Stripcharts explained

Machine load

The system load values correspond to the kernel metric, queried by 'w', 'top' or 'uptime' commands in Unix shell. The last minute load value is shown on the stripchart

Number of active hosts

Number of hosts that contacted the BOINC server in past several hours. By contact we mean any RPC call, initiated by a BOINC client

Number of unfinished workunits

The number of workunits in the BOINC database, the calculation of which has not been accomplished yet. There might be accomplished workunits in the database, but they are unrelated to this statistics

Days of CPU utilized

The statistics is calculated by summing up the following expression for all the hosts that contributed any credit at all: total_credit / credit_per_cpu_sec. Both total_credit and credit_per_cpu_sec are accumulated in the database on a per-host basis. The sum of ratios reflects the CPU days utilization in the long term

Active assistant hosts

Measures the number of hosts, belonging to the assistant user, that contacted the BOINC server in past several hours. The assistant user hosts are hosts where BOINC clients are restricted to run via loadd monitoring daemon

Performance of assistants hosts

By performance of a hosts group we mean the number of GFlops, contributed by the group, normalized by the number of the group hosts. The GFlops number can be derived from the credit, given to hosts. For the dedicated homogenous cluster of nodes the value of this statistics should strive to the clock rate of a cluster host. This statistics concerns only the hosts, belonging to the assistant user

Active farm hosts

Measures the number of hosts, belonging to the farm user, that contacted the BOINC server in past several hours. The farm user hosts are hosts where BOINC clients are free to run whenever the host is unoccupied by users

Performace of farm hosts

The statistics is the performance of the farm user hosts group. Look above for the definition of the hosts group performance

Active farm hosts

Measures the number of hosts, belonging to the DSL user, that contacted the BOINC server in past several hours. The DSL user hosts are hosts where BOINC clients are restricted to run via loadd monitoring daemon

Performace of DSL hosts

The statistics is the performance of the DSL user hosts group. Look above for the definition of the hosts group performance

Active LCCN hosts

Measures the number of hosts, belonging to the LCCN user, that contacted the BOINC server in past several hours. The LCCN user hosts are hosts where BOINC clients are free to run whenever the host is unoccupied by users

Performace of LCCN hosts

The statistics is the performance of the LCCN user hosts group. Look above for the definition of the hosts group performance

Active EGEE hosts

Measures the number of hosts, belonging to the EGEE user, that contacted the BOINC server in past several hours. The EGEE user hosts are hosts in EGEE, to which BOINC clients are submitted as usual EGEE jobs

Performance of EGEE hosts

The statistics is the performance of the EGEE user hosts group. Look above for the definition of the hosts group performance

Execution time of EGEE clients

The accumulated execution time of EGEE clients from the submission till the removal of the client's job from the pool

Turnaround time of EGEE clients

The turnaround time of EGEE clients from the submission till the removal of the client's job from the pool

Queuing time of EGEE clients

The queuing time of EGEE clients from the submission till the removal of the client's job from the pool: the turnaround time minus the accumulated execution time

Number of restarts of EGEE clients

The number of restarts of EGEE clients in EGEE pool till the removal of the client's job from the pool

Number of workunits per EGEE client

The number of workunits that an EGEE client computed from the submission till the removal of the client's job from the pool

Running EGEE jobs

The number of running jobs that are accounted by the EGEE batch system

Normalized efficiency of EGEE pool

Defined as (expavg_credit / 100 / avg(p_fpops) / MAXIMAL_RUNNING_JOBS), this statistics measures the percent of efficiently running machines out of all the machines that were requested by the virtual pool scheduling from EGEE submission machine

Active Madison Condor hosts

Measures the number of hosts, belonging to the Madison Condor user, that contacted the BOINC server in past several hours. The Madison Condor hosts are hosts in Madison Condor pool, to which BOINC clients are submitted as usual Condor jobs

Performance of Madison Condor hosts

The statistics is the performance of the Madison Condor user hosts group. Look above for the definition of the hosts group performance

Execution time of Madison Condor clients

The accumulated execution time of Madison Condor clients from the submission till the removal of the client's job from the pool

Turnaround time of Madison Condor clients

The turnaround time of Madison Condor clients from the submission till the removal of the client's job from the pool

Queuing time of Madison Condor clients

The queuing time of Madison Condor clients from the submission till the removal of the client's job from the pool: the turnaround time minus the accumulated execution time

Execution time of Madison Condor clients

The accumulated execution time of Madison Condor clients from the submission till the removal of the client's job from the pool

Turnaround time of Madison Condor clients

The turnaround time of Madison Condor clients from the submission till the removal of the client's job from the pool

Queuing time of Madison Condor clients

The queuing time of Madison Condor clients from the submission till the removal of the client's job from the pool: the turnaround time minus the accumulated execution time

Number of workunits per Madison Condor client

The number of workunits that an Madison Condor client computed from the submission till the removal of the client's job from the pool

Running Madison Condor jobs

The number of running jobs that are accounted by the Madison Condor batch system

Normalized efficiency of Madison Condor pool

Defined as (expavg_credit / 100 / avg(p_fpops) / MAXIMAL_RUNNING_JOBS), this statistics measures the percent of efficiently running machines out of all the machines that were requested by the virtual pool scheduling from Madison Condor submission machine

Active OSG hosts

Measures the number of hosts, belonging to the OSG user, that contacted the BOINC server in past several hours. The OSG hosts are hosts in OSG pool, to which BOINC clients are submitted as usual Condor jobs

