Skip to content

Methodology · Carrier Ethernet turn-up

ITU-T Y.1564: every EVC on the circuit, checked against its SLA before the first customer frame

Y.1564 (EtherSAM) activates an Ethernet circuit in two passes. The configuration test steps each service to 25, 50, 75 and 100 % of its CIR, then to CIR + EIR and above it to prove the policer. The performance test runs all services at CIR together for 15 minutes, 2 hours or 24 hours and records IR, FTD, FDV, FLR and availability per service against the SLA thresholds. A 1 GbE UNI with three services takes under 20 minutes to configure and one long run to sign off.

  • Two tests in one run: configuration service by service, then performance with every service at CIR at the same time
  • Six steps per service: 25, 50, 75 and 100 % of CIR, CIR + EIR, and a policing step above the line
  • Five metrics per service per second: IR, FTD, FDV, FLR and availability, each with its own pass threshold from the SLA
Get a Y.1564 test plan Jump to the six steps
SERVICE CONFIGURATION TEST IR PER STEP · Y.1564 §8.1 STEPS TO CIR CIR + EIR OFFERED ABOVE THE LINE, POLICED CIR + EIR CIR 25% 50% 75% 100% +EIR +25% STEP 1 STEP 2 STEP 3 STEP 4 EIR POLICING Four steps up, one step over the line, then the long run

Configuration test of one service: four steps to CIR, one step with EIR, one step above the line to prove policing

Where it fits

Turn-up of a circuit that already carries an SLA

RFC 2544 proves a device. Y.1564 proves a delivered service, with the numbers the contract names, in a fraction of the time

Carrier Ethernet hand-over

E-Line and E-LAN services from a provider or a wholesale partner, before the customer is switched on. The report is the acceptance document for both sides

Several classes on one port

Voice, video and data on one UNI, each with its own CIR, EIR and thresholds. Y.1564 loads them together, the way the customer will

After a change, before the ticket

New policer, new QoS map, new path in the aggregation. A two-hour run shows whether the SLA still holds, with the same figures as at turn-up

Test bed

Generator on one side, loopback or a second unit on the other

The circuit sits between two UNIs. Our bench generates and analyses every service at once; the far end returns the frames or measures them on its own clock

GENERATOR · ANALYSER · SITE A TX + RX PER SERVICE · ONE CLOCK ONE STREAM PER SERVICE VLAN + PCP or DSCP as the SLA marks it EMIX: 64 · 128 · 512 · 1024 · 1518 B Step load to CIR, then CIR + EIR Sequence number + timestamp per frame GREEN AND YELLOW FRAMES COUNTED APART UNI A 3 SERVICES NETWORK UNDER TEST EVERY HOP, EVERY POLICER ON THE PATH EVC 1 · VOICE · CoS 5CIR 20 M EVC 2 · VIDEO · CoS 3CIR 100 M EVC 3 · DATA · CoS 0CIR 500 M + EIR 200 M ONE CIRCUIT · THREE CLASSES · ONE RUN COLOUR-AWARE 802.1Q PCP · DSCP CBS · EBS BURSTS 1G · 10G · 100G UNI GPS OR PTP FOR ONE-WAY UNI B RETURNED FAR END · SITE B LOOPBACK, OR A SECOND UNIT WHAT THE FAR END GIVES Loopback: MAC swap, round-trip FTD Second unit: one-way FTD and FDV Counters per service, per colour Both directions when the SLA is two-way IR · FTD · FDV · FLR · AVAILABILITY BED FIRST, NETWORK SECOND generator looped on itself through the same optics and cables: 100% of CIR + EIR, zero loss, FDV under 10 µs, before the circuit is plugged in

Generator and analyser

One stream per service with its VLAN, PCP or DSCP marking, a sequence number and a timestamp in every frame. EMIX frame sizes by default; a fixed size when the SLA asks for it

Network under test

UNI to UNI across every hop and every policer of the provider. The bed is first looped on itself through the same optics: full rate, zero loss, before the circuit is plugged in

Far end

A loopback with MAC swap gives round-trip FTD and all counters. A second unit with GPS or PTP sync gives one-way delay and jitter, and both directions at once

Rates we run

1 GbE, 10 GbE and 100 GbE UNIs. For 100G services the step load and the long run go at line rate; the limit is the far end, so we agree the loopback type up front

Parameters

What is fixed before the run

Y.1564 leaves the SLA numbers to the contract and the duration to the parties. The plan lists each of them, with the value we use when the contract is silent

