Nov. 2010Geib/Morton/Fardid/Steinmitz / draft-ietf-metrictest-011 Testing Standards Track Metrics...

5
Nov. 2010 Geib/Morton/Fardid/Steinmitz / draft-ietf- metrictest-01 1 Testing Standards Track Metrics Draft-ietf-ippm-metrictest- 01 Geib, Morton, Fardid, Steinmitz IETF-79 November 2010

Transcript of Nov. 2010Geib/Morton/Fardid/Steinmitz / draft-ietf-metrictest-011 Testing Standards Track Metrics...

Page 1: Nov. 2010Geib/Morton/Fardid/Steinmitz / draft-ietf-metrictest-011 Testing Standards Track Metrics Draft-ietf-ippm-metrictest-01 Geib, Morton, Fardid, Steinmitz.

Geib/Morton/Fardid/Steinmitz / draft-ietf-metrictest-01 1Nov. 2010

Testing Standards Track Metrics

Draft-ietf-ippm-metrictest-01

Geib, Morton, Fardid, SteinmitzIETF-79 November 2010

Page 2: Nov. 2010Geib/Morton/Fardid/Steinmitz / draft-ietf-metrictest-011 Testing Standards Track Metrics Draft-ietf-ippm-metrictest-01 Geib, Morton, Fardid, Steinmitz.

Geib/Morton/Fardid/Steinmitz / draft-ietf-metrictest-01 2Nov. 2010

Draft-ietf-ippm-metrictest-01

Proposed definition of easy and difficult metrics to be tested

Frequency of events and temporal properties of impairment processes are to be respected by a metric test.

A frequent event produces one outcome helpful to capture a distribution of results

(almost) per singleton. This is true for Delay, Delay Variation or capacity metrics. The proposed test allows comparisons of implementations.

A rare event produces infrequent outcomes helpful to capture a distribution of rates (these events are at least often charaterised by rates). Packet Loss and Duplication fit here. The proposed test allows comparisons of implementations.

Some rare events probably can’t be tested with implementations sending streams of different packet frequency (and may be sizes), like Loss Pattern or Reordering. The temporal properties of the impairment process impact the test. A suitable test set up requires further study.

Page 3: Nov. 2010Geib/Morton/Fardid/Steinmitz / draft-ietf-metrictest-011 Testing Standards Track Metrics Draft-ietf-ippm-metrictest-01 Geib, Morton, Fardid, Steinmitz.

Geib/Morton/Fardid/Steinmitz / draft-ietf-metrictest-01 3Nov. 2010

Draft-ietf-ippm-metrictest-01.

Test set up comparison.

Test Set Up Evaluation

Single test lab, emulate the Internet or measure across the Internet.

Causes travel cost, time consuming for one party.

Remote test labs, emulate the Internet with identically configured impairment generators.

Requires test specification and may introduce the impairment generator as additional source of error.

Remote test labs, measure across the Internet.

The same tunnel segments are passed by the separate measurements flows in different order. Only metrics capturing frequent events can be tested.

Remote test labs, measure across the Internet and include a single impairment generator to be passed by the tunnel.

The same tunnel segments are passed by the separate measurements flows in different order. Metrics capturing frequent and rare events can be tested.

Page 4: Nov. 2010Geib/Morton/Fardid/Steinmitz / draft-ietf-metrictest-011 Testing Standards Track Metrics Draft-ietf-ippm-metrictest-01 Geib, Morton, Fardid, Steinmitz.

Geib/Morton/Fardid/Steinmitz / draft-ietf-metrictest-01 4Nov. 2010

Draft-ietf-ippm-metrictest-01.

New experimental set up avoiding

Compare two implementations by a two site measurement configuration That is an experiment, to be carried out as a diploma thesis. Aim:prove IPPM standard compliance of the RIPE Test box and of Telekom’s Perfas

(which is a self development only used internally). Pro: Simple, an no equipment is moved. Con:The same L2TPv3 tunnel segments are

passed in different order by the different implementations.

U-Turn CrossconnectPort VLAN ID 200 with Port VLAN ID 201

Metric Implementation 1b

InternetF-Eth

10.100.64.32/24 EtherType 0x8100

VLAN ID 100

L2TPv3 Tunnel HeadPort modexconnect 10.100.100.1 1 encapsulation l2tpv3protocol noneip local interface Loopback0

F-Eth

CPE CPE

10.100.64.42/24 EtherType 0x8100

VLAN ID 101

F-Eth

L2TPv3 Tunnel HeadPort modexconnect 10.0.0.1 1 encapsulation l2tpv3protocol noneip local interface Loopback0

Metric Implementation 2b

F-Eth

10.100.124.0/24 EtherType 0x8100

VLAN ID 200

Metric Implementation 2a

F-Eth 10.100.123.0/24 EtherType 0x8100

VLAN ID 201

F-Eth

U-Turn CrossconnectPort VLAN ID 100 with Port VLAN ID 101

F-Eth

Optional ImpairmentGenerator

VLAN 100

VLAN 101

VLAN 200

VLAN 201

Page 5: Nov. 2010Geib/Morton/Fardid/Steinmitz / draft-ietf-metrictest-011 Testing Standards Track Metrics Draft-ietf-ippm-metrictest-01 Geib, Morton, Fardid, Steinmitz.

Geib/Morton/Fardid/Steinmitz / draft-ietf-metrictest-01 5Nov. 2010

Draft-ietf-ippm-metrictest-01.

Next steps.

Carry out the experiment of a complete metric test (not just ADK) and include reports in -02.

If the experiments validates the idea of metric test, finalise draft and go for last call.

We don’t want to worry about what do, if not.