SIP%20End-to-End%20Performance%20Metrics

1
SIP End-to-End Performance Metrics

• 70th IETF Conference
• PMOL
• Daryl Malas

2
Problem Statement

• With widespread implementation of SIP the
  following problems have surfaced
• No standard method for measuring SIP performance
• Enterprise and SIP Service Provider confusion on
  How
• Enterprise and SIP Service Provider confusion on
  Where to measure
• Current reliance on PSTN (and other) metrics
• Draft defines the How and the Where for
  common metrics applicable to ALL SIP applications.

3
Relevant Standards Dev

• IETF
• PMOL WG to develop standard to benchmark
  end-to-end SIP application performance
• BMWG developing standard to benchmark SIP
  networking device performance
• Other
• SPEC developing industry-available test code for
  SIP benchmarking in accordance with IETFs BMWG
  and PMOL standards
• ITU-T (Study Group 12) focused on detailed
  failure/success scenarios and end-to-end
  benchmark guidance

4
History

• Introduced in SIPPING WG at 66th IETF conference
• Confusion around scope relevant to BMWG SIP tests
• Confusion around draft home (e.g. IPPM, SIPPING,
  or BMWG)
• Draft did not fit within the charter of these
  groups
• Unanimous SIPPING WG consensus of interest and to
  continue development of the draft
• Updates and continual progress with draft
  requested from WG and ADs
• Discussion of further updates and progress in
  SIPPING WG at 67th and 68th IETF conference
• Draft still in search of home at 68th conference
• Draft introduced in APM BOF at 69th IETF
  conference

5
Draft Discussion

• Consensus on purpose and scope
• Framework
• Concept LCR (Least Cost Routing) and QBR
  (Quality Based Routing)

6
Example Metric - SER

• Session Establishment Rate (SER)
• Detect ability of a terminating UA or downstream
  proxy to successfully establish new sessions

7
Real World Application

• Enterprise with multiple upstream SIP Service
  Providers (SSP)

• SER SSP A 68
• SER SSP B 87
• What can the Enterprise do with this data?

8
Currently Defined Metrics

• Registration Request Delay (RRD)
• Detect failures or impairments causing delays in
  REGISTER requests
• Session Request Delay (SRD)
• Detect failures or impairments causing delays in
  new session requests
• Session Disconnect Delay (SDD)
• Detect failures or impairments causing delays in
  ending a session
• Session Duration Time (SDT)
• Detect problems (e.g. Poor audio quality) causing
  short sessions
• Average Hops per Request (AHR)
• Detect inefficient routing and failure conditions
  due to number of elements traversed

9
Currently Defined Metrics

• Session Establishment Efficiency Rate (SEER)
• Complements SER, but excludes potential effects
  of terminating UAS
• Session Defects (SD)
• Detect consistent failures in dialog processing
• Ineffective Session Attempts (ISA)
• Detect proxy or UA failures or congestion
  conditions causing setup requests
• Session Disconnect Failures (SDF)
• Detects when an active session is terminated due
  to a failure condition
• Session Completion Rate (SCR)
• Detects failures caused by downstream proxies not
  responding to new session setups
• Session Success Rate (SSR)
• Ratio of successful sessions compared with
  sessions failing due to ISA or SDF

10
Next Steps

• Agreed upon set of metrics
• Too many
• Too few
• Different Metrics
• Concerns or Questions
• Working Group Item?