1 Filename: 155-four-hidden-service-improvements.txt
2 Title: Four Improvements of Hidden Service Performance
5 Author: Karsten Loesing, Christian Wilms
8 Implemented-In: 0.2.1.x
12 25-Sep-2008 Initial proposal for or-dev
16 A performance analysis of hidden services [1] has brought up a few
17 possible design changes to reduce advertisement time of a hidden service
18 in the network as well as connection establishment time. Some of these
19 design changes have side-effects on anonymity or overall network load
20 which had to be weighed up against individual performance gains. A
21 discussion of seven possible design changes [2] has led to a selection
22 of four changes [3] that are proposed to be implemented here.
26 1. Shorter Circuit Extension Timeout
28 When establishing a connection to a hidden service a client cannibalizes
29 an existing circuit and extends it by one hop to one of the service's
30 introduction points. In most cases this can be accomplished within a few
31 seconds. Therefore, the current timeout of 60 seconds for extending a
32 circuit is far too high.
34 Assuming that the timeout would be reduced to a lower value, for example
35 30 seconds, a second (or third) attempt to cannibalize and extend would
36 be started earlier. With the current timeout of 60 seconds, 93.42% of all
37 circuits can be established, whereas this fraction would have been only
38 0.87% smaller at 92.55% with a timeout of 30 seconds.
40 For a timeout of 30 seconds the performance gain would be approximately 2
41 seconds in the mean as opposed to the current timeout of 60 seconds. At
42 the same time a smaller timeout leads to discarding an increasing number
43 of circuits that might have been completed within the current timeout of
46 Measurements with simulated low-bandwidth connectivity have shown that
47 there is no significant effect of client connectivity on circuit
48 extension times. The reason for this might be that extension messages are
49 small and thereby independent of the client bandwidth. Further, the
50 connection between client and entry node only constitutes a single hop of
51 a circuit, so that its influence on the whole circuit is limited.
53 The exact value of the new timeout does not necessarily have to be 30
54 seconds, but might also depend on the results of circuit build timeout
55 measurements as described in proposal 151.
57 2. Parallel Connections to Introduction Points
59 An additional approach to accelerate extension of introduction circuits
60 is to extend a second circuit in parallel to a different introduction
61 point. Such parallel extension attempts should be started after a short
62 delay of, e.g., 15 seconds in order to prevent unnecessary circuit
63 extensions and thereby save network resources. Whichever circuit
64 extension succeeds first is used for introduction, while the other
67 An evaluation has been performed for the more resource-intensive approach
68 of starting two parallel circuits immediately instead of waiting for a
69 short delay. The result was a reduction of connection establishment times
70 from 27.4 seconds in the original protocol to 22.5 seconds.
72 While the effect of the proposed approach of delayed parallelization on
73 mean connection establishment times is expected to be smaller,
74 variability of connection attempt times can be reduced significantly.
76 3. Increase Count of Internal Circuits
78 Hidden services need to create or cannibalize and extend a circuit to a
79 rendezvous point for every client request. Really popular hidden services
80 require more than two internal circuits in the pool to answer multiple
81 client requests at the same time. This scenario was not yet analyzed, but
82 will probably exhibit worse performance than measured in the previous
83 analysis. The number of preemptively built internal circuits should be a
84 function of connection requests in the past to adapt to changing needs.
85 Furthermore, an increased number of internal circuits on client side
86 would allow clients to establish connections to more than one hidden
89 Under the assumption that a popular hidden service cannot make use of
90 cannibalization for connecting to rendezvous points, the circuit creation
91 time needs to be added to the current results. In the mean, the
92 connection establishment time to a popular hidden service would increase
95 4. Build More Introduction Circuits
97 When establishing introduction points, a hidden service should launch 5
98 instead of 3 introduction circuits at the same time and use only the
99 first 3 that could be established. The remaining two circuits could still
100 be used for other purposes afterwards.
102 The effect has been simulated using previously measured data, too.
103 Therefore, circuit establishment times were derived from log files and
104 written to an array. Afterwards, a simulation with 10,000 runs was
105 performed picking 5 (4, 6) random values and using the 3 lowest values in
106 contrast to picking only 3 values at random. The result is that the mean
107 time of the 3-out-of-3 approach is 8.1 seconds, while the mean time of
108 the 3-out-of-5 approach is 4.4 seconds.
110 The effect on network load is minimal, because the hidden service can
111 reuse the slower internal circuits for other purposes, e.g., rendezvous
112 circuits. The only change is that a hidden service starts establishing
113 more circuits at once instead of subsequently doing so.
117 [1] http://freehaven.net/~karsten/hidserv/perfanalysis-2008-06-15.pdf
119 [2] http://freehaven.net/~karsten/hidserv/discussion-2008-07-15.pdf
121 [3] http://freehaven.net/~karsten/hidserv/design-2008-08-15.pdf