Convergence of Wired and Wireless Networks

1
Convergence of Wired and Wireless Networks

• Sushil Singh, Viet Do, Kevin James, Kenneth Yun,
  and Rene Cruz
• UC San Diego

2
Research Goal

• Identification of solutions for ubiquitous access
  of the Internet from any digital mobile or
  stationary device at high-speed
• Leverage existing infrastructure
• Use value-added solution when necessary
• Connect the following devices to the Internet
• Residential home/small business PCs
• Cell phones with data capability
• PDAs

3
Wired Network

• Hierarchical connections of optical rings and
  meshes.
• Internet core Sonet
• Metro optical rings Sonet/Ethernet
• Campus area optical ring/mesh Ethernet
• Cable/DSL optical ring/mesh

• Pro
• Transmission at very high-speed (2.5 40 Gbps)
• Cons
• Requires stationary access points
• Access available only at strategic points

4
Wireless Network

• Cell to base station wireless
• Base station to central office (CO) wireless or
  wired
• Infrastructure independently constructed from
  wired networks today

• Pro
• Almost ubiquitous access provided that cell
  sites exist everywhere
• Cons
• Low transmission speed
• High cost of transmission/cell sites

5
Observations

• Optical fiber is an ideal transmission medium for
  medium-to-long distance air is a lousy one
  although useful for very short distance
• Is it possible to exploit strengths of both
  media?
• The answer is YES!

6
A Potential Solution

• Provide low-cost fiber access points (FAPs)
  everywhere optical fibers are laid
• Instead of just hot spots in commercial centers
• Each FAP in close proximity from the next
• FAPs are not super cell sites
• FAPs are low-cost, low-power femto sites
• Mobile/stationary users access the Internet via
  FAP

7
Conceptual Access Diagram
FAP
FAP
Fiber
802-11a
802-11b
8
System Requirements

• Mobile/stationary devices
• 802-11b/a/g
• FAP
• Matching 802-11
• Antennas
• Modulators/demodulators
• DSPs
• Security/billing processor
• Add/drop multiplexor (ADM) with an extensive QoS
  capability

9
Unique Requirements

• Unlike stationary access of cable or DSL
  infrastructure, FAP access is inherently dynamic
• Must be able to manage bursty traffic in the
  fiber
• If the underlying protocol is
• TDM, need the granularity down to VT level
• IP, need extensive statistical multiplexing
  requires fine-grained QoS

10
Research Proposal

• Phase I study of fine-grained QoS requirements
  for FAP
• Identify killer applications
• Devise and simulate necessary algorithms
• Demonstrate end-to-end solution
• Phase II study of security requirements and
  distributed billing processor requirements

11
Questions and Challenges
FAP

• Is O-E-O conversion necessary?
• Do signals need to be terminated, i.e., de-framed
  and framed again?
• Is buffering required? If so, how much?
• Is it possible to splice traffic without OEO
  conversion or buffering?

12
Test Bed Implementation
13
Test Bed Setup
54G 802.11g wireless router
video server
PC1
statistics
SW
FAP0
2.5 Gbps optical ring
FAP3
FAP1
FAP2
video server
SW
PC0
PC2
14
Test Traffic

• Received by FAP3
• Streaming video from video server 1 (PC1)
• Passed on to and displayed on MAC0
• UDP from PC2 via FAP1
• Filtered and dropped by FAP3
• UDP from PC2 via FAP2
• Filtered and dropped by FAP3
• UDP from PC0 via FAP2
• Filtered and dropped by FAP3

15
Test Traffic

• Received by FAP1
• Streaming video from video server 0 (PC0)
• Passed on to and displayed on MAC1
• UDP from PC2 via FAP2
• Filtered and dropped by FAP1
• UDP from PC0 via FAP3
• Filtered and dropped by FAP1
• UDP from PC1 via FAP0
• Filtered and dropped by FAP1

16
Dest MAC0 Src PC1 via FAP0
PC1
video
SW
MAC0
FAP0
FAP1
FAP3
54G
FAP2
SW
PC0
PC2
17
Dest FAP3 Src PC2 via FAP1
PC1
SW
MAC0
FAP0
FAP1
FAP3
54G
54G
data
FAP2
SW
PC0
PC2
18
Dest FAP3 Src PC2 via FAP2
PC1
SW
MAC0
FAP0
FAP1
FAP3
54G
54G
FAP2
data
PC0
SW
PC2
19
Dest FAP3 Src PC0 via FAP2
PC1
SW
MAC0
FAP0
FAP1
FAP3
54G
54G
FAP2
data
PC2
PC0
SW
20
Dest MAC1 Src PC0 via FAP2
PC1
SW
MAC0
FAP0
FAP3
54G
FAP1
FAP2
video
PC2
PC0
SW
21
Dest FAP1 Src PC2 via FAP2
PC1
SW
MAC0
FAP0
FAP3
54G
54G
FAP1
FAP2
data
PC0
SW
PC2
22
Dest FAP1 Src PC0 via FAP3
PC1
SW
MAC0
FAP0
FAP3
FAP1
54G
data
FAP2
PC2
PC0
SW
23
Dest FAP1 Src PC1 via FAP0
PC1
data
SW
MAC0
FAP0
FAP1
FAP3
54G
54G
FAP2
SW
PC0
PC2
24
Queuing Model

• Global earliest deadline first (EDF) scheduling
• Deadline assignment done in input nodes
• Output queue in each destination node
• Four priority levels

25
Hardware Implementation

• Xilinx Vertex II Pro FPGA
• Queuing and scheduling
• Line (2.5 Gbps serial link) interface (optical
  ring side)
• Ethernet interface (LAN side)
• External SRAM buffer interface
• SRAM buffer for packet storage

26
FAP Board
27
Latency Measurements

• High-priority streaming video has lower latency
  than other lower-priority UDP traffic

28
Dynamic Latency Measurement
29
Long-Term Average Latency