System optimization
Profiling and removing the bottlenecks in both performance and cloud spend.
- Engagement
- Fixed-fee diagnosis with the re-measure included
- Typical timeline
- 3–4 weeks, including the re-measure
- Starts with
- Permission to switch on tracing in production for one week.
A report that used to take four seconds now takes ninety, and the database has been made bigger twice to hide it. Extra hardware buys a few months and raises the bill. Nobody has yet measured what the system is actually waiting for.
Faster systems on a smaller bill.
What we build
Concrete artefacts, handed over and documented.
- 01
Traces and profiles captured under real traffic, so the diagnosis rests on evidence rather than on suspicion
- 02
A ranked list of findings, each carrying its current measurement, the expected gain and the effort to fix it
- 03
Fix specifications your own engineers can implement, written at the level of the query, the index or the call
- 04
A paired session in which your team lands the first two fixes while we watch the numbers move
- 05
A re-measurement a fortnight later, comparing like with like, published whether or not it flatters us
What changes
The slow path is identified with a number attached to it instead of a theory
Hardware bought to mask the problem can usually be given back
Your engineers keep the method, so they catch the next regression without us
How it runs
01Instrument
Tracing and query statistics switched on in production, then a week of real traffic worth looking at.
02Diagnose and specify
Findings ranked, and the two highest-value fixes written up and paired on with whoever will maintain them.
03Re-measure
The same measurements repeated once the fixes are live, with the difference stated plainly in one page.
Chosen per project. Named here so you can see the shape of it.
- OpenTelemetry
- Grafana
- Sentry
- k6
Questions we get asked
Related work
Tell us the outcome, not the tooling.
Send us the situation you are in. We will tell you which discipline it belongs to, what we would do first and what it costs, including when the answer is to wait.