A general purpose proxy filtering mechanism applied to the mobile environment

1. A general purpose proxy filtering mechanism applied to the mobile environment Author：Bruce Zenel From ： Wireless Networks

2. Outline • Introduction • Architecture • filters • Evaluation • Conclusion

3. Introduction(1/3) • It provides a mechanism for downloading and executing proxy programs ( called “filters” ) • Filters are often application specific, and usually act under control of an application running on a mobile host. • Filters can drop, delay, or transform data moving to and from the mobile host in order to improve the perceived quality of the network.

4. Introduction(2/3) • “last link” to the mobile host is the likely problem area. Figure 1.Proxy computing environment

5. Introduction(3/3) • The purpose of Filtering is to allow dynamic run-time, situation-specific, adjustment of design-time decisions. • The value of a proxy is that it permits client or server, or both, to remain unchanged or to be changed very little.

6. Architecture(1/2) • Proxy Server：Filtering unit. • PMICP (Proxy Mobile Internetworking Control Protocol)：Protocol which moves data to and from the Proxy Server. • Filter Control ：Enables filters to adapt to their environment.

7. Architecture(2/2) • PMICP protocol guarantees that all traffic moving to and from the Mobile Host will pass through the Proxy Server. Figure 2. Proxy Server/PMICP interaction

8. PMICP(1/2) • by keeping track of the location of the MH and using Mobile Support Routers(MSR) to direct data to its current location. • PMICP allows a MH to choose an MSP to be its Proxy MSR(PMSR). • Once chosen, the PMICP protocol guarantees that all traffic will pass through the PMSR.

9. PMICP(2/2) Figure 3. PMICP environment

10. Proxy Server(1/3) • It provides a dynamic execution environment for filters. • HLP is to filter application layer protocol, while LLP filters transport and network layer protocols. Figure 4. Block diagram of the Proxy Server

11. Proxy Server(2/3) The HLP uses the notion of filter insertion. Figure 5. Filter download and insertion

12. Proxy Server(3/3) • The Proxy Server Run-Time Environment is the execution environment in which a filter is run. • Filter execution environment can broken down into two groups： 1.Native/binary execution environments --advantage is speed of execution, and disadvantage is security. 2.Interpreted execution environments (execute within the Proxy Server)

13. Filter Control • Filters may be controlled either internally or externally. --internal control refers to explicit control from the parent application on the MH. --external control refers to an indirect form of control based on environment information

14. Filters • The protocols chosen for filtering were selected using the following criteria： 1.utilization/impact – chose protocols that are commonly used on the Internet, such as TCP,HTTP. 2.Architecture versatility – developed filters that processed data in different ways, e.g., compression, selective discard, protocol optimization. 3.evaluation – developed filters that could be measured in some way, e.g., data transfer latency and cost.

15. Evaluation • Evaluated the system both quantitatively and qualitatively. • Quantitative analysis was directed towards the overall performance of the system including the filters. • Qualitative evaluation was less precise, we looked at various issues.

16. Quantitative analysis(1/2) • Chose three filters to be evaluated ： -- HTTP, NFS, and TCP. -- because their performance could be readily measured, and they are well distributed across our architecture. -- HTTP and NFS filters use the High Level Proxy using TCP and UDP, while the TCP filter uses the Low Level Proxy.

17. Quantitative analysis(2/2) • The proxy host performs filtering, while the gateway merely provides connectivity for the client. • Three heterogeneous clients are 10Mbps Ethernet, 2Mbps Wavelan, and 33,3Kbps SLIP. Figure 6. Experimental testbed design

18. HTTP filter(1/4) Figure 7. Two proxy environment used for the HTTP filter

19. HTTP filter(2/4) • Filtering configurations were tested includes： 1.Baseline – no filtering proxy present. 2.ZLIB Compression – slow but efficient in terms of space, prefers space over time. 3.LZO Compression – is faster than ZLIB, prefers time over space.

20. HTTP filter(3/4) • Process by the HTTP filter would increase the delay seen by the client Figure 8. Ethernet testbed： Comparison of text file transfer time using HTTP

21. HTTP filter(4/4) • The SLIP link is so slow that the ZLIB filter has enough time to use its slow but better compression algorithm. Figure 9. SLIP testbed： Comparison of text file transfer time using HTTP

22. Qualitative analysis • Security -- establish trust between the proxy and the application -- runtime execution environment for filters • Proxy mobility -- reject this because too complicated • Number of proxies -- 1, 2, and N

23. Conclusions • The proxy and associated filters performed well in heterogeneous environments such as the Wavelan and SLIP cases. • It is better to filter the underlying protocols used by applications than to operate on the applications themselves.