About Us
Corporate & Contacts
Products
SPS Phrap
SPS SWAT
SPS Cross_Match
SPS Pfam
SPS Search
Support
Technical Information
Documentation
SPSOFT Phrap Benchmark
The following chart illustrates the performance scaling of SPS Phrap compared to the University of Washington's Phrap v.960731 on a multi-CPU system. This benchmark was run on a SGI Origin 2000 with 16 195mhz CPUs. The data used came from Genome Therapeutics. The data file contained 48815 entries and was 36mb in size.
UW Phrap v.960731 completed the job in 5 hours 39 minutes and 23 seconds. With 16 CPU's, SPS Phrap completed the job in just over 32 minutes.
| Program Used | Relative performance to UW Phrap v.960731 |
| UW Phrap v.960731 | 1 |
| SPS Phrap 1 CPU | 1.3x |
| SPS Phrap 2 CPUs | 2.56x |
| SPS Phrap 4 CPUs | 4.55x |
| SPS Phrap 8 CPUs | 7.4x |
| SPS Phrap 16 CPUs | 10.5x |
It's important to note:
- Performance scaling of SPS Phrap is better with jobs that take a while to run.
- For smaller jobs, it is more efficient to run SPS Phrap with 2 to 4 CPUs.
- SPS Phrap and UW Phrap v.960731 run better on systems with a lot of RAM. This benchmark required about 2gb of RAM.
- Long phrap jobs will consume a lot of RAM. Since UW Phrap v.960731 only runs on 1 CPU, a multi-CPU system can sit mostly idle while phrap uses all the system RAM. SPS Phrap improves system utilization by completing these jobs much faster.
Copyright © 2000 Southwest Parallel Software, Inc.
All Rights Reserved.