Changes between Version 99 and Version 100 of ProjectOptions


Ignore:
Timestamp:
Jun 2, 2010, 10:08:49 PM (14 years ago)
Author:
davea
Comment:

--

Legend:

Unmodified
Added
Removed
Modified
  • ProjectOptions

    v99 v100  
    9393Otherwise, the overall maximum is '''N*NCPUS + M*NGPUS)'''.
    9494
    95 {{{
     95
     96{{{
     97<gpu_multiplier> GM </gpu_multiplier>
     98}}}
     99If your project uses GPUs, set this to roughly the ratio
     100of GPU speed to CPU speed.
     101Used in the calculation of job limits (see next 2 items).
     102{{{
     103<max_wus_to_send> N </max_wus_to_send>
     104}}}
     105Maximum jobs returned per scheduler RPC is '''N*(NCPUS + GM*NGPUS)'''.
     106You can use this to limit the impact of faulty hosts.
     107Default is 10.
     108{{{
     109<max_ncpus>N</max_ncpus>
     110}}}
     111An upper bound on NCPUS (default: 16)
     112
     113{{{
     114<daily_result_quota> N </daily_result_quota>
     115}}}
     116Each host has a field MRD in the interval [1 .. daily_result_quota];
     117it's initially daily_result_quota,
     118and is adjusted as the host sends good or bad results.
     119The maximum number of jobs sent to a given host in a 24-hour period is
     120'''MRD*(NCPUS + GM*NGPUS)'''.
     121You can use this to limit the impact of faulty hosts.
     122
     123== Job limits (advanced) ===
     124The following is a
     125more adaptable way of expressing limits on the number of jobs in progress on a host.
     126You can specify limits for specific apps, and for your projects as a whole.
     127Within each of these, you can specify limits for CPU jobs, GPU jobs, or total.
     128In the case of CPU and GPU jobs, you can specify whether the limit should be
     129scaled by the number of devices present on the host.
     130
     131This uses a separate config file, '''config_aux.xml'''.
     132The syntax is:
     133{{{
     134<?xml version="1.0" ?>
     135<config>
    96136<max_jobs_in_progress>
    97137        <project>
     
    114154        ...
    115155</max_jobs_in_progress>
    116 }}}
    117 A more adaptable way of expressing limits on the number of jobs in progress on a host.
    118 You can specify limits for specific apps, and for your projects as a whole.
    119 Within each of these, you can specify limits for CPU jobs, GPU jobs, or total.
    120 In the case of CPU and GPU jobs, you can specify whether the limit should be
    121 scaled by the number of devices present on the host.
    122 
    123 {{{
    124 <gpu_multiplier> GM </gpu_multiplier>
    125 }}}
    126 If your project uses GPUs, set this to roughly the ratio
    127 of GPU speed to CPU speed.
    128 Used in the calculation of job limits (see next 2 items).
    129 {{{
    130 <max_wus_to_send> N </max_wus_to_send>
    131 }}}
    132 Maximum jobs returned per scheduler RPC is '''N*(NCPUS + GM*NGPUS)'''.
    133 You can use this to limit the impact of faulty hosts.
    134 Default is 10.
    135 {{{
    136 <max_ncpus>N</max_ncpus>
    137 }}}
    138 An upper bound on NCPUS (default: 16)
    139 
    140 {{{
    141 <daily_result_quota> N </daily_result_quota>
    142 }}}
    143 Each host has a field MRD in the interval [1 .. daily_result_quota];
    144 it's initially daily_result_quota,
    145 and is adjusted as the host sends good or bad results.
    146 The maximum number of jobs sent to a given host in a 24-hour period is
    147 '''MRD*(NCPUS + GM*NGPUS)'''.
    148 You can use this to limit the impact of faulty hosts.
     156</config>
     157}}}
    149158
    150159=== Job-cache scheduling ===