FAQ:Einstein@Home

From Unofficial BOINC Wiki

Contents 1 General 2 What should my Work Buffer settings Be Set To? 2.1 If your machine is on ~ 24/7 2.2 If your machine is regularly turned off, but runs >40hr / week 2.3 If the machine is on less than about 40 hours / week 3 How large are the Work Units for Einstein@Home?

[edit] General

Welcome to the Einstein@Home Frequently Asked Questions (also known as F.A.Q. or FAQ)! In here we hope that you will find answers to some of the questions that you might have. The material in this FAQ is going to be oriented towards only those subjects that are strictly related to the Einstein@Home Project as it relates to BOINC. Items that are better covered in other places will not be explained here. We will try to put your feet on the right road however.

We suggest that you read the BOINC FAQ first, and then read the FAQ for the BOINC Powered Project that you intend to join. Though we are trying for complete coverage, there are things that are discussed in those other places there that may not be covered in this FAQ.

If you cannot find your question on this page (or any other page for that matter), please submit it to me, Paul D. Buck (call me Paul), and I will try to add your question to the appropriate section.

[edit] What should my Work Buffer settings Be Set To?

The following advice is based on my experience with my 700MHz machines.

[edit] If your machine is on ~ 24/7

Settings that work for me are to have just E@H on board, and a cache = 0.7 day. This keeps one WU in hand at all times, and for me it swaps the returned one for a new one a few hours after the start of the next. If you find it swapping Work Unit just before the next one finishes, reduce the number in steps of 0.2 till it gets it right. I'd suggest, from my own experience, don't go above 0.7.

If you have a cache setting much bigger (over about 2.5) the scheduler can get be too optimist and take on too much work. I lost about a dozen Work Unit over my two machines in my second week! (Gary will remember helping me sort it out)

Certainly with a machine in this speed range you do not want more than 3 Work Unit held locally at any time (including completed ones) preferably only 2.

I ran both machines on these settings 24/7 till I stopped leaving them on all he time (ie once summer arrived).

[edit] If your machine is regularly turned off, but runs >40hr / week

This advice also applies if you have the machine on 24/7 and more than one project, if E@H has a resource share from 30% to 90%.

Please use a cache setting of 0.1. Before turning off, please check if a Work Unit is complete. If so, then click on 'update now' before power down, to let the returned Work Unit be registered.

This is especially important at weekends: the client will try to return a Work Unit automatically when it is within 24hours of deadline, but if you have a Work Unit that expires Sunday and you turn off the machine on Friday it will not have been returned automatically; then on Monday it will be too late.

Again I write from experience: one of my machines is running this sort of pattern now, and without losing Work Unit.

[edit] If the machine is on less than about 40 hours / week

Then I'd really advise against running E@H, much as we appreciate the intention. Other projects have different needs, predictor@h for example will run very nicely with the machine going on and off - again I write with experience, both my machines have crunched Predictor@H when they've been on less than 40h /wk:

[edit] How large are the Work Units for Einstein@Home?

The Work Units for Einstein@Home are very large compared to most of the other BOINC Powered Projects. In fact they can take up to an hour to download when using a 56 kbps Dial-Up Modem. This means that if you are using Dial-Up service, particularly if you are paying by the hour; the Einstein@Home Project may not be one that is right for you.

The Result Data Files may be equally large. Though most of the time they seem to be in the area of one to two megabytes. So, just a word of caution if you use a Dial-Up Modem.

However, in an attempt to mitigate the cost of the download, the Scheduler will attempt to assign you several Results from the one Work Unit Data File. As an example, from 553 total Einstein@Home Work Units, the average number of Results I got from each Work Unit Data File was 3.6, the distribution looked as shown in the table.

In another example, a message log from Mary Jo Kelleher showed a total of 13 Results from one Work Unit Data File. Way better than my one 8 bagger. Still, getting 4 Results from one download is not that bad.

A feature that the Einstein@Home staff has put into the Scheduler is a feature that makes it more likely that high-bandwidth connections will get more downloads with less reuse. Some Participants that are on dial-up connections, report an average that is a lot higher than the numbers I quote above. In fact, I would go so far as to guess that with my high-speed connection I see a much lower than average reuse. The amount of reuse also depends on the characteristics of the data run and the Average Turnaround Time of the Host.

With this news, hopefully you modem users will be a little less scared about joining the project! If I get good reported numbers I will try to let you know ... So, dial-up users, make a log and if I get reports I will post them.