Parameter Our default Allowed Why it matters
CIR per serviceFrom the SLAAny, up to line rateThe rate the provider commits to; the step load ends here and the long run stays here
EIR per serviceFrom the SLA, 0 if absent0 to line rate minus CIRExcess traffic the network may carry; frames above CIR are yellow and may be dropped without a fail
Step load25, 50, 75, 100 % of CIRAny list of 2 or more stepsShows at what load the shaper or policer starts to bite, before the full-rate step
Step time30 s10 s to 60 sLong enough for the policer bucket to settle and for FDV to be sampled; short enough to keep the test under half an hour
Frame sizesEMIX: 64, 128, 512, 1024, 1518 BAny list, or one fixed sizeReal traffic is mixed; a single size hides shaper behaviour at small frames
Performance run2 h15 min, 2 h, 24 hShort runs catch configuration faults; long runs catch congestion from other customers and the daily cycle
Colour modeColour-aware when EIR > 0Colour-blind or colour-awareDecides whether yellow frames count toward FLR; a colour-blind run on a colour-aware network fails the policing step
Pass thresholdsFLR, FTD, FDV, availability from the SLAPer service, per metricEvery metric at every step is compared to its threshold; one miss marks the service failed and the run goes on

Defaults follow the Y.1564 recommendation and the MEF 10 service attributes. When the contract names a value, the contract wins

Always in our runs

  • Both tests, configuration then performance (§8)
  • Every service of the circuit in one run (§6)
  • Colour-aware marking when the SLA has an EIR (§7.1)
  • Policing step above CIR + EIR (§8.1.2)
  • Counters from both ends, per colour
  • Loopback pre-test of the bed

Both tests in order, every service of the circuit, colour-aware marking where there is an EIR, a policing step above the line, counters from both ends. This is what makes the result an activation and not a demo.

Agreed per circuit

  • Performance run 15 min, 2 h or 24 h (§8.2)
  • Round trip by loopback or one-way with sync
  • Frame size list: EMIX or a fixed size
  • Step count and step time (§8.1.1)
  • CBS and EBS burst test (§8.1.3)
  • Both directions, or the direction in the SLA

Run length, direction, frame sizes, step count and burst tests are set in the plan with you. Each one changes how long the run takes and what the report can claim.

Inside the method

Six steps per service, then the long run together

Y.1564 defines a configuration test and a performance test. The first goes EVC by EVC, the second loads all of them at once

  1. 8.1.1Configuration

    Step load to CIR

    Each service alone, at 25, 50, 75 and 100 % of its CIR, one step time per level. IR must equal the offered rate at every step and FTD, FDV and FLR must stay inside the SLA. A shaper set too low shows up as IR below offered at the 100 % step.

    §8.1.1 · per service · 4 × step time
  2. 8.1.1Configuration

    CIR + EIR

    Offered rate rises to CIR + EIR. Green frames must still arrive without loss; yellow frames may be dropped and are counted on their own. In colour-blind mode the network only has to pass at least the CIR.

    §8.1.1 · per service · 1 × step time
  3. 8.1.2Configuration

    Traffic policing

    Offered rate goes 25 % above CIR + EIR. The policer must cut the received rate back to CIR + EIR within a tolerance, with no loss among green frames. A policer that passes everything fails here, and so does one that drops green.

    §8.1.2 · per service · 1 × step time
  4. 8.1.3Configuration

    CBS and EBS bursts

    Bursts of the committed and excess burst sizes on top of the CIR. The network must absorb a CBS burst without loss and an EBS burst without dropping green. Optional in the recommendation; we run it when the SLA names burst sizes.

    §8.1.3 · per service · when agreed
  5. 8.2Performance

    All services at CIR

    Every service of the circuit at its CIR at the same time, for the agreed length. IR, FTD, FDV and FLR are sampled every second per service and compared to the thresholds; one second over the limit is recorded, a pattern of them is a fail.

    §8.2 · all services · 15 min, 2 h or 24 h
  6. 8.2Performance

    Availability

    Each second of the long run is available or unavailable by its FLR, availability per ITU-T Y.1563, counted in 1 s intervals. The share of available seconds per service goes against the SLA figure, typically 99.95 % or better.

