Name Lookup in Named Data Networking(NDN) Bin Liu Email: [email protected] Dept of Computer Science and Technology Tsinghua University Dec 7,2013

Name Lookup in Named Data Networking(NDN)

Bin Liu

Email: [email protected]

Dept of Computer Science and TechnologyTsinghua University

Dec 7,2013

Name Lookup in NDN HotICN

——Outline

1. Introduction to NDN Name Lookup

2. Name Lookup Evaluation Criteria

3. Name Lookup: Algorithms and Data Structure ---NSDI’13

---INFOCOM2013

---ICDCS2012+Computer Networks’13

---CCNx/Hashing

4. NDN Name Lookup TestBench

5. Conclusion and Future Work


——Outline




---INFOCOM2013


---CCNx/Hashing



Name Lookup in NDN HotICN—— Introduction: NDN

Namelookup

Named Data Networking (NDN) Named Data Networking is proposed recently as the

clean-slate network architecture for future Internet, which no longer concentrates on “where” the information is located, but “what” the information (content) is needed.

NDN uses names to identify every piece of contents instead of IP addresses for hardware devices attached to IP network.

Name Lookup in NDN HotICN——Introduction: NDN

Namelookup

Naming in NDN An NDN name is hierarchically structured and composed of

explicitly delimited components

/cn/edu/tsinghua/cs/s-router/~liubin/research/papers

cn edu tsinghua

Interest and Data Packets in NDN

cs s-router ~liubin research papers

6/30

Interest Packet

Data Packet

Client

Content ProviderContent Store

Replica

Name-based forwarding/routing In-network caching Multi-path transmission/delivery

——Introduction: NDN Namelookup


Namelookup

A forwarding table consists of name prefixes. A core challenge and enabling technique in

implementing NDN is exactly to perform name lookup for packet forwarding at wire speed.


Namelookup

Three kinds of tables in an NDN router, each has different function


Namelookup

Packet Forwarding Process

Name Lookup in NDN HotICN——Introduction: NDN Namelookup

Two High-level Requirements for NDN Name Lookup:

1) Longest name Prefix Matching(LPM)

2) Strict latency requirement (<100us)

T


Namelookup

The detailed challenges of name lookup Complex name structure 1) consists of digits and characters;

2) variable length name;

3) without an externally imposed upper bound. The large-scale name table


Namelookup

Backbone IP Table Size Number of Active Web-site Worldwide

400k Prefixes

100M Active Domain Names

Name tables could be 2~3 orders of magnitude larger than IP lookup table


Namelookup

The detailed challenges of name lookup Complex name structure 1) consists of digits and characters;

2) variable length name;

3) without an externally imposed upper bound.

The large-scale name table (2~3 orders larger ) Frequently update Wire Speed (100Gbps Ethernet, OC-3072)


Namelookup

IPv4 Prefixes Name Prefixes

203.84.197.*/24 /com/yahoo/news

74.125.*.*/16 /com/google/news

166.111.4.*/24 /cn/edu/tsinghua/cs/s-router/wy/papers

123.196.117.*/24 /cn/cae/www/html/main/col66

A quick comparison between IP lookup table and NAME lookup table


——Outline




---INFOCOM2013


---CCNx/Hashing




——Namelookup Evaluation Criteria

1. Throughout Wire speed lookup (100Gbs_GE, 160Gbps_OC-3072) Gbps MSPS(Million Search per Second)

2. Memory efficiency hundreds of millions of entries each entry has tens, even hundreds of characters

3. Update network topology changes and routing policy modifications one new type of FIB updates----contents published/deleted

4. Search latency (< 100us)

5. Scalability


——Outline




---INFOCOM2013


---CCNx/Hashing



Name Lookup in NDN HotICN—NSDI’13: Algorithm & Data Structure

Character Trie Two-Dimensional State Transition Table (STT)

Name Table/a/bc//ab/c/a//ab/

06

2 3

71

8

4

9

5a

b

/cb/

/ c /Character Trie

… / … a b c …0 11 2 62 33 44 556 77 88 99…


Two-Dimensional State Transition Table (STT) Advantage

Easy to build Fast speed: One State Transition needs one memory

access only Disadvantage

Too much memory required to be implemented


STT Address Transition… …

1045 /,10011046 /,1003

… …1096 b,9991097 a,9981098 /,9971099 b,10021100 /,10041101 c,10511102 c,1053… …

1000

999

… / … a b c …0 11 2 62 33 44 55…

Aligned Transition Array

1000+a=1097

Aligned Transition Array (ATA) to compress STT

offset + character’s ASCII code

999+b=1097

998

998+b=1096


ATA Advantage

Keep fast speed: one state transition needs one memory access

Low memory space Disadvantage

Building speed is too slow for large-scale name table Cannot support incremental updates


Multiple Stride Character Trie


4

1 2

50

6

3

ab

/bca/

/ c/2-stride Character Trie

ATAAddress Transitio

n01 a/,……

24879 ab,……

“a/” = 24879

d-stride character trie, every state may have 28d transitions at most.

