1. Packet Classification using Tuple Space Search
Author: V. Srinivasan, S. Suri, G. Vargheset
Publisher: ACM SIGCOMM 1999
Presenter: Sih-An Pan
Date: 2013/11/27

2. Tuple Space Search
Our scheme is motivated by the observation that while filter databases contain many different prefixes or ranges, the number of distinct prefix lengths tends to be small.
Suppose we have a filter database FD with N filters, and these filters result in m distinct tuples.
Since m tends to be much smaller than N in practice, even a linear search through the tuple set is likely to greatly outperform the linear search through the filter database.

3. Tuple Space Search
We can define a tuple for each combination of field length, and call the resulting set tuple space.
By concatenating these bits in order we can create a hash key, which can then be used to map filters of that tuple into a hash table.

4. Tuple Space Search
The port numbers, however, are often specified using ranges, and the number of bits specified is not clear.
To get around this, we define the length of a port range to be its nesting level.
RangeID
Level 1 2

5. Tuple Pruning Algorithm
Packet = {0000, 01*}
Source Trie
Destination Trie
Source Tuple List = {[0,0], [0,4], [2,3], [4,1], [4,4]}
Destination Tuple List = {[0,0], [4,1], [3,2]}
Intersected Tuple List = {[0,0], [4,1]}

6. Tuple Pruning Algorithm

7. Markers and Precomputation
Consider a tuple 𝑇 𝑖 = [ 𝑙 1 , 𝑙 2 , , 𝑙 𝑘 ].
𝑇 𝑗 = [ ℎ 1 , ℎ 2 , , ℎ 𝑘 ] belongs to set Short ( 𝑻 𝒊 ) if and only if ℎ 𝑖 ≤ 𝑙 𝑖 ,for all 𝑖=1,2,…,𝑘 ,and 𝑇 𝑖 ≠ 𝑇 𝑗
𝑇 𝑗 = [ ℎ 1 , ℎ 2 , , ℎ 𝑘 ] belongs to set Long ( 𝑻 𝒊 ) if and only if ℎ 𝑖 ≥ 𝑙 𝑖 ,for all 𝑖=1,2,…,𝑘 ,and 𝑇 𝑖 ≠ 𝑇 𝑗
A tuple 𝑇 𝑗 , where 𝑇 𝑖 ≠ 𝑇 𝑗 , that is neither in Short( 𝑻 𝒊 )nor in Long( 𝑻 𝒊 ) belongs to the incomparable set IC( 𝑻 𝒊 ).

8. Markers and Precomputation
The main idea is that a probe into a tuple 𝑇 𝑖 can be used to eliminate a subset of the tuple space.

9. Markers

10. Precomputation

11. Rectangle Search The search terminates when the we reach the
rightmost column or the
first row.
Since there are W
rows and W columns,
the number of probes
needed is at most 2W - 1

12. Rectangle Search

13. Storage Efficiency

14. Throughput (Worst Case)

15. Throughput (Avg. Case)