Distributed Throughput Optimization for ZigBee ClusterTree Networks.

(AN IEEE JOURNAL PAPER, MARCH 2012)

By: I V Surendra Varun Kumar 1PI11SCS05 2nd Sem, M.Tech (CSE)

Under the Guidance of: Mr. Chidambara K Professor, Dept. of CSE PESIT, Bangalore

Distributed Throughput Optimization for ZigBee ClusterTree Networks.

Contents

ZigBee & Properties. Devices, States & Networks ZigBee. System Architecture. Problem Formulation. Algorithm. Properties of Algorithms.

1PI11SCS05 | Varun Kumar.I

Jan 2012 - Jun 2012

2 /26

Distributed Throughput Optimization for ZigBee ClusterTree Networks.

ZigBee & Properties

Recent advances in wireless communication has strong impact on development of wireless sensor networks. 802.15 Standards Committee Wireless Personal Area Network ( WPAN ). 802.15.4

Standards Committee ZigBee.

Unique communication standard designed for LRWPAN. It has extremely low Complexity, Cost, Power consumption. Used in inexpensive, portable, and mobile devices. It requires only about 10% of the software of 1PI11SCS05 | Varun 3 Jan 2012 - Jun 2012 Kumar.I typical Bluetooth or Wireless Internet /26 a

Distributed Throughput Optimization for ZigBee ClusterTree Networks.

ZigBee & PropertiesSmall in size. USER Application Reliable and self healing. Framework Alliance N/W / Security Layer Supports large number of nodes. Easy to deploy. MAC Layer IEEE PHY Layer Very long battery life. 802.15.4 Protocol Stack Secure - Maintain an access control list - Use symmetric cryptography Can be used globally. Used in home automation systems, remote control and monitoring systems, and health care devices, etc.. 1PI11SCS05 | Varun 4Kumar.I Jan 2012 - Jun 2012 /26

Application/Profiles

Distributed Throughput Optimization for ZigBee ClusterTree Networks.

Device Types: ZigBee coordinator(ZC) Initialization,

Devices, States & Networks.maintenance & control function.

ZigBee Router (ZR) Forwards

and routes the sensed data to sink.

ZigBee End Device (ZED)

States: Active State. Inactive State.1PI11SCS05 | Varun Kumar.I Jan 2012 - Jun 2012 5 /26

Distributed Throughput Optimization for ZigBee ClusterTree Networks.

Network Topologies:

Devices, States & Networks.

1PI11SCS05 | Varun Kumar.I

Jan 2012 - Jun 2012

6 /26

Distributed Throughput Optimization for ZigBee ClusterTree Networks.

System Architecture.

IEEE 802.15.4 defines the physical layer (PL) and medium access control (MAC) layer for low-rate wireless personal area networks (LR-WPANs). IEEE 802.15.4 defines a super frame structure that begins by transmitting a beacon issued by a PAN coordinator. The process consist of

ACTIVE Period: Coordinator and the devices communicate each other. INACTIVE Period: Enter into Low-Power phase. 1PI11SCS05 | Varun 7Kumar.I Jan 2012 - Jun 2012 /26

Distributed Throughput Optimization for ZigBee ClusterTree Networks.

System Architecture.

An active portion comprise of 16 time slots, and divided into 3 parts.A

Beacon A Contention Access Period (CAP) A Contention Free Period (CFP)

A beacon is transmitted by the coordinator at slot-0. Followed by CAP, during this period it can transmit non time-critical messages and mac commands. Followed by CFP, during this period Guaranteed time slot -(GTS) is provided 1PI11SCS05 | Varun 8 Jan 2012 Jun 2012 Kumar.I /26 exclusively for transmission.

Distributed Throughput Optimization for ZigBee ClusterTree Networks.

System Architecture.The device that require GTS in the next super frame of CFP, they send GTS requests to the coordinator. ZEDs are directly connected to ZC in star topology. In cluster-tree & mesh topology communication is done in multi-hop fashion through ZR. 1PI11SCS05 | Varun 9Kumar.I Jan 2012 - Jun 2012 /26

Distributed Throughput Optimization for ZigBee ClusterTree Networks.

