Some General Advice to guide Seti Refugees unfamiliar with Einstein

Gary Roberts
Gary Roberts
Moderator
Joined: 9 Feb 05
Posts: 5,119
Credit: 36,756,289,611
RAC: 37,674,448
Topic 221745

Seti Refugees are very welcome here!  There will be a number of regulars who will do their best to help you if you have problems.

If you are new to Einstein but with a Seti background, you may find some things a bit different.  I'll document some stuff as a series of short points.  I won't add too much detail at first but if it seems helpful later on, I'll come back and add more.  If anyone wants a more complete description, post specific (but separate) questions in this forum, via your own thread.  For the time being this thread will be locked.  Depending on what gets asked elsewhere, I am likely to add more here.  The intention is to make it easy for people to review the full advice without the risk of things wandering off-topic.  I don't claim to be infallible.  Please point out anything you feel that needs to be corrected/modified/better explained.

I started with Seti Classic in 1999, then Seti-BOINC and switched to Einstein when it first opened in 2005.  I stopped running Seti soon after that.  I have also run LHC, Nanohive, POGS, Milkyway and Collatz.  For around the last 8 years or so, I've essentially run only Einstein.  For that reason, I have little knowledge of how other projects behave these days.  However, I think I have some clues to share as to how to make your transition to Einstein as painless as possible.

1. At the time of writing, there are 4 main searches of most interest to the project.  These are:-

(i) The gravitational wave search in LIGO data (GW) - CPU cores - Observation Run 2 Multi-Directional (O2MD1).

(ii) The gravitational wave search in LIGO data (GW) - GPUs - Observation Run 2 Multi-Directional on GPUs (O2MDF).

(iii) The gamma-ray pulsar search in LAT data from Fermi satellite - using CPU cores - FGRP5.

(iv) The gamma-ray pulsar binary search in LAT data from Fermi satellite - using discrete GPUs - FGRPB1G.

If using a GPU, you should change your project preferences to ONLY have one GPU search ticked, initially.  You may also find it easier to have just one CPU search selected.

2. For complicated reasons, running both GPU searches will tend to make it hard to have stable operating conditions.

To guide your choice as to which GPU search to run, ask yourself this question, "Do I want maximum science benefit for the project or do I want maximum credit for myself with an efficient and long tested (less problems) app?"  For the former -> O2MDF,  for the latter -> FGRPB1G.  Both CPU apps are stable so choose one or both (or none) as you see fit.

3. Seti people may be used to needing large work cache sizes.  If you don't change that you will likely have problems.

To get started here, temporarily set your cache size as low as practicable and preferably use just the first of the two settings. Using 0.1 days as the minimum plus zero extra days would be ideal. You are trying to start with just a few tasks at first, until you see how they behave.  Once things are working well you can always increase.

4. Einstein uses Duration Correction Factor (DCF) to adjust the estimated crunch time of tasks (and probably always will).

There is only one DCF for the entire project.  This means that badly estimated run times for one search will affect the run times of other searches through changes in the DCF.  This can cause quite violent changes in estimated times which will upset the true size of work caches.  In particular, this is why you should at least start with just one GPU search selected.

5. If you have a discrete GPU and a GPU search is selected, make sure BOINC is allowed to use it if you aren't getting tasks.

6. If you have selected just the searches you want, turn off the settings for beta test apps and non-preferred apps.

If you don't turn these off, you can be given tasks/apps that you haven't specifically selected.  Of course, once you understand the consequences, you can always turn these on later.

7. If you run multiple projects, remember that the last set of website compute preferences you change will be used for all.

It's probably a good policy to have a 'master' project and always change your compute prefs there.  Some compute preferences can be modified locally in BOINC Manager.  These will override website preferences so do think about the implications of that.

8. The Einstein website is probably quite different to what you have been used to.

Because of this, it's very worthwhile to spend some time exploring.  If you change anything here, remember to scroll down to the bottom of the page and find the 'save changes' link.

9. If you have multiple computers and need different prefs for some, explore the use of 'locations'.

These used to be called 'venues' and the standard 4 are available.  Please explore what is available if you need them.

Please have fun.  While you're doing that, please remember how big a deal it was when the very first detection of GW occurred just a few years ago via a BH-BH merger.  Just imagine how big a deal it's going to be when some lucky volunteer's computer processes the task that leads to the first ever detection of continuous gravitational waves that surely permeate the entire universe.  So far they are undetected.  They are just waiting for you to find them :-).

Last Updated: April 07, 2020.

Cheers,
Gary.