Check out the new USENIX Web site. next up previous
Next: CPU utilization Up: Macro-benchmarking Experiments Previous: TPC-C and TPC-H Results


Other Benchmarks

We also used several simple macro-benchmarks to characterize the performance of iSCSI and NFS. These benchmarks include extracting the Linux kernel source tree from a compressed archive (tar xfz), listing the contents (ls -lR), compiling the source tree (make) and finally removing the entire source tree (rm -rf). The first, second and fourth benchmarks are met-data intensive and amenable to meta-data caching as well as meta-data update aggregation. Consequently, in these benchmarks, iSCSI performs better than NFS v3. The third benchmark, which involves compiling the Linux kernel, is CPU-intensive, and consequently there is parity between iSCSI and NFS v3. The marginal difference between the two can be attributed to the impact of the iSCSI protocol's reduced processing length on the single-threaded compiling process.


Table 8: Completion times for other benchmarks.
Benchmark NFS v3 iSCSI
tar -xzf 60s 5s
ls -lR $ >$ /dev/null 12s 6s
kernel compile 222s 193s
rm -rf 40s 22s


next up previous
Next: CPU utilization Up: Macro-benchmarking Experiments Previous: TPC-C and TPC-H Results
2004-02-10