protocol converter synthesis using timed petri nets

Protocol Converter Synthesis Using Timed Petri Nets

Kevin CameraEE249 Fall 2000

10/24/2000

Motivation

Evolution of heterogeneous, distributed networks Protocol converters act as mediators

between otherwise incompatible protocols

This work improves on previous converter modeling and synthesis methodologies

Services and Protocols

Layer N services provide functions for layer N+1 via service access points User-based functional specification

Protocol entities (PE) exchange protocol data units (PDU) via lower and upper SAPs Low-level behavioral specification

Converter Design Approaches

Top-down:service-level conversion Easy to implement,

tends to be “passive”

Bottom-up:protocol-level conversion Very powerful, very

complex

Converter Properties

Safety Free from deadlock or livelock, and is

complete Liveness

Performs the required functionality Timeliness

Satisfies the timing of both protocols

Design Criteria

Modeling formalism CFSM, TPN, etc.

Design approach Service level, protocol level, or hybrid

Design methodology Analytic: trial-and-error iterations Synthetic: systematic, safe

generation

Design Criteria (con’t)

Information transfer issues Direct: no buffers, messages

transmitted immediately to each protocol

Indirect: messages stored in non-FIFO buffer, re-ordered, and transmitted

Synchronization issues Mapping of messages (traces) to

ensure compatbility

Design Criteria (con’t)

Timeliness Internal timing and protocol

requirements Data loss and recovery

Dynamicity Self-induced, active communication

Concurrency Complexity

Timed Petri Net Model

Standard Petri net with predicated and timed transitions

New notations Input/Output actions marked with +/- Parallel composition: PN1 || PN2

Trace “schuffling”: t1 t2

Complement of a trace: ~t

Example: Alternating Bit

Example: Poll-End

Synthesis Technique

Greatest common service definition

Trace generation and collection Trace synchronization Synthesis of Petri Net model

Greatest Common Service

Start with both service descriptions I/O operations are

service primitives Map equivalent

primitives into a service interface converter (SIC)

Remove primitives not mapped in SIC

)||||(}_,_{ MNMupperNupper SSICSPGCSD

Example: GCSD

?

Trace Generation

Interested in traces of each separate network which contribute to the GCSD

Can be found with following analysis: Let TN be set of traces at {lower,upper}_N Find N’, a pruning with contributions to S’N

Find us_N, composition of lower services and communication channel

TN = N’ || us_N

Example: Trace Generation

TABP ={ACCEPT –DATA(bit) (+ACK(~bit) –DATA(bit))* +ACK(bit),+DATA(bit) (-ACK(~bit) +DATA(bit))* DELIVER -ACK(bit),+DATA(bit) DELIVER (-ACK(~bit) +DATA(bit))* -ACK(bit),+DATA(bit) (+DATA(bit))* DELIVER -ACK(bit),+DATA(bit) DELIVER (+DATA(bit))* -ACK(bit)}

TPE ={SEND +poll (-data SEND) (-data SEND)* -end,+poll SEND (-data SEND) (-data SEND)* -end,(+poll)*,-poll (+data RECEIVE) (+data RECEIVE)* +end,(-poll)*}

Trace Synchronization

For trace sets TN and TM found above: Prune the protocol components TRN, TRM

Take complements to get TCN

Schuffle the complements (TCN TCM)

14 rules for ordering data, confirmation, ack, and nack messages safely (N+m,N-c) (M-m) = (N+m,M-m,N-c)

Example: Trace Synchronization

TCABP = {lower_ABP} ~TCABP ={-DATA(bit) (+ACK(~bit) -DATA(bit))* +ACK(bit),+DATA(bit) (-ACK(~bit) +DATA(bit))* -ACK(bit),-DATA(bit) (-DATA(bit))* +ACK(bit)}

TCPE = {lower_ABP} ~TCABP ={-poll (+data SEND) (+data SEND)* +end,(-poll)*,+poll (-data SEND) (-data SEND)* -end, (+poll)*}

TC = TCABP TCPE

Result (after PN synthesis)

Summary

Hybrid approach Starts with service specification, but

performs all synthesis on protocols Timed Petri net model

Can incorporate timing in specification Models concurrency and comes with well-

known analysis algorithms Resulting converter is safe and

functional