Download presentation
Presentation is loading. Please wait.
Published byPhillip Perkinson Modified over 9 years ago
1
Optimal Fast Hashing Yossi Kanizo (Technion, Israel) Joint work with Isaac Keslassy (Technion, Israel) and David Hay (Hebrew Univ., Israel)
2
Hash Tables for Networking Devices Hash tables and hash-based structures are often used in high-speed devices Heavy-hitter flow identification Flow state keeping Flow counter management Virus signature scanning IP address lookup algorithms
3
Hash tables In theory, hash tables are particularly suitable: O(1) memory accesses per operation (element insertion/query/deletion) for reasonable load But in practice, there is a big difference between an average of 1.1 memory accesses per operation, and an average of 4 Why not only 1 memory access? Collisions
4
Hash Tables for Networking Devices 123 Collisions are unavoidable wasted memory accesses For load≤1, let a and d be the average and worst- case time (number of memory accesses) per element insertion Objective: Minimize a and d 123456789 Memory
5
Why We Care On-chip memory: memory accesses power consumption Off-chip memory: memory accesses lost on/off-chip pin capacity Datacenters: memory accesses network & server load Parallelism does not help reduce these costs d serial or parallel memory accesses have same cost
6
Traditional Hash Table Schemes Example 1: linked lists (chaining) 123456789 Memory 12 3 45 6 7 8 9
7
Traditional Hash Table Schemes Example 1: linked lists (chaining) Example 2: linear probing (open addressing) Problem: the worst-case time cannot be bounded by a constant d 123456789 Memory 12345 6 8
8
High-Speed Hardware Enable overflows: if time exceeds d → overflow list Can be stored in expensive CAM Otherwise, overflow elements = lost elements Bucket contains h elements E.g.: 128-bit memory word h=4 elements of 32 bits Assumption: Access cost (read & write word) = 1 cycle 123456789 Memory 4 7 15 3 6 28 h CAM 9
9
Possible Settings Static setting - Insertions and queries only Dynamic setting – Insertions, deletions, and queries. Generalized setting – Balancing between the buckets’ load.
10
Problem Formulation 123456789 Memory 4 7 15 3 6 28 h CAM 9 Given average a and worst-case d of memory accesses per operation, Minimize overflow rate Given average a and worst-case d of memory accesses per operation, Minimize overflow rate
11
Example: Power of d-Random Choices d hash functions: pick least loaded bucket. Break ties u.a.r. [Azar et al.] Intuition: can reach low … but average time a = worst-case time d wasted memory accesses 123456789 Memory 4 7 15 3 6 28 h CAM 9 10 11 12
12
Other Examples d-left [Vöcking] Same as d-random, but break ties to the left. Cuckoo [Pagh et al.] Whenever collision occurs, moves stored elements to their other choices. Typically, uses much more than d memory accesses on average.
13
Outline Static Case Overflow Lower Bound Optimal Schemes: SIMPLE, GREEDY, MHT. Dynamic Case Comparison with Static Case. Overflow Lower Bound Overflow Fraction Depending on d.