Performance of OSG hosts

The statistics is the performance of the OSG user hosts group. Look above for the definition of the hosts group performance

Execution time of OSG clients

The accumulated execution time of OSG clients from the submission till the removal of the client's job from the pool

Turnaround time of OSG clients

The turnaround time of OSG clients from the submission till the removal of the client's job from the pool

Queuing time of OSG clients

The queuing time of OSG clients from the submission till the removal of the client's job from the pool: the turnaround time minus the accumulated execution time

Execution time of OSG clients

The accumulated execution time of OSG clients from the submission till the removal of the client's job from the pool

Turnaround time of OSG clients

The turnaround time of OSG clients from the submission till the removal of the client's job from the pool

Queuing time of OSG clients

The queuing time of OSG clients from the submission till the removal of the client's job from the pool: the turnaround time minus the accumulated execution time

Number of workunits per OSG client

The number of workunits that an OSG client computed from the submission till the removal of the client's job from the pool

Running OSG jobs

The number of running jobs that are accounted by the OSG batch system

Normalized efficiency of OSG pool

Defined as (expavg_credit / 100 / avg(p_fpops) / MAXIMAL_RUNNING_JOBS), this statistics measures the percent of efficiently running machines out of all the machines that were requested by the virtual pool scheduling from OSG submission machine

Active Technion Condor hosts

Measures the number of hosts, belonging to the Technion Condor user, that contacted the BOINC server in past several hours. The Technion Condor hosts are hosts in Technion Condor pool, to which BOINC clients are submitted as usual Condor jobs

Performance of Technion Condor hosts

The statistics is the performance of the Technion Condor user hosts group. Look above for the definition of the hosts group performance

Execution time of Technion Condor clients

The accumulated execution time of Technion Condor clients from the submission till the removal of the client's job from the pool

Turnaround time of Technion Condor clients

The turnaround time of Technion Condor clients from the submission till the removal of the client's job from the pool

Queuing time of Technion Condor clients

The queuing time of Technion Condor clients from the submission till the removal of the client's job from the pool: the turnaround time minus the accumulated execution time

Number of restarts of Technion Condor clients

The number of restarts of Technion Condor clients in Technion Condor pool till the removal of the client's job from the pool

Number of workunits per Technion Condor client

The number of workunits that an Technion Condor client computed from the submission till the removal of the client's job from the pool

Running Technion Condor jobs

The number of running jobs that are accounted by the Technion Condor batch system

Normalized efficiency of Technion Condor pool

Defined as (expavg_credit / 100 / avg(p_fpops) / MAXIMAL_RUNNING_JOBS), this statistics measures the percent of efficiently running machines out of all the machines that were requested by the virtual pool scheduling from Technion Condor submission machine

Incoming job requests

Measures the number of results, returned by all the contacting BOINC clients, per minute

Outgoing job requests

Measures the number of results, sent to all the contacting BOINC clients, in minute. In long term the value must strive to that of incoming job requests, per minute

Rejected jobs with error

Measures the number of results that were rejected by the BOINC backend because of an error on a client side (application crash, dll missing), per minute

Rejected jobs with result over

Measures the number of results that were rejected by the BOINC backend because the result was not needed, already reported or was never sent, per minute

Running jobs time

Measures running times of standalone non-bundled workunits, in seconds

Turnaround jobs time

Measures turnaround times of bundled workunits, that is the time that it takes for a bundle to be sent, computed and assimilated back, in seconds

Results over deadline (invalid)

Measures the number of invalid results that were received from the BOINC clients, per minute. The invalid results are results that do not pass sanity checks, posed by the backend: for instance, probability results must fall into the [0, 1] range, and any result deviating from the range would be classified as invalid

Results over deadline (no reply)

Measures the number of results that were considered by the BOINC backend as expired, per minute

Aborted/overall results ratio

The ratio of aborted results to the total number of results in the database

Error/overall results ratio

The ratio of results that ended with an error on a client side to the total number of results in the database. The blue line designates the level of incorrect results plus results with detached clients

Over deadline/overall results ratio

The ratio of results that were considered by the BOINC backend as expired to the total number of results in the database

Efficient running machines

Measures how much computing power you have currently got in units of machines of a dedicated cluster (where each machine is estimated to be a 1GHz machine), broken down by users of the 'Clusters' team

Usage breakdown

Stacked histogram of breakdown of users, belonging to the 'Clusters' team, according to the number of workunits accomplished. The more workunits were accomplished by an user, the wider its colored share in the stack will be

Errors breakdown

Stacked histogram of breakdown of users, belonging to the 'Clusters' team, according to the number of workunits errored out. The more workunits errored out on the hosts of an user, the wider its colored share in the stack will be

Over deadline breakdown

Stacked histogram of breakdown of users, belonging to the 'Clusters' team, according to the number of workunits whose deadline expired. The more workunits' deadline expired on the hosts of an user, the wider its colored share in the stack will be

Waste

The ratio of results that were computed successfully, but their result was rejected since another host has returned result for the same workunit

Restarts distribution

The graph displays the distribution of workunits in the database according to the number of times, they were restarted on different BOINC clients. 1 restart means there was only one client that successfully and timely calculated the result, corresponding to a workunit. 6 restarts means that the workunit restarted 5 times and finally was rerouted to the local Condor pool due to the inability of the BOINC pool to calculate the result

Duplications breakdown

Stacked histogram of duplications breakdown according to the number of days in the past. The zeroth column is supposed to grow during the current day


Return to Superlink@Technion main page

Copyright © 2012 Computational Biology Laboratory, Technion, Israel