Daily ISP comparison

Sunday 23 August 2026

Latency and packet loss by ISP — nbn FTTP, one NTD, four UNI-D ports.

Midnight to midnight, Australia/Perth · Browse all reports

Scorecard

The day at a glance

Ranks are ordinal across the 12 global endpoints. Lower is better.

Telstra

UNI-D1
Packet loss
0.0484%
Coverage
100%
Avg latency rank
2.00
Endpoints won
0 / 12
Full report

Ozot

UNI-D2
Packet loss
0.0467%
Coverage
100%
Avg latency rank
1.00
Endpoints won
12 / 12
Full report

Aussie Broadband

UNI-D3

Not yet deployed

Superloop

UNI-D4

Not yet deployed

How to read this

Same building, same fibre, same second

Every ISP in this report is plugged into the same nbn FTTP connection box at the same address — Telstra on UNI-D1 and Ozot on UNI-D2. One fibre, one NTD, four ports.

Each port has an identical Raspberry Pi behind it, running identical software, sending identical test traffic to the same 12 servers around the world, at the same moment. Nothing about the building, the fibre or the equipment differs between them.

That means every difference you see below comes from the ISP’s own network — their backhaul, their peering, and how congested they get — and nothing else.

Latency
How long a packet takes to reach the far server and come back, in milliseconds. Lower is better. Min is the best it ever got, Avg is what it was like most of the time, and Max is the single worst packet in the whole 24 hours.
Packet loss
The share of test packets that never came back at all. 0% is perfect. Loss is what makes a video call freeze or a game rubber-band, and even a fraction of a percent is noticeable when it arrives in a burst.
Jitter
How much the latency wobbles from packet to packet. Steady is better than fast-but-erratic for calls and video.
Coverage
How much of the 24 hours we actually have measurements for. Below 100% means that connection stopped reporting — which we show separately, so an outage can never be mistaken for a clean result.
Head to head

Latency by endpoint

Minimum, average and maximum round-trip time over the full 24 hours. The lowest average for each endpoint is marked.

EndpointTelstraOzot
MinAvgMaxMinAvgMax
Perth2.9ms10.9ms458.3ms1.8ms2.5ms101.7ms
Sydney49.0ms59.1ms618.7ms43.0ms43.8ms957.9ms
Singapore282.6ms294.2ms396.3ms48.3ms50.1ms1,026.5ms
Seoul188.7ms197.5ms356.4ms127.6ms128.6ms1,010.8ms
Tokyo163.8ms173.2ms423.1ms129.9ms130.6ms205.0ms
Los Angeles192.0ms200.2ms293.8ms180.4ms181.1ms325.3ms
Chicago241.0ms248.9ms365.5ms219.2ms220.0ms390.1ms
Paris323.6ms332.7ms426.5ms197.3ms203.4ms599.7ms
Frankfurt331.0ms340.7ms442.7ms199.2ms207.5ms661.2ms
Amsterdam338.0ms347.8ms923.7ms202.9ms211.6ms874.7ms
London323.3ms331.5ms425.4ms200.4ms209.5ms654.6ms
Stockholm355.0ms364.2ms465.4ms220.7ms229.3ms621.8ms

Packet loss by endpoint

The share of probe packets that never came back, over the full 24 hours. Lower is better; the best result for each endpoint is marked.

EndpointTelstraOzot
Perth0.0041%0.0083%
Sydney0.0053%0.0409%
Singapore0.075%0.0682%
Seoul0.0418%0.0733%
Tokyo0.2659%0.0108%
Los Angeles0.0378%0.0112%
Chicago0.0065%0.0167%
Paris0.0089%0.1486%
Frankfurt0.0081%0.0306%
Amsterdam0.0346%0.008%
London0.0339%0.0981%
Stockholm0.0594%0.0456%
Over the day

Latency by endpoint, all ISPs

Median latency per five-minute bucket, in milliseconds. The vertical axis starts at zero on every chart. Spikes are reported in the p95/p99 table and the loss timeline rather than allowed to distort the trend.

Perth
051014
TelstraOzot
Sydney
0234668
TelstraOzot
Singapore
0133266399
TelstraOzot
Seoul
074149223
TelstraOzot
Tokyo
083166248
TelstraOzot
Los Angeles
076151227
TelstraOzot
Chicago
095190285
TelstraOzot
Paris
0125250375
TelstraOzot
Frankfurt
0128256385
TelstraOzot
Amsterdam
0131261392
TelstraOzot
London
0125249374
TelstraOzot
Stockholm
0137274411
TelstraOzot
Loss timeline

Where the loss actually happened

Each cell is five minutes. Grey means no data was received, which is reported separately from zero loss.

No loss< 0.01%0.01 – 0.1%0.1 – 1%> 1%No data

Telstra UNI-D1

Perth
Sydney
Singapore
Seoul
Tokyo
Los Angeles
Chicago
Paris
Frankfurt
Amsterdam
London
Stockholm
00:0006:0012:0018:0024:00

Ozot UNI-D2

Perth
Sydney
Singapore
Seoul
Tokyo
Los Angeles
Chicago
Paris
Frankfurt
Amsterdam
London
Stockholm
00:0006:0012:0018:0024:00
Methodology

How these numbers were produced

Every ISP in this report is measured on identical hardware, with an identical probe configuration, through the same nbn FTTP NTD at the same premises, over the same time window. Each ISP terminates on its own UNI-D port.

  • ProbeUDP to port 1337, 10 packets per second per endpoint, 1000ms timeout.
  • BucketCounters aggregated in 60-second buckets.
  • LossPacket loss is the sum of reset-safe counter deltas. Buckets where a counter decreases are treated as an agent restart and excluded rather than counted as loss.
  • CoverageCoverage is delivered samples divided by expected samples. Missing data is reported as a gap, never as zero loss.
  • Percentilesp50/p95/p99 are computed over the full day of samples, not averaged from buckets. Approximate (t-digest) quantiles.
  • TimeoutsA sample with latency <= 0 or null is a timeout.
  • RankingEndpoints are ranked ordinally. No weighted composite score is used.
  • Ingest delay75s ingest settling delay applied before a day is generated.

Scope. This is a single-premises measurement against one nbn point of interconnect. It reflects each ISP’s backhaul, peering and congestion as experienced at this location on this date. It is not a nationwide comparison, and results at other locations may differ. All four UNI-D ports share the same physical fibre and GPON access path, so access-network conditions affect every ISP equally and cancel out of the comparison.

Generated 2026-08-24T00:15:31+08:00. The complete underlying dataset for this report is published as JSON.