BUFFER on, off, empty on, off, full O PPORTUNISTIC NETWORKS S - - PowerPoint PPT Presentation

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

HYBRID PERFORMANCE MODELLING OF

OPPORTUNISTIC NETWORKS

Luca Bortolussi1 Vashti Galpin2 Jane Hillston2

1Dipartimento di Matematica e Geoscienze

Università degli studi di Trieste luca@dmi.units.it

2Laboratory for Foundation of Computer Science

University of Edinburgh Vashti.Galpin@ec.ac.uk Jane.Hillston@ec.ac.uk

QAPL 2012, Tallinn, Estonia, April 1, 2012

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

OUTLINE

1 OPPORTUNISTIC NETWORKS 2 STOCHASTIC HYPE 3 OPPORTUNISTIC NETWORKS IN SHYPE

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

OPPORTUNISTIC NETWORKS

characteristics and challenges

disconnectedness: very low probability of direct path between any two nodes high latency and low data rate: due to lack of connectivity, queuing at intermediate nodes limited resources: battery-driven, hostile environment, constraints on storage and communication

protocols

approach: store-carry-forward main objective: maximise probabality of packet reaching destination secondary objectives: minimise delivery delay and resources usage

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

OPPORTUNISTIC NETWORKS: PROTOCOLS

classification of protocols

deterministic versus stochastic with or without infrastructure flooding versus forwarding

flooding

each packet is forwarded to many nodes epidemic routing prevents forwarding of already-seen packets

forwarding

each packet is forwarded to one node information gathered during communication used to choose example: historical information about path likelihood

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

STOCHASTIC HYPE: OVERVIEW

MAIN FEATURES OF HYPE Modelling of the continuous dynamics is focussed on the notion of flow. Control structure of events is modelled separately from the reaction of continuous dynamics to events. Events can either happen at stochastic times (exponentially distributed) or instantaneously, when guards become true. Modularity in the definition of the actual form of guards, resets, functions. In summary: compositionality.

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

A SIMPLE (FLUID) BUFFER MODEL

BUFFER

input

• utput
• n, off, full
• n, off, empty

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

HYPE ACTIONS AND ACTIVITIES

EVENTS Events can either be instantaneous (buffer becomes full or empty: full, empty ∈ Ed)

• r they can happen at exponential random times

(input/output of data in buffer: onin/out, offin/out ∈ Es) ACTIVITIES OR INFLUENCES Description of flows: influences on continuous variables. For instance, the effect of data inflow in the buffer is α ∈ A α( X) = (in, rin, I( X))

✟✟ ✟ ✯ ✻ ❍ ❍ ❍ ❨

influence name rate influence type here I( X) := const

where X is a formal parameter.

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

HYPE ACTIONS AND ACTIVITIES

EVENTS Events can either be instantaneous (buffer becomes full or empty: full, empty ∈ Ed)

• r they can happen at exponential random times

(input/output of data in buffer: onin/out, offin/out ∈ Es) ACTIVITIES OR INFLUENCES Description of flows: influences on continuous variables. For instance, the effect of data inflow in the buffer is α ∈ A α( X) = (in, rin, I( X))

✟✟ ✟ ✯ ✻ ❍ ❍ ❍ ❨

influence name rate influence type here I( X) := const

where X is a formal parameter.

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

HYPE ACTIONS AND ACTIVITIES

EVENTS Events can either be instantaneous (buffer becomes full or empty: full, empty ∈ Ed)

• r they can happen at exponential random times

(input/output of data in buffer: onin/out, offin/out ∈ Es) ACTIVITIES OR INFLUENCES Description of flows: influences on continuous variables. For instance, the effect of data inflow in the buffer is α ∈ A α( X) = (in, rin, I( X))

✟✟ ✟ ✯ ✻ ❍ ❍ ❍ ❨

influence name rate influence type here I( X) := const

where X is a formal parameter.

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

HYPE ACTIONS AND ACTIVITIES

EVENTS Events can either be instantaneous (buffer becomes full or empty: full, empty ∈ Ed)

• r they can happen at exponential random times

(input/output of data in buffer: onin/out, offin/out ∈ Es) ACTIVITIES OR INFLUENCES Description of flows: influences on continuous variables. For instance, the effect of data inflow in the buffer is α ∈ A α( X) = (in, rin, I( X))

✟✟ ✟ ✯ ✻ ❍ ❍ ❍ ❨

influence name rate influence type here I( X) := const

where X is a formal parameter.