Problem Formulation.

The vertex-constraint flow network can be formulated as a directed graph G=(V,E)V

- routers in the network E communication link b/w pair of routers

Each vertex v V is associated with nonnegative capacity (v)0. Each directed edge (u,v) is associated with an implicit capacity c(u,v) = if (u,v) E c(u,v) = 0 if (u,v) EJan 2012 - Jun 2012 10 /26

1PI11SCS05 | Varun Kumar.I

Distributed Throughput Optimization for ZigBee ClusterTree Networks.

Problem Formulation.

Two vertices are distinguished in network:

a source s (sender) a sink t (receiver)

Let f be the flow in network G as f: VVR, where f(u,v) is the net flow from vertex u to vertex v. A flow network that satisfies the following 3 properties Capacity Skew Flow

u V {f(u,v) | f(u,v) >0} (v), v V. f(u,v) = -f(v,u), u,v V.

constraint:

symmetry:

1PI11SCS05 | Varun u V Kumar.I

conservation:

Jan v - 2012 f(u,v) = 0, 2012 VJun {s,t}.

11 /26

Distributed Throughput Optimization for ZigBee ClusterTree Networks.

Problem Formulation.

Value of flow f is defined as the total net flow into the sink. uV f(u,t)Given network G with s snd t, objective is to find maximum flow from s to t in G.

1PI11SCS05 | Varun Kumar.I

Jan 2012 - Jun 2012

12 /26

Distributed Throughput Optimization for ZigBee ClusterTree Networks.

Algorithm.

Pull-Push-Relabel (PPR) algorithm is proposed to maximize flow in ZigBee clustertree network. Residual capacity with respect to f: in in

vertex, f(v) = (v) uV {f(u,v) | f(u,v)>0} edge, cf(u,v) = c(u,v) f(u,v)

Residual graph Gf is defined as (V,Ef)Ef =

1PI11SCS05 | Varun Kumar.I

Jan 2012 - Jun 2012

13 /26

Distributed Throughput Optimization for ZigBee ClusterTree Networks.

Algorithm.

Total net flow into a vertex v is defined as e(v) = uV f(u,v). A vertex vV {s,t} is said to be over flowing if e(v) > 0. Height function h: V N for f is defined as h(s)=|V|, h(t) = 0 and h(u) h(v) + 1 for every residual edge (u,v) Ef.

1PI11SCS05 | Varun Kumar.I

Jan 2012 - Jun 2012

14 /26

Distributed Throughput Optimization for ZigBee ClusterTree Networks.

Algorithm.PULLIn a PULL(u,v) operation, a lower vertex u pulls the flow of a higher vertex v downward to itself. Conditions to be satisfied: v is overflowing, i.e., e(v) > 0; There is an edge from v to u in G, i.e., c(v,u) = ; The residual capability of u is positive, i.e., f (u) > 0; 1PI11SCS05 | Varun 15 Jan 2012 - Jun 2012 Kumar.I /26 u is lower than v by 1, i.e., h(v) = h(u) + 1.

Distributed Throughput Optimization for ZigBee ClusterTree Networks.

Algorithm.PULLProcedure 1. PULL(u,v) Applicability: e(v) > 0, c(v,u) = , (u) > f 0, and h(v) = h(u) + 1 Action: u pulls = min( e(v), (u)) units of f flow from v 1: f(v,u) f(v, u) + 2: f(u, v) -f(v, u) 3: e(v) e(v) - 4: e(u) e(u) +Jan 2012 - Jun 2012 1PI11SCS05 | Varun 16Kumar.I /26

Distributed Throughput Optimization for ZigBee ClusterTree Networks.

Algorithm.PUSHIn a PUSH(u,v) operation, a higher vertex u pushes the over-pulled flow back to a lower vertex v along the edge (v,u)G. Conditions to be satisfied: u is overflowing, i.e., e(u) > 0; There is no edge from u to v in G, i.e., c(u,v) ; There is a positive net flow on (v,u), i.e., c f (u, v) > 0; 1PI11SCS05 | Varun 17 Jan 2012 - Jun 2012 Kumar.I u is higher than v by one, i.e., h(u) = h(v) /26 +

Distributed Throughput Optimization for ZigBee ClusterTree Networks.

