. +---------------------+
+---|---------------------|-------. .---_||||||_-----------------.
|:--|-----\\\\\//////-----|--. o ---- (\)|
|#|||------ ------|--.|| [ Peter A. H. Peterson |
|-::|-----|.-, ,-.|-----|===========[ Researcher / PhD Candidate |
|o+o|----(| @ | @ |)----|:::::::::::[ 3564 Boelter Hall |
||o||-----# '-' #-----|:::::::::::[ Los Angeles, CA 90095 |
|||||------\"===" /------|===========[ 310.924.2425 |
|||||-------`-----'|------| | o [ pedro@tastytronic.net |
||/ |.----/'\'-----'/\---.|-+-' (/)|
||^_| '''| |'' |-----'''----||||-||||||||||-||||||||||. |
||| +---------------------+-----------' | :|
|:| |:.|
| '------------. [back to research] |.:|
| home | |:.|
'------------. | Representative Synthetic Benchmarking |.:|
.------------' | |:.|
| cv | One can hardly work in the area of operating |.:|
| .------------' systems, security, or power measurement, |:.|
| '------------. without becoming deeply interested in, and |.:|
| publications | perhaps dismayed by, the state of |:.|
'------------. | benchmarking and system testing. One would |.:|
.------------' | like to test systems meant for the real world |:.|
| research | with representative workloads. But what |.:|
| .------------' benchmark workloads are representative of |:.|
| '------------. real users? |.:|
| interests | |:.|
'------------. | Because of the difficulty of these and other |.:|
.------------' | questions, researchers often fall back on |:.|
| blog | microbenchmarks, standardized benchmarks |.:|
| .------------' (such as SPEC), and old standbys such as the |:.|
| '------------. Modified Andrew Benchmark. While each of |.:|
| reading | these benchmarks have their place, the |:.|
'------------. | conventional wisdom is that these workloads |.:|
.------------' | are not representative of how real users use |:.|
| quotes | their computers. Instead, benchmarks like |.:|
| .------------' this were designed primarily to quantify raw |:.|
| '------------. computational power. |.:|
| contact | |:.|
| | Today's laptop is faster than yesterday's |.:|
|''''''''''''''| supercomputer, and another bit of |:.|
| 001010101101 | conventional wisdom is that most personal |.:|
| 011011011001 | computers are underutilized. If that is the |:.|
| 010011011011 | case, then users can benefit -- with little |.:|
| 110011010111 | to no visible performance degradation -- |:.|
| 100100001100 | when this surplus computational power is |.:|
| 101001100111 | used to perform other tasks on behalf of the |:.|
| 001101100101 | user. Software-based full disk encryption |.:|
| 001111000000 | (FDE) is one such example; while it once |:.|
| 001010101101 | seemed completely impractical, it is quite |.:|
| 011011011001 | common and essentially invisible today. Other |:.|
| 010011011011 | potential uses for unused cycles include |.:|
| 110011010111 | speculation, machine learning, compression, |:.|
| 100100001100 | other cryptographic uses, and more. |.:|
| 101001100111 | |:.|
| 001101100101 | These are potentially exciting areas for |.:|
| 001111000000 | research, but without defensible research |:.|
| 001010101101 | showing acceptable performance, it can be |.:|
| 011011011001 | difficult to convince skeptics that the |:.|
| 010011011011 | technique can be practical, even if they |.:|
| 110011010111 | accept its functional value. And without |:.|
| 100100001100 | well-founded "real world workloads," it is |.:|
| 101001100111 | difficult to perform these experiments. I'm |:.|
| 001101100101 | interested in quantifying the "typical user |.:|
| 001111000000 | workload," so that we can use these workloads |:.|
| 001010101101 | (in conjunction with other techniques) to |.:|
| 011011011001 | test experimental systems. |:.|
| 010011011011 | |.:|
| 110011010111 | Additionally, a strong notion of typical user |:.|
| 100100001100 | workloads would also be valuable for power |.:|
| 101001100111 | research, because they could be used to drive |:.|
| 001101100101 | systems for power measurement. Research into |.:|
| 001111000000 | capturing, generating, and using these kinds |:.|
| 001010101101 | of workloads is ongoing. |.:|
| 011011011001 | |:.|
| 010011011011 | [back to research] |.:|
| 110011010111 | |:.|
|'''''''''''''''-------------------------------------------------'::'.
|::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::::|
`=======---------.______.-------:::::::-:------------------------. ? |
---'