ATA cannot support multiple Stride Character Trie


Multiple Stride Character Trie


4

1 2

50

6

3

ab

/bca/

/ c/2-stride Character Trie

Address Transition

012 bc,…

Multi-ATAAddress Transitio

n01 a/,…23 ab,…456

“a/” Mod 7=1

“ab” Mod 7=3

“bc” Mod 7=1

“bc” Mod 3=2


MATA Advantage

Improve lookup throughput: one state transition consumes multiple characters, and each state transition requires only one memory access

Further compress memory space Small ATAs in MATA are easier to build and manage Support fast incremental update

Name Lookup in NDN HotICN—NSDI’13: Name Lookup Engine

Implementation

Name Table

CharacterTrie

Aligned Transition

Array

Update Forwarding

CPU

Name Lookup Engine

Name Trace

PCIe

PCIe

GPU

NVIDIA GeForce GTX590 GPU board

Name Lookup in NDN HotICN—NSDI’13: Latency Optimization

Name Table

CharacterTrie

Aligned Transition

Array

Update Forwarding

CPU

Name Lookup Engine

Name Trace

PCIe

PCIe

GPU

Name Batch

A

B

Name Lookup in NDN HotICN——NSDI’13: Experimental

Results

Platform: A commodity PC

Name Table Download from DMOZ website: 3M Crawl from Internet: 10M

Name Trace Average workload: random name prefix + suffix Heavy workload: the longest 10% name prefix + suffix


Results

Memory Space 3M name table

ATA vs STT: 101× MATA vs STT: 130 ×

10M name table ATA vs STT: 102× MATA vs STT: 142×


Results

Lookup Speed (Million Searches per Second, MSPS) 10M, Average Workload 10M, Heavy Workload


Results

Scalability Lookup speed Memory Latency


Results

Update

More than 30K insertion per second

Nearly 600K deletion per second


——NSDI’13: Conclusion

1. MATA is proposed to compress memory space and improve name lookup speed

2. Implement a wire speed name lookup engine based on a commodity PC installed with a GTX590 GPU board

3. Extensive experiments demonstrate: Name lookup speed: 63.52 MSPS,>127 Gbps wire-speed Latency: <100us Memory: compress >140× Good Scalability

Open Website: http//www.namelookup.org


——Outline




---INFOCOM2013


---CCNx/Hashing



Name Lookup in NDN HotICN—INFOCOM2013_NameFilter:Algorithm & Data Structure

NameFilter: Basic FrameworkFirst Stage: Bloom filters for ascertaining

the prefix length of a name

/com/google/maps/…/

prefixes:

name:

Bloom filters:

…

1-th 2-th 3-th B-th

…

…Match Vector:

Second Stage: Hash Table for ascertaining forwarding ports

/com/google/maps/prefix:

Prefix Next Ports List

…

Prefix Hash Table:

Next Port

prefix: /com/google/maps/

[5] Sarang Dharmapurikar, P. Krishnamurthy, and D. E. Taylor, “Longest prefix matching using bloom filters,” in Proceedings of ACM SIMCOMM’03, 2003.

Name Lookup in NDN HotICN—NameFilter:Algorithm & Data Structure

NameFilter: Improved FrameworkFirst Stage: Bloom Filters for ascertaining

the prefix length of a name

/com/google/maps/…/

prefixes:

name:

Bloom filters:

…

1-th 2-th 3-th w-th

…

…Match Vector:

Second Stage: Merged Bloom Filters for ascertaining forwarding port(s)

/com/google/maps/prefix:

…

Merged Bloom Filters:

Next Ports

Name Lookup in NDN HotICN—NameFilter: Algorithm & Data Structure

NameFilter: Improved Framework Merged Bloom Filter

1 0 0 1 0 1 1 1 0 0 0 0 1

1 0 1 0 1 1 0 0 0 1 0 0 1

1-th Bloom filter:

11..100..001..110..000..111..110..010..000..001..100..001..011..1

Hash functions (prefix) Hash functions (prefix)

