This changes the files table to use normalisation for the names and
versions. The idea is that these are often common between all remote
devices, and repeating an integer is more efficient than repeating a
long string. A new benchmark bears this out; for a database with 100k
files shared between 31 devices, with some worst case assumption on
version vector size, the database is reduced in size by 50% and the test
finishes quicker:
Current:
db_bench_test.go:322: Total size: 6263.70 MiB
--- PASS: TestBenchmarkSizeManyFilesRemotes (1084.89s)
New:
db_bench_test.go:326: Total size: 3049.95 MiB
--- PASS: TestBenchmarkSizeManyFilesRemotes (776.97s)
The other benchmarks end up about the same within the margin of
variability, with one possible exception being that RemoteNeed seems to
be a little slower on average:
old files/s new files/s
Update/n=RemoteNeed/size=1000-8 5.051k 4.654k
Update/n=RemoteNeed/size=2000-8 5.201k 4.384k
Update/n=RemoteNeed/size=4000-8 4.943k 4.242k
Update/n=RemoteNeed/size=8000-8 5.099k 3.527k
Update/n=RemoteNeed/size=16000-8 3.686k 3.847k
Update/n=RemoteNeed/size=30000-8 4.456k 3.482k
I'm not sure why, possibly that query can be optimised anyhow.
Signed-off-by: Jakob Borg <jakob@kastelo.net>
88 lines
2.1 KiB
Go
88 lines
2.1 KiB
Go
// Copyright (C) 2014 The Syncthing Authors.
|
|
//
|
|
// This Source Code Form is subject to the terms of the Mozilla Public
|
|
// License, v. 2.0. If a copy of the MPL was not distributed with this file,
|
|
// You can obtain one at https://mozilla.org/MPL/2.0/.
|
|
|
|
//go:build !solaris && !windows
|
|
// +build !solaris,!windows
|
|
|
|
package main
|
|
|
|
import (
|
|
"fmt"
|
|
"os"
|
|
"runtime"
|
|
"syscall"
|
|
"time"
|
|
|
|
"github.com/syncthing/syncthing/lib/build"
|
|
"github.com/syncthing/syncthing/lib/locations"
|
|
"github.com/syncthing/syncthing/lib/osutil"
|
|
"github.com/syncthing/syncthing/lib/protocol"
|
|
"golang.org/x/exp/constraints"
|
|
)
|
|
|
|
func startPerfStats() {
|
|
go savePerfStats(fmt.Sprintf("perfstats-%d.csv", syscall.Getpid()))
|
|
}
|
|
|
|
func savePerfStats(file string) {
|
|
fd, err := os.Create(file)
|
|
if err != nil {
|
|
panic(err)
|
|
}
|
|
|
|
var prevTime time.Time
|
|
var curRus, prevRus syscall.Rusage
|
|
var curMem, prevMem runtime.MemStats
|
|
var prevIn, prevOut int64
|
|
|
|
t0 := time.Now()
|
|
syscall.Getrusage(syscall.RUSAGE_SELF, &prevRus)
|
|
runtime.ReadMemStats(&prevMem)
|
|
|
|
fmt.Fprintf(fd, "TIME_S\tCPU_S\tHEAP_KIB\tRSS_KIB\tNETIN_KBPS\tNETOUT_KBPS\tDBSIZE_KIB\n")
|
|
|
|
for t := range time.NewTicker(250 * time.Millisecond).C {
|
|
syscall.Getrusage(syscall.RUSAGE_SELF, &curRus)
|
|
runtime.ReadMemStats(&curMem)
|
|
in, out := protocol.TotalInOut()
|
|
timeDiff := t.Sub(prevTime)
|
|
|
|
rss := curRus.Maxrss
|
|
if build.IsDarwin {
|
|
rss /= 1024
|
|
}
|
|
|
|
fmt.Fprintf(fd, "%.03f\t%f\t%d\t%d\t%.0f\t%.0f\t%d\n",
|
|
t.Sub(t0).Seconds(),
|
|
rate(cpusec(&prevRus), cpusec(&curRus), timeDiff, 1),
|
|
(curMem.Sys-curMem.HeapReleased)/1024,
|
|
rss,
|
|
rate(prevIn, in, timeDiff, 1e3),
|
|
rate(prevOut, out, timeDiff, 1e3),
|
|
osutil.DirSize(locations.Get(locations.Database))/1024,
|
|
)
|
|
|
|
prevTime = t
|
|
prevRus = curRus
|
|
prevMem = curMem
|
|
prevIn, prevOut = in, out
|
|
}
|
|
}
|
|
|
|
func cpusec(r *syscall.Rusage) float64 {
|
|
return float64(r.Utime.Nano()+r.Stime.Nano()) / float64(time.Second)
|
|
}
|
|
|
|
type number interface {
|
|
constraints.Float | constraints.Integer
|
|
}
|
|
|
|
func rate[T number](prev, cur T, d time.Duration, div float64) float64 {
|
|
diff := cur - prev
|
|
rate := float64(diff) / d.Seconds() / div
|
|
return rate
|
|
}
|