Getting work to do

[Paratima]

Sent. Will post reply.

Having said that, let me point out that answers to this type of question have been posted before. Essentially, (as well as I understand it) the parameters of the current situation are flung out to the troops. Answers come back and are applied to the problem. Fine, so far. Rinse and repeat indefinitely. Then the server gets to a point where it has to figure out a way "over the hump". Processing slows. If it's a major hump, processing stops. ZZZ's happen. Impatience ensues. Sometimes, the server gets SO hung up that only a kick in the butt will get it going again. Most times, it gets there by itself, although it does seem to slow down overall with time and eventually requires a reboot to get the ball rolling.

The situation is vastly improved over the pre-Lijun period. He has really improved the situation, but it appears to be limited in scope, possibly by the way the problem is being approached. In other words, it might be possible to get a quantum leap, but at this stage it might require a major rewrite.

I have forwarded your concerns and also asked for a project progress update, as we haven't heard anything in quite a while.

Stand by.

[Chuck]

Paratima, thanks for sending out the email promptly.... Greatly appreciated! I guess the decision to inquire to Lijun was a nobrainer.

I noticed a lot more work starvation today, but in total, it seems as if we all are getting more work done. Bravo, and thank you Alpha!!!

This is one of those times when I wish I knew/remembered the win32 internals.

Alpha, again thanks for being there with the knowledge and right tools for the job at the right time as well as the 'manual way' (love working with regedit) of getting it done. I use PC-Tools other products, but not this one yet... guess now I will be needing it as I lose that knowledge/memory of the internals.

Thanks to EVERYONE who has been keeping their eyes open and looking in documentation to find the parameter names and values. I suspect this will have impact on MANY of our projects!!!!!

Good night all.... I look forward to some great news about Eon improvements as well as our other DC projects. I smell some great impact on other projects and can't wait to 'catch the other teams with their pants down'...


Cheers,
Chuck

[Paratima]

Interestingly, I checked my registry and found that the change to 30 seconds had already been made!

Then I remembered that, using Firefox as my browser, I had installed a helper app named, ingeniously, Tweak Network Settings. Ran it once after installation, selected "Max Everything". (Well, of course!) And never thought about it again. Seems to have made that adjustment, among others.

Will be interested at Lijun's response - also cc'd Graeme (Dr. Henkelman) who is the overall project admin and who wrote the original code, btw. He it was who brought EON from its birthplace in UWashington to UTexas.

[Paratima]

PS. Ironbits and I pretty much agree on this about EON:

1. Put as much horsepower on it as you feel moved to do.
2. Make sure everything's running nicely.
3. Turn your monitor's ON/OFF switch to the OFF position
4. Go play with your kids/dog/cat/wife/motorcycle/whatever. DON'T watch it and you'll be happier.

[alpha]

Excellent, I'm glad to hear it if you're experiencing improvements since making that change to the registry. I'm not sure if a reboot is required or not after the tweak. I don't think I've rebooted yet so I can't really comment on whether I'm seeing any changes or not. Actually, I'll reboot now.

Of course, it would be great if we could get less ZZZs and jump on this as a primary project instead of a side runner. We could then maybe push for top place. I use a tool called SetPriority by Digital Parasite to make sure eOn gets the cycles when work is available.

Edit: the reboot seems to have done the trick, I'm seeing the sockets closing much quicker now.

[Paratima]

OK, here's the best I can do at the moment.
First, see this link, which describes the start-up.

Then see 2nd link, which talks about the changes made a year ago.

See my feeble description of the processing cycle above.

After I passed along your query, I got this:
Hi Les,
We are getting a new graduate student to work on this EON project. It is gonna take some time for him to get used to it. Thanks for advice.
Lijun

So, some days, you're the pigeon; some days, you're the statue. Progress is sometimes just about as slow as the processes being calculated. Could be wrong, but I read the response as being, "We're reasonably happy with what we've got and any changes are going to take a while."

They may not have the resources to do any better right now. As I explained above, it's not particularly a lack of processing power as much as the cycle of the processing algorithm. Not knowing any more about the overall scheme of things makes it hard to make meaningful suggestions. I, and others, have found some correctable items in the past, but unless you're there...

[movieman]

If you want to make EON "crank" do a ton of installs.
One day for chuckles I decided to see if I could match that 'Welltest" account which is their computer lab with 75 dual cores.
Tossed 2 clovers, a Q6600 and an old FX51 at it.
60 installs on each of the clovers, 35 on the Q6600 and 7 on the FX51 to max out the machines..
Yup, I got 'em that day..