    §8.2 · all services · same run

Procedure

From the plan to the signature

The whole sequence for a circuit with three services on a 1 GbE UNI fits in a working day, most of it the long run

  1. 00

    Test plan

    Services, CIR, EIR, colour mode, frame sizes, run length and every threshold in one table, signed off by both sides before anything is plugged in

  2. 01

    Bed check

    Generator looped on itself through the test optics and cables: CIR + EIR of every service, zero loss, FDV under 10 µs. The circuit is connected only after this passes

  3. 02

    Configuration test

    Service by service: four steps to CIR, the EIR step, the policing step, bursts when agreed. A failed step is reported at once so the configuration can be fixed and the step rerun

  4. 03

    Performance test

    All services at CIR together for the agreed time, counters per service per second from both ends. A run that shows a fault is kept, the time series is the evidence

  5. 04

    Report and hand-over

    Pass or fail per service and per metric, with the one-second series and the raw counters attached. A failed service comes with the step and the minute where it went wrong

ONE RUN · 3 SERVICES · Y.1564 §8 CONFIGURATION FIRST, THEN PERFORMANCE · EVERY KPI AT EVERY STEP SERVICE CONFIGURATION TEST · ONE SERVICE AT A TIME 6 STEPS × 3 SERVICES · 10 s TO 60 s A STEP EVC 1 · VOICE EVC 2 · VIDEO EVC 3 · DATA 25 · 50 · 75 · 100% CIR +EIR · POLICING same six steps same six steps SERVICE PERFORMANCE TEST ALL AT CIR, TOGETHER EVC 1 EVC 2 EVC 3 CIR 20 M · FLR, FTD, FDV sampled every second CIR 100 M · availability counted in 1 s intervals CIR 500 M · 15 min, 2 h or 24 h, as agreed IR ≥ CIR AT STEP 4 = SHAPING IS RIGHT IR ≤ CIR + EIR UNDER POLICING = THE POLICER WORKS FLR OVER THE SLA IN THE LONG RUN = NOT READY FOR TRAFFIC

One run, three services: six steps each in turn, then the long run together. The pass rules are the same at every step

Pass rules we apply

  • IR ≥ CIR

    At the 100 % step and through the long run the received rate per service equals the committed rate, within the tolerance of the frame size mix

  • 0 green loss

    No green frame lost at any step of the configuration test, and FLR under the SLA figure in the performance test

  • ≤ CIR + EIR

    Under policing the received rate is capped at CIR + EIR; a cap too high or too low fails the step

  • FTD · FDV

    Delay and jitter per service inside the SLA at every step, measured on the same frames that carry the rate

Time budget

  • Bed check10 minutes, once per set of optics and cables
  • Configuration test3 to 6 minutes per service with a 30 s step; three services in under 20 minutes
  • Performance test15 minutes for a quick turn-up, 2 hours as our default, 24 hours when the SLA names a daily availability figure
  • ReportSame day for a 2 h run, the day after for 24 h

Report

What you get at the end

A verdict per service and per metric, and the series behind it. Enough to sign the acceptance, or to send the provider to the exact minute and hop

Per service

  • IR at every step of the configuration test, offered against received
  • Green and yellow frame counts at CIR + EIR and under policing
  • FTD and FDV at every step: minimum, average, maximum
  • FLR over the performance run and per second
  • Availability: available and unavailable seconds
  • Pass or fail against each SLA threshold

Context

  • Service table: VLAN, marking, CIR, EIR, CBS, EBS, colour mode
  • Bed check result with the optics and cables used
  • Direction of each measurement, loopback type or sync source
  • Frame size list and step time
  • Timestamps of every step and of the long run
  • Firmware and software versions of the test units
Y.1564 · ACTIVATION REPORT 1 GbE UNI · 3 EVC · 2 h RUN 8.1EVC 1 voice · CIR 20 M, steps 8.1EVC 2 video · CIR 100 M, steps 8.1EVC 3 data · CIR 500 M + EIR 200 M 8.1EVC 3 policing at CIR + EIR + 25% 8.2EVC 1 · FLR / FTD / FDV 8.2EVC 2 · FLR / FTD / FDV 8.2EVC 3 · FLR / FTD / FDV 8.2Availability, all services 20.0 M 100.0 M 699.8 M 700.1 M 0.000% 0.001% 0.42% 100% IR at every step · FTD 4.1 ms · FDV 0.2 ms · FLR 0 IR at every step · FTD 4.3 ms · FDV 0.4 ms · FLR 0 green 500 M, yellow 199.8 M · FTD 4.6 ms offered 875 M, passed 700.1 M · policer holds FTD 4.1 ms · FDV 0.3 ms · SLA: 1% · 10 ms · 2 ms FTD 4.4 ms · FDV 0.6 ms · SLA: 0.1% · 20 ms · 5 ms loss from minute 38 · SLA 0.1% · green frames only 0 unavailable seconds in 7,200 ATTACHED SERVICE TABLE STEP GRAPHS 1 s KPI SERIES COUNTERS · PCAP Values are illustrative. Your report carries your services, your SLA thresholds and your run length.

