
Old man PrimeNet Jan 2003 Altitude>12,500 MSL 101 Posts	At least I'm not as slow as the IETF! I believe it's important not to rush a new server, particularly an open source design. Some advance work to properly stage this effort has been underway for a while, maybe emerging January '04. When I wrote PrimeNet v1 in winter '96/'97 and v2 in summer '97 (for v14 clients) I pretty much crashed through it, unconcerned (er, ignorant) as everything was fresh ground. Then came the hard work. There are some subtleties unique to the task longevity and state transitions of GIMPS exponent tests that took a few rewrites (v3-v4 servers, v15-v19 clients) over a few years to converge upon what worked well, lessons hard won. To illustrate why using another distributed system would probably not be a quick solution, here's a code-by-volume breakdown of server v4: managing GIMPS exponents 32% managing accounts, teams 21% database reporting 15% operational administration 12% web site integration 8% managing machine resources 7% validation & security 5% Managing exponents, operational admin, reporting and validation - having requirements distinct to GIMPS - are over half the code. These must be implemented anew for a v5 server regardless of what platform may provide the other functions. In any case, the results will be worth everyone's patience.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Replacing Entropia's primenet server	Prime95	PrimeNet	228	2006-02-04 07:40
Report of monitoring primenet server unavailability	Peter Nelson	PrimeNet	13	2005-10-18 11:17
mprime stalls if primenet server is unavailable :(	TheJudger	Software	1	2005-04-02 17:08
Primenet Server Oddity	xavion	PrimeNet	28	2004-09-26 07:56
So, What's the status of the "Server Replacement"	Joe O	PrimeNet	4	2002-12-11 05:31

2003-10-24, 13:02	#2
eepiccolo Dec 2002 Frederick County, MD 2×5×37 Posts	I purpose that before anyone starts making a bunch of suggestions, we wait for news from Prime95 or Old Man Primenet about server replacement so that if it doesn't happen, we will not have made a bunch of suggestions in vain.

2003-10-27, 20:42	#3
PrimeCruncher Sep 2003 Borg HQ, Delta Quadrant 2·3³·13 Posts	eepiccolo, I agree completely which is why I waited a few days to see if anyone would respond. Since no one has, I guess the answer to my question is probably one of two possibilities: 1) No 2) Haven't decided yet Anyone know for sure?

2003-10-27, 21:49	#4
QuintLeo Oct 2002 Lost in the hills of Iowa 2⁶×7 Posts	I believe the correct answer is "eventually, but it's lower priority than working, eating, sleeping, school, etc." 9-)

2003-11-19, 02:08	#9
PrimeCruncher Sep 2003 Borg HQ, Delta Quadrant 2·3³·13 Posts	I see. No rush, I'm sure the wait will be worth every second. But just out of curiosity, when do you think v5 will be ready? End of 2004, sometime in 2005?

2003-11-19, 05:31	#10
Old man PrimeNet Jan 2003 Altitude>12,500 MSL 101 Posts	Once the staging work is done things should pick up fairly quickly. Assuming George and I get this out by mid-January, a wild guess from experience... maybe we'd see a new server in late testing by April/May '04. An initial project outline & timeline will be posted, revised then periodically updated as participation engages and moves forward.

2003-11-19, 06:38	#11
GP2 Sep 2003 101000011001₂ Posts	The thing is, SETI has literally 100 times the number of users as GIMPS. This means that BOINC, however limited or even flawed its initial implementation might be (I haven't had the chance to look at it closely yet), is likely to become some kind of de facto standard platform for "volunteer distributed computing". What I'm referring to "volunteer distributed computing" is really a very different thing than for-pay grid computing where a central organization pays for and fully controls the computing resources. With volunteer computing, things like "customer service" (teams, transferring credit, handling special requests, etc), fancy statistical tables and graphs, accomodating unusual environments (firewalls, proxies, sneakernet), limiting CPU or memory usage at certain times and days, allowing users to split their CPU time easily among different distributed computing projects... all of those things are really not luxury frills, but increasingly essential for participant recruitment and retention. The point with BOINC is that we're likely to see third-party freeware add-ons or plug-ins, or contributed code improvements, for handling stats and teams and routine admin tasks or routine user tasks (for instance, splitting work around a home farm and centrally monitoring all your home machines from a single display). Those "nice-to-haves" might soon come to be considered essential, and potential participants might shun projects that don't provide them. We might be out in the cold if we're a PrimeNet Mac in a BOINC Windows world... Time and again, the history of computing shows that a product with a large user base usually wins against even a technically far superior product with a smaller user base. They get the third-party contributed added value, become the center of an "ecosystem", and gradually improve over time to become the dominant standard. So what I'm saying -- mind you, without ever having seen the PrimeNet code or looked too closely at the BOINC code, just making the market share argument based on the past history of software -- should we seriously consider BOINC even if it means abandoning work done thus far on PrimeNet 5? PS, United Devices delayed BOINC with litigation, since the head of SETI@Home worked for them for a couple of years, alleging misappropiation of trade secrets. While some say this was more for strategic reasons rather than based on a valid claim, and a settlement was reached after a few months, nevertheless... could Entropia pursue the same tactic?