1. Only what you look at is calculated
Exclusion TABs is on from the first start. The summary view is always available, and every other view (latency, throughput, samples, errors, degradation, trend, modal and table analysis, compare and complements) stays switched off until you enable it.
A view that is switched off is not loaded at all: it does not aggregate the samples, does not compute its percentiles, does not draw its charts and does not keep its series in memory. That is where most of the work of an analysis goes, so most of the saving comes from here.
2. A sensible ceiling on series
Every file opens with a ceiling of 1,200 user levels and 25 labels per chart. 1,200 is the maximum number of virtual users of our own runner, so a run made with i9x is never trimmed. You can raise the ceiling up to 10,000 users and 100 labels, and the sidebar tells you how many are being left out.
To be precise: the file is still read in full, because the global figures must include every sample. The ceiling reduces how many series are built, kept in memory and drawn, which on large files is the slowest part.
3. Your device, not a data centre
- Your JTL/CSV is processed on the computer you are already using and is not uploaded (how to check it). Hundreds of megabytes are not sent over the network and no copy of them is stored on our side.
- When you save a run, your account keeps its aggregated KPIs, not the individual samples: a small fraction of the size of the file.
4. No servers on standby
The website and the account service run on serverless functions: they use resources while answering a request and nothing in between. We do not keep our own machines switched on around the clock waiting for traffic.
5. Habits that save even more
- Switch on only the views you need for the question you are answering, and switch them off again when you are done.
- Keep the Exclusion TABs master switch on: when it is off, every view is treated as enabled and processed.
- Narrow the time range to the part of the test you care about, such as the steady state.
- Raise the Top N ceiling only when you really need more series on a chart.
- In Compare, select only the files you want to compare.
6. Calculate the impact of fixing defects
Put in your own numbers: the hours your team stops spending on building charts by hand, and the energy, heat and CO2e your infrastructure stops producing when a performance defect is fixed and the same load needs fewer servers.
Engineering hours
Time spent building charts by hand from a results file and searching for the cause of each finding, against the time it takes with i9x.
Energy and heat
i9x does not save energy by itself: fixing the defects it helps you find does. Compare before and after the fix, and enter how many servers the same load needs.
The starting values are an example: change them for yours. These are estimates with your data, not measurements of i9x customers. Energy follows the energy × carbon-factor part of the SCI standard (ISO/IEC 21031:2024), without the emissions of manufacturing the hardware; 1 W of a server is 3.412 BTU/h of heat. The grid factor of 0.44 is approximate for Mexico: use the official one for your country.
7. What we do not claim
- We have not measured the carbon footprint of i9x Performance and hold no environmental certification. This page describes design decisions, not an audited result.
- We do not claim that i9x Performance is carbon neutral.
- How much is saved depends on the size of your file, the views you enable and your device. The energy behind our hosting depends on the policies of our providers.
Ideas to make i9x lighter are welcome: support@i9xperformance.com.