Delay&Measurement&Challenges&in&Mobile&Access& …Delay&Measurement&Challenges&in&Mobile&Access&...

13
Delay Measurement Challenges in Mobile Access Networks Joachim Fabini, Tanja Zseby TU Wien AIMS 2016

Transcript of Delay&Measurement&Challenges&in&Mobile&Access& …Delay&Measurement&Challenges&in&Mobile&Access&...

Delay  Measurement  Challenges  in  Mobile  Access  Networks

Joachim  Fabini,  Tanja  ZsebyTU  Wien

AIMS  2016

Problem  Statement• Motivation:  Smart  Grid  Communication  

– Use  of  cellular  access  networks– But  strict  latency  requirements

• Past  (RFC  2330):  State-­less  history-­less  network  behavior• Now:  “Reactive”  Cellular  Access  Networks

– Demand-­driven  resource  allocation– Time-­slotted  operation

• Examples:  HSPA,  LTE  – Time  slotting– Higher  capacity  for  active  users

• Problem:  repeatable  measurements– Same  patterns  è different  results– Concatenated  paths  è results  biased  by  first  segment

2016-02-11Fabini, Zseby

2

Packets  in  Cadence

2016-02-11Fabini, Zseby

3

(a)  Send  time  (client) (b)  Receive   time  (server)

Fabini,  Zseby,  Hirschbichler:  “Representative  Delay  Measurements  (RDM):  Facing  the  Challenge  of  Modern  Networks,”  I8th  International  Conference  on  Performance  Evaluation  Methodologies  and  Tools,  ICST,  Brussels,  Belgium,  2014

HSPA  Uplink HSPA  Uplink

è network  period  of  10  ms

Random  Send  Schedule Synchronized  Arrivals  

Delay  Dependence  on  Absolute  Send  Time  (LTE)

2016-02-11Fabini, Zseby

4

Suboptimal  Sampling

2016-02-11Fabini, Zseby

5

Fabini,  Zseby,  “The  Right  Time:  Reducing  Effective  End-­to-­End  Delay  in  Time-­Slotted  Packet-­Switched  Networks,”  IEEE/ACM  Transactions  on  Networking,  2015.

Optimized  Sampling

2016-02-11Fabini, Zseby

6

LTE  Delay  

2016-02-11Fabini, Zseby

7

Abs.  Send  Time:  1  -­ 2  ms (mod  10) Abs  Send  Time:  

4.5  – 5.5  ms (mod  10)

Rate  Dependence   (HSPA)

2016-02-11Fabini, Zseby

8

Low  Bit  Rate  Scenario   High  Bit  Rate  Scenario

Representative  Delay  Measurements   (RDM)

2016-02-11Fabini, Zseby

9

Fabini,  Zseby,  Hirschbichler:  “Representative  Delay  Measurements  (RDM):  Facing  the  Challenge  of  Modern  Networks,”  8th  International  Conference  on  Performance  Evaluation  Methodologies  and  Tools,  ICST,  Brussels,  Belgium,  2014

RDM• Scenario  concept

– Different  network  loads– Highly  compressed  packet  payloadè preventing  optimizers  to  further  compress  packet

• Check  for  reactive  network  behavior– Identical  scenario  definitions– Multiple  subsequent  test  runs

• Detecting  – Reactive  network  behavior  (e.g.  load  dependence)– Optimizers  (compressed  vs.  non-­compressed)

• Time  slotting  è Randomness  regeneration– Server  delays  packet  before  reflecting  it

2016-02-11Fabini, Zseby

10

Randomness  Regeneration   (HSPA)

2016-02-11Fabini, Zseby

11

RFC  2330  compliant Server  regenerated  randomness

Conclusion• Networks  have  evolved

– Stateful,    load  dependence,  history,  time-­slotting,  …• RFC  2330  conformant  measurements

– Impaired  repeatability  for  cellular  access  networks– Influence  to  concatenated  paths

• è RFC  7312:  Advanced  Stream  and  Sampling  Framework  for  IPPM– Influence  of  Packet  length,  payload  type– Influence  of  history  (rate,  inactivity)– Randomness  cancellation

• RDM  Tool  implements  ideas  from  RFC  7312– Stream  Description  (RFC7312  Options)– Randomness  Regeneration

2016-02-11Fabini, Zseby

12

Thank  You!

• 3  year  position,  full  time  • Topic:  Network  Traffic  Analysis,  Security• Contact:  [email protected]

2016-02-11Fabini, Zseby

13

Open  PhD  Position  at  TU  Wien