1: \begin{abstract}

2: We propose a new and easily-realizable distributed hash table

3: (DHT) peer-to-peer structure, incorporating a random caching

4: strategy that allows for {\em polylogarithmic search time} while

5: having only a {\em constant cache} size. We also show that a very

6: large class of deterministic caching strategies, which covers

7: almost all previously proposed DHT systems, can not achieve

8: polylog search time with constant cache size. In general, the new

9: scheme is the first known DHT structure with the following

10: highly-desired properties: (a) Random caching strategy with

11: constant cache size; (b) Average search time of $O(log^{2}(N))$;

12: (c) Guaranteed search time of $O(log^{3}(N))$; (d) Truly local

13: cache dynamics with constant overhead for node deletions and

14: additions; (e) Self-organization from any initial network state

15: towards the desired structure; and (f) Allows a seamless means for

16: various trade-offs, e.g., search speed or anonymity at the expense

17: of larger cache size. \iffalse Hence, unless totally new design

18: paradigms are invented, deterministic caching strategies do not

19: seem to be able to

20: outperform this design.\\

21: Constant cache size, allows for constant overhead for entering and

22: departure of nodes into the network. No previously known DHT

23: structure can achieve this with less than $\Theta(log(N))$

24: overhead. \\

25: The randomized nature of the design will allow for a high degree

26: of flexibility and dynamics, making the network very hard to track

27: and attack. The caching scheme is totally decentralized and does

28: not depend much on the availability of certain nodes. A totally

29: distributed algorithm for continuous update of cache contents are

30: also provided and the global convergence of any initial network

31: towards the desired structure is proved.\\

32:  Other generalizations

33: and modifications for trading off the performances is also

34: provided. \fi

35: \end{abstract}

36: