real-time scheduling for wirelesshart networks by abusayeed saifullah, you xu, chenyang lu, and...

35
Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua Ivaldi Dept of Electrical and Computer Engineering Center for Clean Energy Engineering University of Connecticut 1

Upload: regina-norton

Post on 25-Dec-2015

215 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

Real-Time Scheduling for WirelessHART Networks

byAbusayeed Saifullah, You Xu, Chenyang Lu,

and Yixin Chen

A Presentation of Findings for CSE5095

Joshua IvaldiDept of Electrical and Computer Engineering

Center for Clean Energy Engineering

University of Connecticut

1

Page 2: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

Introduction to WirelessHART

• Wireless Highway Addressable Remote Transducer, or WirelessHART is a

network standard tailored for sensors and actuators – or rather, the “internet

of things”.

• In wireless automated system applications, the level of human interaction

must be reduced as much as possible!

• So why do we care?

– Feedback based control loops in industrial environments require strict

adherence to end-to-end latency requirements on data exchange.

– Networks failing to meet feedback data transmission requirements may

result in production line inefficiency, destruction of equipment, and

highly impactful financial or environmental damage.

2

Page 3: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

Introduction to WirelessHART

• For the previously stated goals of the

communication standard, the protocol exhibits

the following key features:

– Centralized Network Management

architecture (controlled through the

gateway).

– Multi-Channel Time Division Multiple

Access (TDMA) Transmission (for

predictable latencies in RT Comm.)

– Redundant Routes (for self-healing

networks).

– Real Time operation.

– Avoidance of spatial reuse of channels for

enhanced performance and reliability.

3

Page 4: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

Features: Route and Spectrum Diversity

• Spatial diversity of routes allows messages to pass

through many paths. This increases the reliability of

the network!

• This is a useful feature in the event that there is

interference, physical obstacles, or broken links.

• Spectrum diversity allows the network to fully utilize

16 physical layer channels. Time slotted channel

hopping mitigates jamming and interference.

4

Page 5: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

Features: Handling Internal Interference

• Detection of interference between nodes is

challenging.

• To avoid this, WirelessHART allows only one

transmission in each channel per time slot, across

the entire network.

• Therefore, the maximum number of concurrent

transmissions cannot exceed the number of channels

in the network!5

Page 6: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

Developing a Mathematical Model

• Suppose that the mesh network can be

modeled as a graph, , where the nodes V

represent network devices, and E is the set of

edges between the devices.

• An edge, , exists in G iff the nodes u and v can

communicate with each other.

6

Page 7: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

Developing a Mathematical Model

• Each , is able to send packets to other network

devices.

• For any transmission, , occurring at an edge, ,

u is designated as the sender, and v is

designated as the receiver.

7

Page 8: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

Developing a Mathematical Model

• Suppose that two transmissions,

Occur simultaneously.

• And if,

Then we should expect that the transmissions are conflicting

because the channels cannot be shared! These events are

mutually exclusive (cannot occur simultaneously).

8

Page 9: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

Developing a Mathematical Model

• In the network, , we may consider N end-to-end

packet flows comprising the set, originating at a

network device, , then passing through the gateway

and finally terminating at a destination in the

network,.

• For a packet released at time slot k, and delivered at

time slot j, the end to end latency can be described

by,9

Page 10: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

Developing a Mathematical Model

• For each flow, Fi, the end-to-end latency is the

maximum end-to-end latency amongst all

packets during the transmission period

(channel/path with the greatest latency).

• Thus, each flow, Fi , is characterized by a total

period, Pi, a deadline, Di, where , and a set of

possible routes, Φi.10

Page 11: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

Developing a Mathematical Model

• Given the set of all flows, F, our main

objective is to schedule all transmissions

propagating through, m channels s.t.

• For the above problem, a scheduling algorithm,

Α, is optimal if A can schedule all transmission

whenever a feasible schedule exists.

11

Page 12: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

Theorem 1

• Given an instance of the

real-time scheduling

problem for a

WirelessHART network

with N flows, it is NP-

complete to decide whether

it is schedulable or not.

• This is verifiable in O(n2)

time. (proven in paper)

12

Page 13: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

Conditions for Scheduling

• In a WirelessHART network, simultaneous

transmissions are a major cause of

communication delays and difficulty in

scheduling signal flows.

• For a given set of flows, let T be the hyper

period, or least common multiple of the periods

of flows.

13

Page 14: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

Conditions for Scheduling

• Using to denote the j-th packet, , generated by

a flow Fi.

• For this packet, its release time, should be

upper-bounded by its absolute deadline,

14

Page 15: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

Conditions for Scheduling

• Given that we must meet our transmission

deadlines, there arises the issue of node hops.

• Suppose that we want to send a packet from

sender to receiver, which are postk hops away

from eachother. Then some transmission, τk,

needs to happen no later than slot,

15

Page 16: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

Conditions for Scheduling

• Before a transmission, τk can occur, prek,s

transmissions of a packet may be required,

because the message is prek,s hops away from

the source of τk.

• The anticipated release time of τk at time slot s,

can be described by,

16

Page 17: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

Conditions for Scheduling

• This implies that a transmission, τk, exists between the time period

rk to dk, and needs at least one time slot.

• If prek,s > 0 at time slot s, then the lifetime of a preceding

transmission is [rk – 1, dk – 1].

• If postk,s > 0 at time slot s, then the lifetime of a preceding

transmission is [rk + 1, dk + 1].

• In general, when there are finite pre- and post- hops to the

destination, the window of time that a transmission, τk, exists can

be described by,

17

Page 18: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

Conditions for Scheduling

• Since the total number of transmissions that can occur in one time

window is limited by the conflicting transmissions and number of

m channels, let be the number of transmissions in the largest set

of mutually exclusive transmissions in a window, [a, b]. qa,b is the

total number of transmissions whose lifetimes are contained

within [a, b]. Then the measure of schedulability is,

• Where the maximum number of transmissions able to be

supported in one time window is .

18

Page 19: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

Theorem 2

• For a set of flows, F, let Γs be the set of

unscheduled transmissions at slot s. If these

transmissions are schedulable, then there

exist a finite number of transmissions that are

elements of Γs. (proven in paper)

19

Page 20: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

Optimal Branch-and-Bound Scheduling

• Let the minimum number of time slots

required for scheduling be,

• And for unscheduled transmissions, the above

equation can be used as an upper bound

regarding the schedule laxity (remaining time

slots for a transmission to complete) where,20

Page 21: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

Optimal Branch-and-Bound Scheduling

• The following is the developed scheduling algorithm,

21

Page 22: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

Conflict-aware Least Laxity First

• The gateway and the nodes with high connectivity in the

network tend to experience significantly higher degrees of

conflicts, and WirelessHART must be aware of these

conflicts.

• Based upon these occurrences, the authors develop

Conflict-aware Least Laxity First (C-LLF) which

considers the length of time windows that allow

transmission scheduling, as well as any potential conflicts

that the transmission may experience.

22

Page 23: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

Conflict-aware Least Laxity First

• For a transmission, , let Λ be the set of deadlines of

transmissions that involve a node u handling .

• Since conflicting transmissions must be scheduled

during different time slots, at slot s, we may assess the

transmission priority based on the current time slot,

window length, b, and anticipated conflicts (other

transmissions at node u during [s,b]), σ,

23

Page 24: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

Conflict-aware Least Laxity First

• The conflict aware laxity at time slot s is

defined as,

• In other words, the smaller the lambda, the

more critical the transmission is!

24

Page 25: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

Conflict-aware Least Laxity First

25

Page 26: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

Evaluation

• The proposed algorithms are compared with

several real-time scheduling algorithms:

Deadline Monotonic (DM), Earliest Deadline

First (EDF), Proportional Deadline Monotonic

(PD), Earliest Proportional Deadline First

(EPD) and Least Laxity First (LLF)

26

Page 27: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

Evaluation

• Using a C code implementation on a Mac OS

X with 2.4 GHz Intel Core 2 Duo, an

evaluation of each algorithm is carried out

with the following parameters,

27

Page 28: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

Evaluation – Branch and Bound Scheduling vs Scheduling Upper Bound (UP) and Optimal Algorithm

28

Page 29: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

C-LLF vs Other Algorithms

29

Page 30: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

C-LLF vs Number of Nodes

30

Page 31: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

C-LLF vs Number of Routes

31

Page 32: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

C-LLF + B&B vs Other Algorithms

32

Page 33: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

Test-Bed Results @ Washington University in St. Louis

33

Page 34: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

Summary

• WirelessHART is applicable in sensor/actuator networks.

• Communication is TDMA and controlled/scheduled by the gateway

using various algorithms.

• Because of the TDMA operation, we must carefully schedule

transmissions based on expected delivery times for each transmission

while considering the traffic of transmissions passing through a node

(Branch and Bound + Conflict Aware Least-Laxity First)

• The proposed algorithm was compared with several common real-

time scheduling algorithms, and in general, exhibited improved

performance.

34

Page 35: Real-Time Scheduling for WirelessHART Networks by Abusayeed Saifullah, You Xu, Chenyang Lu, and Yixin Chen A Presentation of Findings for CSE5095 Joshua

End of Presentation

Questions?

35