fnet-h1

Experimental · Go · engine: fnet · standard configuration

fnet v1.0.0's fhttp HTTP/1.1 server: fnet's epoll event loops find where each request message ends and enqueue it, then the standard library parses it and runs an http.Handler on a task-pool worker, so there is no goroutine per connection. Responses are buffered whole. A request whose body is not buffered (larger than 1MB, chunked, or framed unusually) is handed to an internal net/http server, which closes that connection after the one request. fnet's ReusePort option is on: on Linux every sub-reactor listens on :8080 with a socket of its own, sharing the port through SO_REUSEPORT so the kernel spreads the connections over the loops. A second fhttp.Server on :8081 speaks HTTPS through fnet's ftls (crypto/tls inside the connection callbacks, ALPN http/1.1) for the TLS profiles. Static files are read per request from the mounted directory, serving the pre-compressed .br/.gz siblings when the client accepts them; /async-db and the crud routes use pgx's pool sized from DATABASE_MAX_CONN, with Redis as a cache-aside store for item reads where REDIS_URL is provided.

Composite rank

Each profile of a family is worth 100 to the entry that leads it, and the composite is the sum over the family, less 2.5% for each of routing, middleware, request and response the entry does not do for you - its completeness factor, which the WebSocket and gRPC families do not carry. The field is this entry's own league: engines and reverse proxies are scored apart from frameworks. How it works.

Family	Rank	Composite
HTTP/1.1	13 of 22	4585

Every result

Alpha Round, 13 runs. Requests per second is the best of three; latency, CPU and memory come from that run.

Category	Profile	Conns	Req/sec	Avg	p99	CPU	Memory
Connection	Baseline	4,096	2,404,492	1.70ms	8.09ms	6605.8%	156MiB
Connection	Pipelined	4,096	13,726,828	4.77ms	21.70ms	6470.2%	228MiB
Connection	Short-lived	4,096	2,010,211	2.02ms	9.10ms	6363.5%	301MiB
Concurrency	Async Delay	32,000	1,494,896	21.21ms	41.20ms	5125.4%	841MiB
Efficiency	Latency-10K	1,024	9,981	84.5us	124.0us	40.7%	86MiB
Efficiency	Latency-1M	1,024	997,258	151.4us	518.0us	3238.3%	132MiB
Efficiency	Latency-500K/8	1,024	404,289	1896352.2us	4004864.0us	755.5%	33MiB
Workload	JSON Comp	4,096	230,866	17.56ms	248.80ms	5403.3%	560MiB
Workload	JSON Comp	16,384	228,590	67.68ms	701.70ms	5222.6%	596MiB
Workload	JSON TLS	4,096	457,968	29.67ms	801.33ms	5031.3%	275MiB
Workload	8Gbit	512	49,243	107.1us	171.0us	294.2%	126MiB
Workload	Static TLS	1,024	244,877	27.36ms	564.47ms	4176.1%	298MiB
Database	Async DB	1,024	116,347	8.28ms	20.60ms	4353.7%	236MiB

Compared with

The Go entries around fnet-h1 on the HTTP/1.1 composite, and the one leading them. The last column shows every profile side by side.

Entry	Composite	vs fnet-h1
fib-tuned	6061	+32%	head to head
fiber	5637	+23%	head to head
fib	5439	+19%	head to head
go-fasthttp	3505	-24%	head to head
echo	2118	-54%	head to head

Compare fnet-h1 with any Go entry · Compare fnet-h1 with any entry, in any language · Every Go entry, ranked

Results updated 2026-11-01 · How the benchmark is run · The machine · Add or fix an entry