Benchmark Result: Scanning
The Cloud-Edge technology scanned a one (1) billion row, 100-bytes per row, table from disk in 29.619 seconds. The table was conservatively compressed at a 3.7:1 ratio to simulate actual customer data. The benchmark was run on a realistic customer configuration using RAID5 LUNs on a 3Par SAN, connected via standard Brocade 8/24 switches4, resulting in uncompressed throughput of 3.26 Gb/sec.
Modifying the scanning query described above to include a filter (which produced 3.9 million output rows) ran in 29.748 seconds.
Modifying the scanning query described above to sort the entire table and output it to disk, ran in less than 2 minutes (119.167 seconds). The in-memory sort portion of this test was performed in less than 6 seconds.
The Cloud-Edge Sort technology was also able to perform an all-columns scan, from disk, of 1 Billion rows (100 bytes wide), doing complex calculations on every single row (all the columns, not just an index or extract), in 29 seconds. That rate (roughly 33 1⁄2 million rows per second) scales linearly, even if the scan were increased to trillions of rows.
It should be emphasized that Cloud-Edge Sort does not require any of the preprocessing of data during the load phase that many other solutions require. Eliminating the need for optimization of a pre-selected set of access paths.