 |
» |
|
|
|
The following are the packages originally developed by the HP Labs
storage system department and made available as open source. Feedback
on this software can be sent to software (at) cello.hpl.hp.com. If
you send patches, please indicate the Revision from the Release-Info
file of recent releases or the filename for older releases.
Lintel is the support package for a large number of other SSD
packages. It includes library code and utilities that are used in
multiple other packages, and are not big/cohesive enough to be treated
as a separate package. Some of the features included in Lintel are:
- Statistics class hierarchy -- a collection of classes that
compute statistics in different ways (e.g. mean+stddev, exponentially
moving average, quantiles). All the classes use the same interface
making it easy to replace statistics.
- Improved datastructures (HashTable, Deque, PriorityQueue) --
improved versions of the standard C++ data structures either offering
improved performance or more functionality.
- mercury-plot -- a script that ties together an SQL database
with a gnuplot plotting engine making it relatively easy to explore
data.
- batch-parallel -- a script for running large collections of
jobs either on a single node or a batch cluster. Jobs can be
calculated either as a set of files creating outputs, or as a set of
options describing the jobs.
DataSeries is our toolset for manipulating large datasets. It uses a
data model similar to SQL. Data is grouped by types (ala SQL tables)
that consist of a number of rows. Each row consists of a number of
fields (ala SQL columns) of various types. DataSeries includes the
C++ interface to quickly read and analyze DataSeries files. We have
used it to store and analyze datasets containing more than 100 billion
rows. We have pre-calculated a snapshot of the tr (pdf) found
in the doc/tr directory. We have made some of our data available for analysis
at http://apotheca.hpl.hp.com/pub/datasets/. The SNIA IOTTA
working group also has data
available at http://iotta.snia.org/traces/
|
