Our plan of attack so far 5/13

  • Thread starter Deleted member 88227
  • Start date
D

Deleted member 88227

Guest
We only have three projects left.

Cosmology which lasts until the events end.
LHC which also lasts until the events end.
YOYO which ends, at the time of this post, in 2 days.

If you are just now joining us and/or adding new boxen to the race then put them on YOYO until the 16th. That's when that project ends and we need to score as high as we can on that project to get the points before it's over.

If you're currently running anything then move 75% of your boxen/cores to LHC for now. Most of the smaller teams don't seem to be doing much on this project as of right now. We can maintain our position so any boost and we should be able to maintain it up to about 3 more higher positions.

Except Fastgeek, Applejacks, phoenicis, ChristianVirtual, pututu and coleslaw. Ya'll keep your current boxen running Cosmology. If any of you have anything new adding to the mix then move them to YOYO. Once YOYO is done then you'll most likely move them to LHC, but we may move them to Cosmology depending on how well fastgeek can help us gain on the competition.
 
Switch from cosmo to yoyo last night.

In Einstein, we got bumped down by one place to 13th. The team that beat us to the 12th, just dump a shit load of WUs. It is a very high risk maneuver i.e. there were cases where the server did not respond well, and you know what happened next. Even in this competition, we have server/WU issue with WCG OZ, so I'm a bit wary about server reliability.

This morning, we gain 2 places in yoyo since last night. After this, will switch to LHC. I think we can get some points off yoyo.

With fastgeek in cosmo, I think we are in good shape, so I think the focus will be in the remainder of the two events. Where we are now, a move in one position higher, will net us 3 points, unless you are in top 4.
 
I've got a 64 cores crunching on yoyo right now. Got around double that crunching on LHC. That's close to half my boxen.

The other half is crunching on Cosmo + GPUs are mixed between Moo! Wrappy and E@H with most of them on E@H.

I might move another box to yoyo if we aren't gaining fast enough on those above us.

I do agree though. We only gain 3 points each time we rank up so focusing on Cosmo only would net us 3 points each time we rank up. These other projects have given us a lot more points and we're barely even touching them.
 
Hey guys where do you want me to move my stuff too then?
 
leave them where they are for now. I seen your post about the smaller moves, they're fine.
 
added two EC2 instances to yoyo; but hell, some WU are really long running ... 18 hours ? I hope the spot instance I use lives long enough for it (they might get cancelled if the "market price" is higher then my limit)
 
Actually, take one of the boxes you moved to LHC and move it to YOYO. Once YOYO is done you'll most likely move it back to LHC.
 
so i guess its safe to say we can pull off e@h since thats over? i could deal with less heat :)
 
Yes, you can drop E@H.

Forgot about the time difference. E@H is over. lol
 
haha ok i was wondering what this other race was :)
 
still 11 pages of pending results in E@H ... I hate that late bunkering ... for those races only project with quorum 1 please; else the result is not reflecting our performance.
 
I'm not to worry about catching up in Formula BOINC, of course with you on board.:p

Here is our current status for E@H in Formula BOINC. We are at number #4.

Comparing us with the top three teams, I think we have a good chance of catching up to no. 1 spot. Deadline is sometime in Nov.

Let's put this on the list to do till the end of Nov. Our biggest problem is to get more volunteers and not to relying heavily one or two crunchers. We probably need to discuss this as a team after the pentathlon. Perhaps a post-mortem, no bull-shit, open and frank discussion where every feedback and opinions are considered?

There are a few more projects to focus on. Let's discuss more.

upload_2017-5-13_18-27-44.png



upload_2017-5-13_18-31-48.png
 
applejacks said:
haha ok i was wondering what this other race was :)
Click to expand...

It's not a big concern right now as pututu pointed out. The race ends sometime in November. Currently we're 2nd overall and only 58 points behind 1st. We were 80+ points behind before the start of the Pentathlon and it just so happens that all the projects in the Pentathlon are also in the Formula-BOINC challenge as well. Nothing to worry about right now.

ChristianVirtual said:
still 11 pages of pending results in E@H ... I hate that late bunkering ... for those races only project with quorum 1 please; else the result is not reflecting our performance.
Click to expand...

We don't really have any choice or option in that. Since it was the only GPU project it's not like we could have moved our GPUs to another project to show our production more. Last I checked I had over 70 pages of pending results. So I feel your pain bro.

pututu, yeah I know we're number 4 now, but a week ago we wasn't even in the top 10. ;)
 
Skillz said:
Last I checked I had over 70 pages of pending results. So I feel your pain bro.
Click to expand...
I'm sure I mentioned somewhere in the forum to download as many task as possible early on in the race after the bunker dump.

My 960 (slow) I bunker enough tasks to last for 8-9 days or so and never download any new tasks when the event started, so this has the highest validation number.
970 I bunker late in the first day.
1080 was the first one that has 1/2 day of no task during the bunkering period in which I downloaded as many tasks to last for the next 5 days.

For those event with quorum of 2, you would want to download as many as possible during the early race.

I think LHC require quorum of 2 for the sixtrack and atlas only have quorom of 1. I'm testing atlas but I don't have enough RAM! You have to think of this strategy for LHC.

My E@H as of 6:45pm PST.
upload_2017-5-13_18-56-14.png
 
Yeah, bunkering takes time, time I have little of over during the time it was required.
 
Skillz said:
Except Fastgeek, Applejacks, phoenicis, ChristianVirtual, pututu and coleslaw. Ya'll keep your current boxen running Cosmology.
Click to expand...

