Einstein @ Home

S5R3 Task Crunch Time Ready Reckoner V5A

Please follow these steps to obtain estimates of task behaviour

Step 1

Important Points To Read

  • All times are in seconds.
  • Predictions are designed for data from a single machine and a given application version only.
  • For Step 4 data entry particularly :
    • Minumum two inputs, and generally the more the better.
    • Rigorous input checking applies to CSV input into left hand side text area/pane.
      • Three data values per line.
      • Values interpreted in order as FREQUENCY, SEQUENCE NUMBER and RUNTIME respectively.
      • FREQUENCY interpreted as floating point.
      • SEQUENCE NUMBER interpreted as integer.
      • RUNTIME interpreted as floating point.
      • Comma to separate each value.
      • No spaces except those leading the first data value or trailing the last ( trimming occurs ).
      • Abberant lines that do not parse as above will be discarded.
      • Each line represents a data point for analysis. There is no assumption of any ordering from one line and the next.
    • Choose a spread of task phases for a given frequency. Analysis ouput will be sorted into ascending frequency, then by ascending sequence number within each frequency.
    • An expected peak, average and trough runtimes plus variance and error is calculated given all qualifying tasks at each given frequency. This is subject to triage as regards fitness to predict ( complex ), and is shown on the right hand text area/pane
    • An expected peak, average and trough runtimes plus variance and error is calculated using all qualifying tasks at all frequencies. This appears in the lower "output" boxes. User must decide whether or not to subsequently use these values which may represent data across different search frequencies. I have yet to satisfy myself as to how well cross-frequency predictions stand up - your experience/comments are welcomed!
    • The right hand pane is editable. Any changes you make will be reflected when "Inputs and Outputs Summary" is selected.
    • If no error is quoted then none applies. Zero will be explicit.
  • You may of course return to a previous step and re-input any parameters.
  • Re-calculation is automatic after a given data entry, excepting the use of the buttons in Step 4. Here you must 'nudge' the process along as you please from ( but possibly starting at "Clear Inputs" ) "Use Inputs" to "Use These Values In Step 6" and/or "Inputs and Outputs Summary".
  • NaN in a field indicates either an bad input somewhere earlier, or an unacceptable calculation result ( various causes ).
  • This an UNOFFICIAL toy. No warranties are implied or express whatsoever.
  • I would be very glad to receive any & all feedback ! ;-)

Enjoy!
Cheers, Mike

Step 2

Firstly we need to find and examine the task identifier. It looks rather like this :

h1_XXXX.XX_S5R2__YYY_S5R3a_??

XXXX.XX is a decimal number like 345.67 which is the search frequency F for the task.
YYY is an integer number like 123 which is the sequence number S for the task.

  • Open BOINC Manager and maximise the window.
  • For a running task you will find it by
    • Select the Tasks tab
    • Look in the column under the Name heading
  • Alternatively you could also find it by
    • Select the Projects tab
    • Highlight Einstein@Home in the column under the Project heading
    • Click the Your results button toward the bottom of the Web Sites panel on the lower left
    • This will open your browser at the page which shows your recent tasks including any pending ones
    • Select a task of interest by clicking on one in the column under the task ID click for details heading

Step 3

We will first calculate the period P of the task cycle.
This depends upon the sky search frequency F .

INPUT SEARCH FREQUENCY   ——> = F
OPTIONAL SKY GRID DENSITY CONSTANT   ——> = D
 
OUTPUT CYCLE PERIOD   ——> = P
OUTPUT SEQUENCE NUMBERS WITH HIGHEST RUNTIMES   ——>
OUTPUT SEQUENCE NUMBERS WITH LOWEST RUNTIMES   ——>

Step 4 [ OPTIONAL ]

For a sequence of tasks you may wish to attempt an estimation of runtime parameters.
This depends upon the frequency F and sequence number S and running time T for as many entries as your runtime system can manage. There is no pre-programmed limit, and expect noticeable pauses for large data sets! :-)
Please note this advice!

OUTPUT PEAK RUNTIME   ——> = A Error ~   %
OUTPUT AVERAGE RUNTIME   ——> = M
OUTPUT TROUGH RUNTIME   ——> = L
OUTPUT RUNTIME VARIANCE   ——> = B
   
   

Step 5

Now we will calculate the phase R of the task within the cycle, and sequence numbers with similar runtimes T.
This depends upon the task sequence number S and the period P.

INPUT SEQUENCE NUMBER   ——> = S
 
OUTPUT PHASE WITHIN SEQUENCE   ——> = R
OUTPUT SIMILIAR RUNTIMES WITH   ——>

Step 6

To finish we will calculate the expected runtime T of your task.
This depends upon estimates of the peak runtime A, the runtime variance B and the phase R.

INPUT PEAK RUNTIME ESTIMATE   ——> = A
INPUT RUNTIME VARIANCE ESTIMATE   ——> = B
 
OUTPUT ESTIMATED RUNTIME   ——> = T