Attached

  • One-second KPI series per service as CSV, for your own charts
  • Raw counters from both ends and a PCAP of the first and last minute
  • Step graphs as PNG, ready for the acceptance document

From practice

The service that passed alone and failed in company

A 1 GbE UNI with voice, video and data. Every service passed its configuration test: steps to CIR clean, policer at the right line. In the performance test, from the 38th minute, the data service lost 0.42 % of its green frames while the other two stayed clean.

The three CIRs add up to 620 Mbit/s. The policer on the aggregation node was set to 450 Mbit/s, a figure from an earlier version of the order. Alone, no service ever reached it. Together, they did. The SLA was signed a day later, after the fix and a second run.

Y.1564 performance test on a 1 GbE E-Line, three EVCs, 2 h run

SAME CIRCUIT · FRAME LOSS RATIO · TWO TESTS CONFIGURATION TEST · EVC 3 ALONE 500 M green for 60 s per step · 0 frames lost · passed 0.00% PERFORMANCE TEST · 3 EVC TOGETHER, 2 h 620 M on the trunk from minute 38 · EVC 3 loses green frames · SLA 0.1% 0.42% POLICER ON THE AGGREGATION NODE 450 M Set at 450 M instead of 620 M for the three classes. Found before the SLA started.

Services

How we run Y.1564 for you

On our CloudLAB bench against your equipment, or on your circuit with our units at both ends. Same plan, same report

Turn-up of a new circuit

Plan, bed check, both tests, report. For the provider, or for the customer who wants an independent acceptance

Order a turn-up

Device check before deployment

Your CE or aggregation device on our bench, with the services it will carry. Policers, shapers and QoS maps proven before they go into the field

Bring your device

Dispute between provider and customer

One run, one report, both parties on the distribution list. The numbers are the same for everyone

Ask for a neutral run

Recurring checks

The same plan rerun after each change in the path, with the previous report as the baseline

Set up a schedule
  • Report in the Y.1564 structure, step by step, service by service
  • Test units chosen for the rate and the far end, from more than one vendor
  • Engineers who read the SLA before the counters

Related methodologies

FAQ

Questions we get about Y.1564

How is Y.1564 different from RFC 2544?

RFC 2544 benchmarks a device with one stream at a time and searches for the maximum throughput. Y.1564 activates a service: several services at once, at the rates in the SLA, with pass or fail per metric. A full RFC 2544 set takes hours; a Y.1564 configuration test takes minutes per service.

Do we need a second test unit at the far end?

A loopback is enough for rate, loss and round-trip delay. One-way delay and jitter, or a circuit whose two directions differ, need a second unit with GPS or PTP synchronisation. The plan says which one applies.

What does colour-aware mean in practice?

Frames within CIR are green, frames between CIR and CIR + EIR are yellow. A colour-aware network may drop yellow frames under congestion and must keep the green ones. The test unit marks and counts them separately, so a dropped yellow frame does not fail the service.

How long should the performance test run?

Fifteen minutes catches configuration errors. Two hours is our default and catches most congestion from other traffic. Twenty-four hours is needed when the SLA names a daily availability figure or when the fault is tied to a time of day.

Can you test a 100 GbE service?

Yes. The step load and the long run go at line rate. The limit is the far end: a passive loopback works, a second unit must keep up with the rate. We agree the pair before the plan is signed.

What happens when a service fails?

The run goes on and the service is marked failed with the step and the time. You fix the configuration, we rerun the configuration test of that service and then the performance test with all services. The report shows both runs.

Test plan and quote in 1 to 3 business days

Send the service table or the SLA extract. You get the plan with every parameter and threshold, the run length and a quote for the run

Request a Y.1564 plan Log in to CloudLAB

Send the shortlist

You have the method. Now get the numbers.

Name the models and the KPIs that matter. An engineer replies within one business day with a test plan draft and a quote.

Prefer to chat? Message an engineer on WhatsApp

We use what you send only to answer you. See our Privacy Policy.