Full logs vs sampled logs

A managed log intake bills each line. Keeping every line costs 20,824 USD/mo in this test. A sampler keeps the error lines and a share of the info lines, and costs 3,264 USD/mo. It stays under the 4,600 USD/mo limit, keeps 99 percent of errors and passes at least 400 info lines/s.

Play the level

A graph for each resourceTap to add or remove the probes

One CPU graph · CPU 40% · disk no graph · the fault is hidden

A faster runtime needs fewer hostsArrow keys change the load. Tap or Space changes the runtime

1,500 req/s · Node API · 3 hosts · 18 USD/mo

Case 1

A fleet sends 20,000 info lines/s and 200 error lines/s. For 30 s, errors reach 3,000 lines/s. Keep 99 percent of errors and 400 info lines/s, and stay under 4,600 USD/mo.

Case 1
Measure Goal Keep every line Sample info lines
Error rate (Errors) under 1% 0% Pass 0% Pass
Throughput (Info) at least 400 req/s 8,560 req/s Pass 819 req/s Pass
Monthly cost under 4,600 USD/mo 20,824 USD/mo Fail 3,264 USD/mo Pass
Monthly cost 20,824 USD/mo 3,264 USD/mo
p99 latency 15 ms 15 ms
HTTP/2HTTP/2HTTP/2HTTP/2HTTP/2Partner appsPartner appsRate limiter1 vCPU · 1 GBRate limiter1 vCPU · 1 GBManaged que…2 vCPU · 1 GBLog store2 vCPU · 2 GB
HTTP/2HTTP/2HTTP/2HTTP/2HTTP/2Partner a…Partner a…Rate limi…1 vCPU · 1 GBRate limi…1 vCPU · 1 GBManaged q…2 vCPU · 1 GBLog store2 vCPU · 2 GB
Keep every line: misses a goal
HTTP/2HTTP/2HTTP/2HTTP/2HTTP/2Partner appsPartner appsRate limiter1 vCPU · 1 GBRate limiter1 vCPU · 1 GBManaged que…2 vCPU · 1 GBLog store2 vCPU · 2 GB
HTTP/2HTTP/2HTTP/2HTTP/2HTTP/2Partner a…Partner a…Rate limi…1 vCPU · 1 GBRate limi…1 vCPU · 1 GBManaged q…2 vCPU · 1 GBLog store2 vCPU · 2 GB
Sample info lines: meets the goals