Within the BOINC platform for volunteer computing, the BOINC Credit System helps volunteers keep track of how much CPU time they have donated to various distributed computing projects. The credit system is designed to avoid cheating by validating results before granting credit on projects. This ensures users are returning accurate results for both scientific and statistical reasons.
Contents |
Online distributed computing relies heavily, if not completely, on volunteer computers. For this reason, projects such as SETI@home and other BOINC projects depend on a complicated balance among long-term users and the cycle of new users and retiring users.
The basis for the BOINC credit system is the cobblestone named after Jeff Cobb of SETI@home. The basis of the system is the concept that 200 cobblestones would be claimed for one day of work on a computer with the following specifications:
The actual computational difficulty needed to run a given work unit is the basis for the number of credits that it should be granted. The BOINC system allows for work of any length to be processed and have a user claim identical amounts of credit.
To achieve this BOINC uses benchmarks to measure the speed of a system and in combination with the amount of time it required for a work unit to process can “guess” at the amount of credit it should receive. Since systems have many variables including the amount of RAM, the processor speed, and specific architectures of different motherboards and CPUs, there can be wide discrepancies in the number of credits that different computers believe that each work unit requires to process.
Most projects require a consensus be reached by having multiple hosts return the same work unit. If they all agree then the credit is calculated and all hosts receive the same amount regardless of what they asked for. Each project can use their own policy depending on what they see is best for their specific needs. In general the top and bottom claimed credits are dropped and an average of the remaining is taken.
Credits are tracked internally for computers, users, and teams. When a computer processes and returns a work unit it receives no credit. It must first have that work unit validated by the project specific method. Once validated the computer is granted credit, which can be less than, equal to, or greater than what was requested. This amount is immediately added to the computer, user, and team total. If a work unit is returned past the given deadline or is found to be inaccurate it is marked as invalid and zero credits will be granted. Users and teams commonly determine world rank by comparing the total number of credits accumulated. This highly favors users and teams that have been around for the longest time and makes it extremely difficult for new users to rapidly gain ground in the rankings even if they are running many computers. That being said, given the exponential increase in computing power of the average PC, it is relatively easy to surpass inactive BOINC users, who have earned all of their points on obsolete machines—even if they were at one time ranked highly. Thus, on average, the highest ranked BOINC users will be the ones who are actively crunching.
To calculate the useful amount of work provided by a computer, a special calculation called Recent Average Credit (RAC) is used. This calculation is designed to estimate the number of credits a computer, user, and team will accumulate on an average day. Due to the many variables not taken into account including the inconsistency of host processing, time it takes to validate work units, discrepancies in benchmarks, and possible project down time, the RAC calculation proves to be only a guide. Additionally RAC is independent of computers, users, and teams, meaning they cannot be simply added up. RAC was originally meant to help scientists understand the computational power available to them and to increase competition among users by allowing even new users to quickly move up in rank based on RAC, which should directly reflect how fast work is being processed.
BOINC projects export statistical information in the form of XML files and make it available for anyone to download. Many different third party statistics websites have been developed to track the progress of BOINC projects. The statistics track computers, users, teams, and countries within individual projects and across many projects. Many different sites provide summary graphics, which can be used on web pages that automatically update to contain the statistical information for the specified user or team.
|
||||||||||||||||||||