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

HYPE ACTIONS AND ACTIVITIES

EVENTS Events can either be instantaneous (buffer becomes full or empty: full, empty ∈ Ed)

• r they can happen at exponential random times

(input/output of data in buffer: onin/out, offin/out ∈ Es) ACTIVITIES OR INFLUENCES Description of flows: influences on continuous variables. For instance, the effect of data inflow in the buffer is α ∈ A α( X) = (in, rin, I( X))

✟✟ ✟ ✯ ✻ ❍ ❍ ❍ ❨

influence name rate influence type here I( X) := const

where X is a formal parameter.

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

UNCONTROLLED SYSTEM

SUBCOMPONENTS S ::= a:α.Cs | S + S a ∈ E, α ∈ A Cs( X)

def

= S (simple looping agents) COMPONENTS P ::= C( X) | P ⊲

⊳

L P

L ⊆ E C( X)

def

= P or subcomponent name UNCONTROLLED SYSTEM Σ ::= C( V) | Σ ⊲

⊳

L Σ

L ⊆ E

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

SIMPLE BUFFER IN SHYPE

INPUT OF DATA Input

def

= onin:(in, rin, const).Input + offin:(in, 0, const).Input+ full:(in, 0, const).Input + init:(in, 0, const).Input OUTPUT OF DATA Output

def

= onout:(out, −rout, const).Output +offout:(out, 0, const).Output +empty:(out, 0, const).Output +init:(out, 0, const).Output UNCONTROLLED SYSTEM Sys

def

= Input ⊲

⊳

{init} Output

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

SHYPE CONTROLLED SYSTEM

CONTROLLER M ::= a.M | 0 | M + M a ∈ E Con ::= M | Con ⊲

⊳

L Con

L ⊆ E CONTROLLED SYSTEM ConSys ::= Σ ⊲

⊳

L Con

L ⊆ E WELL-DEFINED HYPE SYSTEM Subcomponents are in bijection with influence names. init:(ι, _, _) appears exactly once a:(ι, _, _) appears at most once synchronisation on all shared events

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

SIMPLE BUFFER IN SHYPE (CONTINUED)

CONTROLLERS Conin

def

= onin.Con′

in

Con′

in

def

= offin.Conin + full.Conin Conout

def

= onout.Con′

• ut

Con′

• ut

def

= offout.Conout + empty.Conout CONTROLLED SYSTEM Con

def

= Conin ⊲

⊳

∅ Conout

SYSTEM Buffer

def

= Sys ⊲

⊳

M init.Con

with M = {init, onin, offin, onout, offout, empty, full}

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

EVENT CONDITIONS AND OTHER STUFF

V is the set of system variables. ec associates event conditions to events. Event conditions for instantaneous events consist of an activation condition (a predicate/boolean formula on system variables) and a reset function. Event conditions for stochastic events consist of a rate function (of system variables) and a reset function.

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

EVENT CONDITIONS AND OTHER STUFF

V is the set of system variables. ec associates event conditions to events. Event conditions for instantaneous events consist of an activation condition (a predicate/boolean formula on system variables) and a reset function. Event conditions for stochastic events consist of a rate function (of system variables) and a reset function.

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

EVENT CONDITIONS AND OTHER STUFF

V is the set of system variables. ec associates event conditions to events. Event conditions for instantaneous events consist of an activation condition (a predicate/boolean formula on system variables) and a reset function. Event conditions for stochastic events consist of a rate function (of system variables) and a reset function.

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

SIMPLE BUFFER IN SHYPE (CONTINUED)

System variable: B (buffer level). EVENT CONDITIONS ec(init) = (true, B′ = b0) ec(full) = (B = maxB, true) ec(empty) = (B = 0, true) ec(onin) = (kon

in , true)

ec(offin) = (koff

in , true)

ec(onout) = (kon

• ut, true)

ec(offout) = (koff

• ut, true)

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

FORMAL SEMANTICS OF SHYPE:

STOCHASTIC HYBRID SYSTEMS

A formal semantics of stochastic HYPE can be given in terms

• f a class of stochastic processes called Piecewise

Deterministic Markov Processes.

