Looking at https://dealbot.fwss.io/
What stands out to me as worth investigating :
- Daily RSR is almost always 100% for direct retrievals. CDN retrievals had a poor RSR on several days, with 0% RSR on September 8 and 34% on September 18.
- Retrieval latency via CDN was much higher than for direct retrievals on Sep 12 (3.7s vs 0.9s) and Sep 3 (6.7s s 2.1s).
Do we know why our RSR is sometimes worse than direct retrievals? I would expect us to always be better than direct retrievals thanks to caching. Is there a bug on our side? Was there an outage on our side that we did notice? Similarly for TTFB.
I want us to have enough visibility into our performance so that when people look at the dealbot dashboard and notice that Beam has worse performance than direct retrievals, then we have a good explanation for why that happened.
RSR
TTFB
Looking at https://dealbot.fwss.io/
What stands out to me as worth investigating :
Do we know why our RSR is sometimes worse than direct retrievals? I would expect us to always be better than direct retrievals thanks to caching. Is there a bug on our side? Was there an outage on our side that we did notice? Similarly for TTFB.
I want us to have enough visibility into our performance so that when people look at the dealbot dashboard and notice that Beam has worse performance than direct retrievals, then we have a good explanation for why that happened.
RSR
TTFB