11..1 & 11..1 & 10..0 & 11..1 => 10..0

Hash functions (prefix)

N-th Bloom filter:

Mergeand

Transpose1 0 0 1 0 1 1 1 0 0 0 0 12-th

Bloom filter:

Hash functions (prefix)

…

Merged Bloom filter:

Memory access times is reduced from k to 1

Name Lookup in NDN HotICN——NameFilter: Experimental

Results

Platform: A commodity PC

Name Table Download from DMOZ website: 3M Crawl from Internet: 10M

Name Trace Average workload: random name prefix + suffix Heavy workload: the longest 10% name prefix + suffix


Results

Lookup Speed (Million Searches per Second, MSPS) One processing thread

NameFilter: 16× of Character Trie 1.8× of Bloomhash


Results

Lookup Speed (Million Searches per Second, MSPS) 24 processing threads

NameFilter 11.7× of Character Trie


Results

Memory Space

Compared with Character Trie, NameFilter saves more than 77% memory space.

Compared with BloomHash, NameFilter saves more than 65% memory space.


Results

Scalability Lookup speed Memory


Results

Update

More than 3M insertion per second

Nearly 3.4M deletion per second


——Outline




---INFOCOM2013


---CCNx/Hashing



Name Lookup in NDN HotICN—Component Coding : Algorithm & Data Structure

/com/google/map/china

/10/10/0

The idea of Name Component Encoding (NCE)


1

2

9

3

7

A

D

com

cn

4

5

8

B

E

yahoo

google

baidu

google maps

map

news

maps

maps

6uk

level-1 level-5level-2 level-4level-3

Csina

Name Pointer/com/yahoo …/com/yahoo/news …/com/yahoo/maps/uk …/com/google …/com/google/maps …/cn/google/maps …/cn/sina …/cn/baidu …/cn/baidu/map …

< yahoo,1> <google, 2> <baidu,1> <sina, 2> <google, 3>

<baidu,1> <google, 2> <google, 3> <sina, 3> <yahoo, 1>

<baidu,1> <google, 4> <sina, 3> <yahoo, 1>


Global-Code Allocation Mechanism


Local-Code Allocation Mechanism


—Component Coding： Implementation

GPU-based Implementation


—Component Coding： Experimental Results

Experimental Result: Memory

Name

Table

Average Name Length

TCT

(MByte)

# of total

Comp.

# of unique Comp.

NCE (MByte)Comress

Ratio

(TCT vs. NCE)

CHT OSTA 总数

3M 21.27 370.87 6223396250570

660.14 60.72

120.86

67.41%

10M 24.481310.0

21263789

4775676

2186.16 343.47

529.63

59.57%



Experimental Result: Throughput



Experimental Result: Update Compared to MATA

Insertion speed: 30× Deletion speed: 1.8×


——Outline




---INFOCOM2013


---CCNx/Hashing



Name Lookup in NDN HotICN—CCNx/hashing: Algorithm & Data Structure

CCNx applies hash table to construct the FIB Lookup process in CCNx

Generates all candidate prefixes of a name Sorts all candidate prefixes in a descendent order Looks up each candidate prefix against the hash table


—CCNx/hashing: Comments and Further Work

Comments: The name search speed in CCNx is much slow; The name lookup in CCNx prototype is basically for the

demonstration purpose

Future Improvements: Improve hash table lookup speed: perfect hash table Compress hash table to reduce memory consumption

[1] Yi Wang, Dongzhe Tai, Ting Zhang, Jianyuan Lu, BoyangXu, Huichen Dai and Bin Liu. Greedy Name Lookup for Named Data Networking. Sigmetrics 2013 Poster.


——Outline




---INFOCOM2013


---CCNx/Hashing




—NDN Namelookup TestBench


——Outline




---INFOCOM2013


---CCNx/Hashing




——Conclusion

1. Name Lookup is a complex search work

2. We have conducted several Name lookup Algorithms and Data Structures to achieve inspiring results

3. Our results are much better than the ones in CCNx prototype

4. The experimental results demonstrate the feasibility towards building high speed NDN routers

5. Name Lookup is also widely used in a broad range of technological fields


——Future Work

1. Larger scale datasets/traces experiments

2. Improving the TestBench

3. Extending the lookup core engine to a whole router

system to examine the performance

4. Improvements on Algorithms and Data Structures


Thank you!

Q & A

Documents

Name Lookup in Named Data Networking(NDN) Bin Liu Email: [email protected] Dept of Computer Science and Technology Tsinghua University Dec 7,2013