14
Overflow Lower Bound Objective: given any online scheme with average a and worst-case d, find lower-bound on overflow . [h=4, load=n/(mh)=0.95, fixed d] No scheme can achieve (capacity region)
15
Overflow Lower Bound Result: closed-form lower-bound formula Given n elements in m buckets of height h: Valid also for non-uniform hashes For n=m and h=1, we get simply Defines a capacity region for high- throughput hashing
16
Lower-Bound Example [h=4, load=n/(mh)=0.95] For 3% overflow rate, throughput can be at most 1/a = 2/3 of memory rate
17
Overflow Lower Bound Example: d-left scheme: low overflow , but high average memory access rate a [h=4, load=n/(mh)=0.95, m=5,000]
18
The SIMPLE Scheme SIMPLE scheme: single hash function Looks like truncated linked list 123456789 Memory 4 7 15 3 6 28 h CAM 9 10 11
19
Performance of SIMPLE Scheme [h=4, load=0.95, m=5,000] The lower bound can actually be achieved for a=1
20
The GREEDY Scheme Using uniform hashes, try to insert each element greedily until either inserted or d 123456789 Memory 4 7 15 3 6 28 h CAM 9 10 11 12 d=2
21
Performance of GREEDY Scheme [d=4, h=4, load=0.95, m=5,000] The GREEDY scheme is always optimal until a co
22
Performance of GREEDY Scheme [d=4, h=4, load=0.95, m=5,000] Overflow rate worse than 4-left, but better throughput (1/a)
23
The MHT Scheme MHT (Multi-Level Hash Table) [Broder&Karlin]: d successive subtables with their d hash functions 1234567 Memory 4 7 1 5 3 6 28 h CAM 9 10 11 1 st Subtable2 nd Subtable3 rd Subtable
24
Performance of MHT Scheme Optimality of MHT until cut-off point a co (MHT) Proof that subtable sizes fall geometrically Confirmed in simulations [d=4, h=4, load=0.95, m=5,000] Overflow rate close to 4-left, with much better throughput (1/a)
25
Outline Static Case Overflow Lower Bound Optimal Schemes: SIMPLE, GREEDY, MHT. Dynamic Case Comparison with Static Case. Overflow Lower Bound Overflow Fraction Depending on d.
26
Dynamic vs. Static Dynamic hash tables are harder to model than the static ones [Kirsch et al.] But past studies show same asymptotic behavior with infinite buckets (insertions only vs. alternations) traditional hashing using linked lists – maximum bucket size of approx. log n / log log n [ Gonnet, 1981] d-random, d-left schemes – maximum bucket size of log log n / log 2 + O(1) [ Azar et al.,1994; Vöcking, 1999] As a designer, using the static model seems natural. Even if real-life devices have finite buckets
27
Degradation with Finite Buckets Finite buckets are used. Surprising result: degradation in performance 1234 12 FiniteInfinite 1 1 2 1234 H(1) = 3H(2) = 3 Remove 1 Element “2” is lost although its corresponding bucket is empty
28
Comparing Static and Dynamic Static setting: insertions only n = number of elements m = number of buckets Dynamic setting: alternations between element insertions and deletions of randomly chosen elements. fixed load of c = n / (mh) Fair comparison Given an average number of memory accesses a, minimize overflow fraction .
29
Overflow Lower Bound Overflow lower bound of where r = ach. Also holds for non-uniformly distributed hash functions (under some constraints). The lower bound is tight (Simple, Greedy)
30
Numerical Example For h=1 and c=1 (100% load) we get a lower bound of 1/(1+a). To get an overflow fraction of 1%, one needs at least 99 memory accesses per element. Infeasible for high-speed networking devices Compared to a tight upper bound of e -a in the static case. [Kanizo et al., INFOCOM 2009] need ~4.6 memory accesses.
31
Outline Static Case Overflow Lower Bound Optimal Schemes: SIMPLE, GREEDY, MHT. Dynamic Case Comparison with Static Case. Overflow Lower Bound Overflow Fraction Depending on d.
32
Overflow Fraction Depending on d So far, we relaxed the constraint on d. We considered n elements with an average of a memory accesses, as n a distinct elements. To take into account d, we must consider each element along with its own hash values.
33
Graph Theory Approach Consider a bipartite graph. Left vertices = Elements Right vertices = Buckets (assume h=1). Edge = The bucket is one of the element’s d choices
34
Graph Theory Approach We get a random bipartite graph where each left vertex has degree d. Expected maximum size matching = Expected number of elements that can be inserted to the table, that is, a lower bound. We derived an explicit expression for d=2. Upper bound can be achieved by Cuckoo hashing (equivalent to finding maximum size matching).
35
Summary We found lower and upper bounds on the achievable overflow fraction both for the static and dynamic cases. Static models are not necessarily exact with dynamic hash tables. Improved lower bound for d=2 and a characterization of the performance of Cuckoo hashing.
36
Thank you.
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.