Oops, the damned time zone difference in the UK means that strategy discussions are starting just after I call it a night. I moved everything over to LHC yesterday ..... the stalwart team mates remaining on cosmo seem to be more than holding the fort but let me know if you think a switch back of some resource would be better.

Thanks for taking an active role in strategy and attempting to herd us cats.
 
I think it'll be fine. fastgeek brought some serious computing power over to cosmology. We've already moved up a couple spots in that and, unless other teams are bunkering, we're gaining on them rather quickly. I think we have 2 or 3 more spots to easily overtake.
 
We gained 4 more spots over the last 24 hours in yoyo. This is a nice project with quorum = 1 BUT the server response is horrible.

After yoyo, I'll focus on LHC and let fastgeek does the job for us:D


upload_2017-5-14_7-41-22.png
 
One of my rigs finally filled the queue up with 700+ tasks that it can now work on. I was having issues with getting tasks yesterday.
 
With about less than 1.33 days left, I would recommend running the "C2xx" WU since this has the shortest completion time.

The "np" WU takes a long time up from 0.5 to 1 full day. I would leave this up to the team if they want to micro-manage the tasks.

The "GW" WU (not many) yields the best point per hour. Note that each WU type has a fixed points no matter how fast you turn in the result. For e.g. the C214/C215 will net you 100.72 points. With such a short challenge, I do not have time to analyze the WU types that give the best PPD since the their server/validator is not co-operating.

upload_2017-5-14_8-12-32.png
 
Sorry guys..forgot Skillz got this approved. Formula-BOINC can be fun if the team focuses on it. There are some tactics to i5 like getting the less popular projects points.

WEP-M+2 is a project a lot of people don't run because it refuses to support Windows. It is the admins objective even though he doesn't list it as part of the projects objective.
Distributed Data Mining (which requires JAVA to be installed)
Quake Catcher Network which uses unique sensors. The move important thing is that the project is on auto pilot and has no support any more. Android devices earlier than v5 may work as well as Macbooks and some IBM/Lenovo laptops. HP laptops that have the 3d Hard drive application/compatibility may still work too.
Lattice Project.... This one is a beast. Some work units may run for a few seconds. Some can run months. Some of these will want like 16GB ram per work unit. So, as you can imagine it isn't very popular these days and needs to have some measures put in place with some micro management.
RNA World has only virtualbox work units even though they claim to have more. They just don't run the others and haven't for years. The virtualbox work units run for weeks/months at a time. They also don't score well and they have to run a script to credit since they don't validate like most projects. Most work units say they fail, so don't get discouraged.
Primaboinca is a poorly managed project and does not get any responses from the admin. It is also pretty much impossible to get an account if you don't already have one. I've been PM'ing sebastion (the admin) for a while to consider removing it since it isn't fair.
MindModeling@home BETA runs out of work often and can be tough to get your fill.
SZTAKI runs good but they run out of work quite often and their work units tend to be short.
Gerasim is a headache in my opinion, but the CPU work runs good. They sometimes have work available for Intel GPU's but pay attention to which version of the c++ they require.
BURP ran well when I ran it, but I think work runs low at times.
Atlas is an LHC project that recently got merged with the main LHC site. The project can not be ran independently any more so those points are locked in.

Other strategies is to pay attention to projects that are closing soon.
A perfect example of this is Atlas as it merged with LHC. anyone that had points at the time the project ends or changes will lock those points in the rest of the year. IMO it is a flaw of the system.
Another example is GoofyxGrid. The project moved to a new project as they plan on having other sciences hosted. Unfortunately, this also means the old project ended. Luckily, I and a few others earned a shit ton of points there from running multiple clients and we locked in #1 position.

This year introduced the Sprints. As far as scoring goes, these are worth just as much as the Marathon. If priority was given, Sprints are most important. This also forces people to move stuff around before the end of the year when they may normally play clean up.

The year ends in November.

It overlaps other challenges as you may have noticed. So priorities have to be given in advance. We typically focus on Pentathlon, WCG x-mas race though we have been moving towards the WCG Birthday challenge in November. Less teams are joining in the x-mas challenge. We also typically have a few interested in the PrimeGrid challenges because the project has more control over the stats. No bunkering allowed. All work units returned regardless of validation count. We also still do DC-Vault, but we don't focus heavily on it due to its nature. I am an admin there and pretty much only one active for testing/making decisions. Rusty pretty much does what I tell him to. But it is a lot harder to move up and down older projects unless one is recently added or removed from the Vault.
 
I'm still in PrimeGrid challenge but will probably continue participating this till the end of this year. There is one team (they called themselves "rats") that dominates this project most of the time. It is their pet project. At one point, fastgeek gave them a run for their money and they took notice.

In Formula BOINC, we are in league 3. Got bumped down from league 1 to league 3 but looks like we are doing a great comeback this year if the current interest in this is maintained throughout the year.

My preference is towards well managed projects. At least my contribution/donation/money spent on them means something but once a while would just run some projects for the fun of it but shouldn't be bulk of my contribution.
 
Same here: thinks like Yo-Yo might be fun for three days but not really what I'm looking for. Also LHC should have enough money to do their own computation.

Hope soon be back to WCG and GPUGrid as primary projects.
 
LHC has many researchers contributing and using the data. It isn't just one organization that uses it. So, like any other research facility the idea is to reduce costs for them so that research is feasible. Some of these projects may never get approved otherwise.
 
You must log in or register to reply here.
Back
Top