1
Brendan Gregg
Brendan Gregg's homepage: I use this site to share various things, mostly my work with computers, specifically, computer performance analysis and methodology.
- Popular posts
- 41
- Total points
- 13,732
- HN comments
- 4,266
- Rank
- #53 personal
Visit brendangregg.com → On Hacker News May 2014 – Feb 2026
Systems & Performance 32Tools & Productivity 20Career & Management 9Tech Industry & Business 8Infrastructure & DevOps 4AI & Machine Learning 2
2
3
4
Linux Load Averages: Solving the Mystery
Linux load averages explained, including why they include the uninterruptible I/O sleep state.
5
6
7
DTrace for Linux 2016
DTrace for Linux 2016: announcing a major milestone: the final kernel capabilities have merged in Linux 4.9-rc1 to enhanced BPF (eBPF) to provide an advanced programmable dynamic tracer similar to DTrace.
8
9
10
11
TCP Tracepoints have arrived in Linux
Linux TCP Tracepoints, and using them via bcc/eBPF.
12
13
Bpftrace for Linux 2018
bpftrace (DTrace 2.0) for Linux 2018: the bpftrace front-end
14
BPF Performance Tools: Linux System and Application Observability
BPF Performance Tools: Linux System and Application Observability (book)
15
Where has my disk space gone? Flame graphs for file systems
Using the flame graph visualization to determine where file system or disk space is consumed, and answer where space has gone.
16
Meltdown Initial Performance Regressions
KPTI/KAISER Meltdown Initial Performance Regressions
17
18
19
20
Computing Performance on the Horizon
USENIX LISA2021 Computing Performance: On the Horizon
21
22
23
AWS EC2 Virtualization 2017: Including Nitro
AWS EC2 Virtualization 2017: explaining the different virtualization types, from emulation and binary substitution, paravirtualization and Xen, PV, HVM, and PVHVM modes, and the new Nitro hypervisor
24
25
26
27
28
OS X 10.9.3 Is Toxic
OS X 10.9.3 Recurring Panics
29
30
Unikernel Profiling from dom0
Unikernel Profiling: Flame Graphs from dom0. This post investigates performance profiling of Unikernels, and with a proof of concept shows that with a little work a lot of observability is possible.