The Morning Paper

NOVEMBER 3, 2019

Perhaps the most interesting lesson/reminder is this: it takes a lot of effort to tune a Linux kernel. Google’s data center kernel is carefully performance tuned for their workloads. A micro-benchmark suite, LEBench was then built around tee system calls responsible for most of the time spent in the kernel. Headline results.

Performance 108

Performance Benchmarking Tuning Hardware 108

Brendan Gregg

JUNE 30, 2018

A co-worker introduced me to Craig Hanson and Pat Crain's performance mantras, which neatly summarize much of what we do in performance analysis and tuning. These have inspired me to summarize another performance activity: evaluating benchmark accuracy. If the benchmark reported 20k ops/sec, you should ask: why not 40k ops/sec?

Benchmarking 47

Benchmarking Latency Cache Network 47

Brendan Gregg

JUNE 29, 2018

A co-worker introduced me to Craig Hanson and Pat Crain's performance mantras, which neatly summarize much of what we do in performance analysis and tuning. These have inspired me to summarize another performance activity: evaluating benchmark accuracy. If the benchmark reported 20k ops/sec, you should ask: why not 40k ops/sec?

Benchmarking 40

Benchmarking Latency Cache Network 40

HammerDB

JANUARY 5, 2019

Some opinions claim that “Benchmarks are meaningless”, “benchmarks are irrelevant” or “benchmarks are nothing like your real applications” However for others “Benchmarks matter,” as they “account for the processing architecture and speed, memory, storage subsystems and the database engine.”

Benchmarking 40

Benchmarking Database Social Media Scalability 40

HammerDB

APRIL 19, 2023

Instead, focus on understanding what the workloads exercise to help us determine how to best use them to aid our performance assessment. Benchmarking the target Two of the more popular database benchmarks for MySQL are HammerDB and sysbench. For the experiments in this blog, we did not tune the system. 4.22 %usr 38.40

Performance 55

Performance Benchmarking Cache Storage 55

John McCalpin

JANUARY 22, 2018

The exercise seemed simple enough — just fix one item in the Colfax code and we should be finished. Published DGEMM benchmark results for the Xeon Phi 7250 processor ( [link] ) show maximum values of about 2100 GFLOPS when using all 68 cores (a very approximate estimate from a bar chart). Instead, we found puzzle after puzzle.

Latency 40

Latency Hardware Code Testing 40

John McCalpin

JANUARY 22, 2018

There was no deep goal — just a desire to see the maximum GFLOPS in action. The exercise seemed simple enough — just fix one item in the Colfax code and we should be finished. of the “adjusted peak performance”, there is no longer a significant upside to performance tuning.

Latency 40

Latency Hardware Code Testing 40