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

MODELLING OPPORTUNISTIC NETWORKS

We assume a large number of packets: fluid representation. Network nodes deal with multiple data streams, may have priorities. Each stream can be buffered, generated or consumed by node. data streams can be uni- and/or bidirectional each stream can connect to many other nodes mobile nodes have routes – geometric or abstract

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

MODELLING OPPORTUNISTIC NETWORKS

We assume a large number of packets: fluid representation. Network nodes deal with multiple data streams, may have priorities. Each stream can be buffered, generated or consumed by node. data streams can be uni- and/or bidirectional each stream can connect to many other nodes mobile nodes have routes – geometric or abstract

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

MODELLING OPPORTUNISTIC NETWORKS IN SHYPE

STREAM ν IN NODE i Nodei,v

def

= Inputi,v ⊲

⊳

∗ Outputi,v ⊲

⊳

∗ Generatei,v

⊲ ⊳

∗

Removei,v ⊲

⊳

∗ Dropi,v ⊲

⊳

∗ KeepIi,v ⊲

⊳

∗ KeepGi,v

CONTROLLER OF STREAM ν IN NODE i ConNodei,v

def

= ConIi,v ⊲

⊳

∗ ConOi,v ⊲

⊳

∗ ConGi,v

⊲ ⊳

∗

ConRi,v ⊲

⊳

∗ ConDi,v

CONTROLLER FOR CONNECTION BETWEEN NODE i AND j ON

STREAM ν

ConStreami,j,v

def

= ConUnii,j,v ⊲

⊳

∗ ConUnij,i,v

⊲ ⊳

∗

CBii,j,v ⊲

⊳

∗ ConTidyi,j,v

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

CASE STUDY

existing wildlife monitoring projects

ZebraNet: collars on zebra, data collected by mobile nodes in vehicles, both flooding and history-based protocols used SWIM: whales are tagged, flooding, collection by fixed stations (buoys) and mobile stations (seabirds)

• ur scenario

video data captured by movement-triggered stationary sensors: 10 video sensors, 250MB disk space, each records 3 times a day for 3 minutes on average newer data overwrites old data if disk is full vehicular ferry collects data and delivers to a base station unidirectional communication abstracted as data flowing from sensor to ferry and ferry to base station

vary ferry parameters such as speed, route and buffer size

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

CASE STUDY DETAILS

parameters

mean-time-to-contact (MTC): 6 values in 5 minute increments ferry buffer size: 5 values in 250MB increments

routes: 4 variants

return-at-end with fixed route (raef) return-at-end with random route (raer) return-when-full with fixed route (rtbf) return-when-full with random route (rtbr)

measurements

data dropped and collected: for MTC and route data dropped and collected: for buffer size and route

We have an implementation of sHYPE that we used to analyse the case study.

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

ROUTE: RETURN-AT-END WITH FIXED ROUTE

V1 V2 V10 Base station

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

ROUTE: RETURN-AT-END WITH RANDOM ROUTE

Vi1 Vi2 Vi10 Base station

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

ROUTE: RETURN-WHEN-FULL WITH FIXED ROUTE

V1 V2 V10 Base station

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

ROUTE: RETURN-WHEN-FULL WITH RANDOM ROUTE

Vi1 Vi2 Vi10 Base station

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

RESULTS: MEAN-TIME-TO-CONTACT AND ROUTE

DATA DROPPED

• 5

10 15 20 25 30 10 20 30 40 50 MTC (min) Total data dropped (MB)

• raef

raer rtbf rtbr

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

RESULTS: MEAN-TIME-TO-CONTACT AND ROUTE

DATA COLLECTED

• 5

10 15 20 25 30 200 400 600 800 1000 MTC (min) Total data collected (MB)

• raef

raer rtbf rtbr

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

RESULTS: FERRY BUFFER SIZE AND ROUTE

DATA DROPPED

• 600

800 1000 1200 1400 10 20 30 40 50 60 Ferry buffer capacity (MB) Total data dropped (MB)

• raef

raer rtbf rtbr

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

RESULTS: FERRY BUFFER SIZE AND ROUTE

DATA COLLECTED

• 600

800 1000 1200 1400 400 600 800 1000 1200 Ferry buffer capacity (MB) Total data collected (MB)

• raef

raer rtbf rtbr

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

CONCLUSIONS

We showed how to model opportunistic networks in HYPE a hybrid process algebra recently developed. Compositionality of HYPE allows a modular description of those networks, that can be easily generated from graphical models. We have implemented a simulator, and the results provided are generated with it. We are currently developing a graphical front-end and a simple textual language

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

THANKS FOR THE ATTENTION Questions?

OPPORTUNISTIC NETWORKS STOCHASTIC HYPE OPPORTUNISTIC NETWORKS IN SHYPE

ANNOUNCEMENT

First International Workshop on

Hybrid Systems and Biology

Newcastle upon Tyne September 3, 2012 at CONCUR 2012

Important dates JUNE 15, 2012 Abstract submission deadline JUNE 22, 2012 Full paper submission deadline