Virtualization in Wireless Networks Feb. 23 th, 2009

Transcription

1 Virtualization in Wireless Networks Feb. 23 th, 2009 Heejin Lee Keum-mo Park Prof. Chong-kwon Kim

2 Outline Introduction Examples: WLAN virtualization Techniques: Wireless resource isolation Embedding problem for wireless virtual networks Conclusions 2

3 Introduction Network virtualization Decoupling network service from network infrastructure Multiple & independent network services on a single shared infrastructure Research issues To implement Virtual Network: How to slice network resources To manage Virtual Network: How to operate after slicing Wireless virtual networks The wireless extensions to the wired virtual networks Wireless virtualization can provide a simple solution for previous problems: Mobility, Security, Address space, Network management, etc. 3

4 GENI wireless sub-networks EMULAB Virtualization of wireless networks: WLAN, Mesh, 3G, WiMax, Sensor, Vehicle, etc. Integration across wired and wireless networks Programmable and cognitive radio tech. ORBIT Testbed/Emulator Sensor Networks Mobile Wireless/ Cognitive Networks Vehicular Networks 4

5 Wireless Virtualization Example: WLAN Virtualization Multiple VAPs over one physical AP One VAP made of multiple physical APs

6 Multiple VAPs over one physical AP Virtual AP Channel 6 AP B STA Channel 6 Multiple ISPs/Services using the existing APs Beacon/ Probe Response VMAC: virtual MAC address BSSID: B Rates: 1, 2, 5.5, 11 Security: SSN Enhanced resource utilization Cost-effective approach AP A Without considering resource isolation Multiple Physical APs Beacon/ Probe Response BSSID: A Rates: 1, 2, 5.5 Security: WEP Beacon/ Probe Response Channel 6 Beacon/ Probe Response BSSID: B Rates: 1, 2, 5.5, 11 Security: SSN STA AP A BSSID: A Rates: 1, 2, 5.5 Security: WEP Multiple Virtual APs 6

7 Channel blanket topology Multiple thin APs supporting a single cell controlled by a central server Interference/handoff free system Scalability problem 7

8 FatVAP [Katabi 08] Split the traffic of a MN over multiple APs Backhaul capacity aggregation using 3ms fast AP switching When the wireless link is the bottleneck, the client stick to the best AP and avoid AP switching 8

9 Fat AP emulation Kernel maps flows with interfaces according to routing info. Reverse NAT architecture replaces source IP address MN side virtualization Useless if the wireless medium is a bottleneck 9

10 Wireless Virtualization Techniques: Wireless Resource Isolation

11 Virtualization Tech.(1/4) SDMA Each experiment is assigned space Size of region controlled by transmission power, channel characteristics, etc. Transmit power control is important 11

12 Virtualization Tech.(2/4) FDMA Different experiments assigned non-interfering channels Limited number of non-interfering channels Channel switching time Can be avoided using multiple NIC cards 12

13 Virtualization Tech.(3/4) TDMA Each experiment is assigned time slots Context switching overhead Time synchronization 13

14 Virtualization Tech.(4/4) Combinatorial manner SDMA SDMA+TDMA SDMA+FDMA SDMA+TDMA+FDMA 14 14

15 Embedding Problem for Wireless Testbed

16 Virtual network embedding problem Mapping logical slices into physical substrates 16

17 Wireless testbed Examples ORBIT, Emulab, etc ORBIT Nodes are placed in Grid Nodes have Multiple Wireless Interface cards 2 transceivers per node The number of channels > the number of interfaces Slicing by FDM+TDM+SDM is recommended Referred by Tech. report from GENI 17

18 TDM based virtualization of ORBIT Performance distortion of TDM 18

19 Embedding for wireless virtualization Slice requirements for wireless testbed Various number of frequencies and interface cards Various time length and space size Goal Accommodate as many VNs as possible, while meeting the constraints of each VN s requirement 19

20 Problem description Frequency 3 NIC per node TDM Super frame VN2 VN1 VN3 VN6 VN4 VN5 VN7 VN8 VN9 VN3 VN1 VN5 VN6 VN2 VN4 VN7 VN8 Time Slot Minimize Assumptions TDM + FDM (A slice is defined by the duration of time and the number of channels) A single radio interface is allocated to each slice Similar to 2D packing problem but it must meet MSC (Maximum Slicing Constraint) Goal: minimize the super frame length 20

21 Packing problem Problem type 2D knapsack problem Strip packing problem Bin packing problem Minimize height??? VN7 VN8 Solution type Exact solution NP-complete Approximation heuristic Additional constraints The number of transceivers Segment of channel dimension < the number of transceiver VN1 VN6 VN4 VN2 VN5 VN3 21

22 Simulation results Type1 Type2 Metric TDM Super-frame length Low Bound W: the number of total channels S i : the slice i Channel Single Single Time slot Single Multiple Ch. by Time 1 by 1 1 by 3~10 The length of TDM Superframe (sec) Type1 : Type2 : Type LB LB LB Type3 Multiple Multiple 3~12 by 1~ The number of network interfaces per node 22

23 Conclusions Network virtualization in terms of future Internet allows Various future internet architectures to coexist Testbed to verify various future internet proposals Wireless virtual networks The wireless extensions A simple solution for previous problems: Mobility, Security, Address space, etc. Two ways of virtualization: WLAN examples Wireless virtual network embedding problem 23

24 Thank you! Q&A Center for Next-generation Network & Service research 24