Top500: Top 500 Datasets

Description Format Details References

Description

All Top500 lists are available up to the release version of the package. For example, package 14.06 contains all Top500 datasets up to June of 2014. Top500 data is released twice a year, every June and November, at the ISC and SC conferences, respectively.

Format

All Top500 datasets are stored as dataframes, essentially being then output of XML::xmlToDataFrame. The raw datasets are available at the official Top500 website, and are mirrored in the GitHub repository of the RTop500 project (along with the parser used by the package).

Details

The naming convention for all datasets goes TOP500_<year><month>, where the year is the 4-digit year and month is the 2-digit month that the list was released (either 06 or 11).

rank Rank of the system based on performance of the Linpack benchmark.
system.id A unique (for Top500 lists) system identifier.
system.name Name of the machine.
manufacturer System vendor.
computer Short description of the nodes.
system.address Machine website.
r.max Max Linpack benchmark performance in GFLOPs.
power Power in megawatts with machine at peak.
r.peak Theoretical peak performance in GFLOPs.
n.max Problem size for Linpack benchmark to achieve given r.max value.
n.half Problem size for Linpack benchmark to achieve half of the given r.max value.
installation.site Location of the machine.
town Town machine is located in.
state State machine is located in (if applicable).
country Country machine is located in.
year Year machine was created.
area.of.installation
number.of.processors Number of processors/cores.

Some older lists don't have all of these fields.

References

Official Top500 website: http://www.top500.org/

Raw Top500 data mirror: https://github.com/wrathematics/RTop500/tree/master/raw


shinra-dev/RTop500 documentation built on May 29, 2019, 9:25 p.m.