Route Party Buy the route, not the specs

Testing a VPS Route: How to Measure at Peak Hour

Vendor speed charts are shot during the best hours. One test between 20:00 and 23:00 tells you whether this box is your workhorse or a toy.

5 min read · updated 2026-10-25

Peak hour CN2 GIA Unicom 9929 Japan SoftBank 163 backbone 0 80 160 240ms 00:0006:0012:0018:0024:00 Vertical axis: latency (illustrative shape, not measured data)
Illustration: a flat line and a line that spikes at peak hour differ by hours, not by percent. The vertical axis is latency; the shape is illustrative, not measured.

Why 20:00–23:00 specifically

Peak hour is when international egress is most congested, and the 163 backbone is exactly where a route turns from “fast by day” into “a lottery at peak hour”. A number measured at 3 a.m. and a number measured at peak hour are not describing the same machine.

Test any VPS at least three times: once in the daytime, once at peak hour, once late at night. Put the three numbers side by side and you have the machine’s real profile.

Four commands that give you the outline

These work on Debian or Ubuntu. Install the tools first with: apt update && apt install -y mtr traceroute

# 1) Latency and packet loss to a public DNS server (the most direct)
ping -c 20 1.1.1.1

# 2) Trace the route (outbound)
traceroute -n 1.1.1.1

# 3) Trace the return path — the single most important step
#    The return path is what you feel when streaming and downloading
mtr -rwzbc 200 1.1.1.1

# 4) Overall benchmark (CPU / disk / network)
curl -sL yabs.sh | bash

How to read mtr output

What you seeWhat it meansWhat to do
59.43-prefixed node on the return pathFull CN2, i.e. GIAUsable as your main box
59.43 mixed with other segmentsHalf CN2, i.e. GTCheck peak-hour loss before committing
Only 202.97 on the return pathOrdinary 163 backboneNever pay a premium for it
One hop shows high loss but later hops are fineThat router ignores ICMPNormal — read the next hop
Final hop loses more than 5%The link really is dropping packetsDo not use it for public traffic

Save the baseline

Save your first mtr and yabs results — a screenshot or a file, either is fine. The next time you feel it has “slowed down”, you will have something to compare against. Without a baseline you are only guessing, and at peak hour a feeling is the least reliable instrument there is.

Routes mentioned

See the vendor shortlist →