Algorithm.PUSHProcedure 2. PUSH(u,v) Applicability: e(u) > 0, c(u,v) , c (u, v) f > 0, and h(u) = h(v) + 1 Action: u = min ( e(u), c (u,v) ) units of f flow to v 1: f(u,v) f(u,v) + 2: f(v,u) -f(u,v) 3: e(u) e(u) - 4: e(v) e(v) +Jan 2012 - Jun 2012 1PI11SCS05 | Varun 18Kumar.I

/26

Distributed Throughput Optimization for ZigBee ClusterTree Networks.

Algorithm.RELABEL It enables a vertex u to increase its height. Conditions to be satisfied:

u is overflowing, i.e., e(u) > 0; (u, v) Ef implies that h(u) h(v) for all vertices vV.

Procedure 3. RELABEL(u) Applicability: e(u) > 0, and (u, v) E h(u) f h(v), v V 1PI11SCS05 | Varun 19 Jan 2012 - Jun 2012 Action: u increases its height to 1 + min{h(v) |/26 Kumar.I

Distributed Throughput Optimization for ZigBee ClusterTree Networks.

Algorithm.PULL-PUSH-RELABEL Procedure 4. PULL-PUSH-RELABEL(u)1: for all v Adj(u) do 2: PULL(u,v) 3: if u cannot be pulled by any other vertex then 4: for all v Adj(u) do 5: PUSH(u,v) 6: RELABEL(u)

1PI11SCS05 | Varun Kumar.I

Jan 2012 - Jun 2012

20 /26

Distributed Throughput Optimization for ZigBee ClusterTree Networks.

Algorithm.Procedure 5. INIT(u) 1: if u is source s then 2: h(u) |V| 3: for all v Adj(u) do 8: else if u Adj(s) then 4: h(v) 0 9: h(u) 0 10: e(u) 0 5: e(v) (v) 11: for all v 6: f(u,v) Adj(u) do (v) 12: f(u,v) 0 7: f(v,u) - 13: f(v,u) 0 1PI11SCS05 | Varun 21 (v) Jan 2012 - Jun 2012 Kumar.I /26

Distributed Throughput Optimization for ZigBee ClusterTree Networks.

Algorithm.

Algorithm 1. PPR1: for all u V do 2: INIT(u) 3: while there exists any overflowing vertex do 4: for all u V do 5: PULL-PUSH-RELABEL(u)

1PI11SCS05 | Varun Kumar.I

Jan 2012 - Jun 2012

22 /26

Distributed Throughput Optimization for ZigBee ClusterTree Networks.

Properties of AlgorithmsLemma 1. During the execution of Algorithm PPR, if a vertex u V is overflowing, u performs a push or a relabel operation; otherwise, a pull operation is performed on u. Lemma 2. Whenever a vertex u performs a relabel operation on itself, its height h(u) increases by at least 1. Lemma 3. If h is initialized as a height function, then it remains a height function throughout Algorithm PPR. 1PI11SCS05 | Varun 23 Jan 2012 - Jun 2012Kumar.I /26

Distributed Throughput Optimization for ZigBee ClusterTree Networks.

Properties of Algorithms

Lemma 4. Let f be a pre-flow in G and let h be a height function of f. Then, there will not be a path from the source s to the sink t in the residual network Gf . Lemma 5. The vertex-constraint maximum flow problem can be reduced to the traditional maximum flow problem.

1PI11SCS05 | Varun Kumar.I

Jan 2012 - Jun 2012

24 /26

Distributed Throughput Optimization for ZigBee ClusterTree Networks.

Properties of Algorithms

Theorem 1 (Optimality of Algorithm PPR). When Algorithm PPR terminates, the preflow f is a maximum flow from the source s to the sink t in G. Theorem 2 (Convergence of Algorithm PPR). Algorithm PPR always terminates within 2|V|2 passes.

1PI11SCS05 | Varun Kumar.I

Jan 2012 - Jun 2012

25 /26

Thanks