How we think about latency (and what we measure)
Round-trip time is the whole product. A breakdown of the budget we hold ourselves to, from input event to painted frame, and how we report it.
By The SisiDesk Team
Remote desktops have a long history of being tolerable rather than good. The difference between the two is almost entirely latency — specifically, the time between a key press and the frame that shows the result.
The budget
We hold a P95 target of 80 ms glass-to-glass within a user's home region. That budget is spent across input capture, network transit to the point of presence, instance render, encode, transit back, and decode in the browser.
- Input capture and transport: target under 12 ms
- Instance render and encode: target under 25 ms
- Return transit to the nearest point of presence: target under 30 ms
- Decode and paint in the browser: target under 13 ms
Why we publish it
Latency claims are cheap. We surface the measured session round trip inside the session viewer itself, so the number you see is the number you are living with, not a marketing average taken on an empty network at 3 a.m.
When a session degrades, we would rather tell you why — region distance, an encode stall, an unstable local link — than quietly drop quality and let you wonder whether it is your Wi-Fi.