Upload
thomhert-siadari
View
106
Download
1
Embed Size (px)
DESCRIPTION
My presentation in NSL Seminar. The paper is not mine, you could download in IEEE website
Citation preview
Low-overhead Uplink Scheduling Through Load Prediction for WiMAX real-time servicesW. Nie, H. Wang, N. Xiong2IET Commun., 2011, Vol. 5, Iss. 8, pp. 1060–1067
Thomhert Suprapto SiadariDept. IT Convergence
Kumoh National Institute of Technology
February 3rd., 2012
Wireless Embedded & Networking System Laboratory
DOC ID
Outline
Introduction
Problems & Solutions
WiMAX Sample Frame
WiMAX Service Classes
Low-overhead Scheduling
Simulation Results
Conclusion & Future Works
DOC ID
Introduction
IEEE802.16
WiMAX 300 trials worldwide
Connection oriented
PHY & MAC Layer
Suffers problem of huge MAC overhead
No scheduling Algorithm standard
DOC ID
Problems and Solutions
Problems:
1. Large overhead uplink scheduling
2. Real-time services
3. Scheduling algorithm
Solutions/ Contributions:
4. Low-overhead uplink scheduling
5. Load prediction
DOC ID
WiMAX Sample Frame
PMP (BS to MSs)
Transmission: Downlink & Uplink
TDD
DOC ID
WiMAX Service Classes
WiMAX service classes:
Unsolicited Grant Service (UGS) fixed-size data packets
Real-time polling service (rtPS) generate variable-size data packets periodically
Non-real-time polling service (nrtPS) bandwidth not on the basis of fixed packet size
Best Effort (BE) efficient service (web surfing)
DOC ID
Low-Overhead Uplink Scheduling
- Earlier Deadline First (EDF) scheduling
- Adaptive Bandwidth Scheduling
DOC ID
EDF Scheduling
- Information Module
- Scheduling Database Module
- Service Assignment Module
f : frame size (ms), uplink and downlink subframe contains;
di : the maximum delay of connection i (ms);
qi(t) : the queue length of connection i at time t(bit);
si [t, t + f ]: the number of bits required to be transmitted for connection i in the time interval [t, t + f ];
ai[t, t + f ]: the number of bits arriving for connection i in the time interval [t, t + f ];
Ndi[t, t + f ]: the number of bits waiting in the queue for connection i, which will expire in the time interval [t, t + f ].
DOC ID
EDF Scheduling
- Selects SSs based on delay requirement
- Suitable for real-time services
- Deadline to each packet
- Allocate bandwidth to SS based on earliest deadline
Information Module Firstly delay requirement
rtPS connection input information module:
Output:
Secondly expiration time
DOC ID
EDF Scheduling
Scheduling Database Module serves as a database of information for all connections
Service Assignment Module
- Determine uplink subframe allocation in terms of the number of bits per SS
DOC ID
EDF Scheduling
Specific implementation steps:
Check BWrtps & Bufferi_deadline (bandwidth required by the deadline frame in cureent time
If
Guarantee the bandwidth of deadline packets
Allocate more bandwidth to active SS
If
The bandwidth requirement will be scheduled:
DOC ID
Low Overhead Algorithm
C: the uplink channel capacity;
F: set of all SSs belonging to the rtPS class;
Bi: bandwidth allocated to connection i;
Dequeue i: remove packet P from the queue of connection i;
amount(P,): retrieve the packets P from the connection i. Convert the packets to number of symbols according to the signal-to-interference noise ratio [SINR(ji)] of connection i.
CreateIE(amount(P, ji)): create an IE for connection i with
amount(P, ji) number of symbols. Then, IE is added to the UL-MAP message.
Drop(rtPS): drop packets from the queues for all connections.
DOC ID
Adaptive Scheduling Scheme
- Modeling the Arrival Process
PDF:
CDF of inter-arrival time:
- Estimation of Time: predict the response time when BS allocates the bandwidth to SS
- Tr = Reuest time
- Ti = Bandwidth response time
- Estimation of Time
DOC ID
Adaptive Scheduling Scheme
Adaptive time slots calculating:
To calculate expected bandwidth:
To calculate required time slots
Given buffer calculate required time slot
Si(0,1) smooth parameter give ratio of the actual allocation bandwidth to previous predictions and requirements
If ε > 1 calculated bandwidth is closer to predicted bandwidth
If ε < 1 calculated bandwidth is closer to requested bandwidth
So, use ω = 0.05 to adjust Si.
DOC ID
Simulation Results
DOC ID
Conclusion & Future Work
- Better performance than WFQ & WRR
- Sharply reduce MAP & MAC SDUs subheader overhead
- Improves system throughput
DOC ID
Yang harus dijawab setelah baca paper
- Problem yg ada itu apa?
- Solusi dan kontribusi yang ditwarkan apa?
- Metodenya? LOH: EDF & Adaptive sched schem?
- EDF utk apa sebenarmya? Ad a 3 module disini? Information module? Sched database module? Service assignment module? Specific implementation steps?
- Adaptive sched scheme: modeling Arrival process? estimation time? Adaptive time slots calculating? apa tujuannya semua ini?
- Simulasi frame ultilisation, average throughput, average queuing delay, packet loss? kenapa dalam real-time communication harus pake ini? Alasannya?
- Dia pake perbandingan WFQ dan WRR? kenapa? Dan hasilnya lebih baik? Kenapa? Ada apa dengann WFQ dan WRR?
- Dia kan pake load prediction? Kalo WFQ dan WRR pake load prediction juga gimana?