Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
등록 안내
운영위원회
행사장 약도
SUMMER WORKSHOP ON COMPUTER COMMUNICATIONS
하계컴퓨터통신 워크샵(SWCC 2011)
•운영위원장 : 김기천 (건국대)•프로그램 위원장 : 정 송 (KAIST)•프로그램 위원회 부위원장 : 고영배 (아주대), 유명식 (숭실대)• 프로그램 위원 : 고석주(경북대), 김동균(경북대), 김범준(계명대),
김승욱(서강대), 박재성(수원대), 박준상(홍익대), 백상헌(고려대), 이정륜(중앙대), 임유진(수원대), 정영욱(광운대), 최선웅(국민대), 최영준(아주대), 한연희(한국기술교육대), 허노정(동양대), 황호영(한성대), 정윤원(숭실대), 조성래(중앙대)
•등록 : 이상환 (국민대)•출판 : 이정륜 (중앙대)•회계 : 김영용 (연세대)•현장 : 김종덕 (부산대), 이혁준 (광운대), 유명식 (숭실대)•홍보 : 최용훈 (광운대)
|일 시|2011년8월25일(목)–8월27일(토)|장 소|부산대학교과학기술연구동|주 관|한국정보과학회정보통신소사이어티|주 최|한국정보과학회정보통신소사이어티|후 원| 한국통신학회통신네트워크연구회,(사)OSIA,
한국정보처리학회정보통신응용소사이어티,(주)KT,IT기계융합연구센터,부산대차세대물류IT사업단
●부산대학교과학기술연구동
●교통편1. 김해공항에서 오실 때: 버스 307 좌석버스 → 동래역 하차 → 지하철 1호선(
노포동 방향) 부산대역(3번 출구) → 셔틀버스 → 학생회관 앞 하차2. 부산역에서 오실 때 : 지하철 1호선 부산역 → 부산대역 → 셔틀버스 → 학
생회관 앞 하차3. 노포동 시외 버스 터미널에서 오실 때 : 지하철 1호선 노포동역 → 부산대역
→ 셔틀버스 → 학생회관 앞 하차4. 서부 시외 버스 터미널에서 오실 때 : 지하철 2호선 사상역 → 서면역 하차
→ 1호선(노포동 방향) 부산대역 → 셔틀버스 → 학생회관 앞 하차
●숙박숙박은 개인 별로 아래 연락처를 이용하여 예약 하셔야 합니다. - 상남국제회관 : 051-510-7000 (부산대학교라고 말한 후, 숙박을 예약하여야 할인 금액으로 숙박이 가능합니다), 일반 객실 1박 47,500원, 비지니스 객실 1박 56,000원
- 센텀 호텔 : 051-720-9000 (부산대학교라고 말한 후, 숙박 예약) 1박 80,000원
- 농심 호텔 : 051-550-2100 (부산대학교라고 말한 후, 숙박 예약), 1박 139,150원
- 씨클라우드 호텔 : 051-993-1000 (부산대학교라고 말한 후, 숙박 예약 / 씨클라우드의 경우, 부산대 교수님의 명함이 필요하오니, 꼭 소지하시고 가셔야 합니다), 1박 80,000원
SUMMER WORKSHOP ON COMPUTER COMMUNICATIONS
SWCC 2011하계컴퓨터통신 워크샵
SUMMER WORKSHOP ON COMPUTER COMMUNICATIONS
하계컴퓨터통신 워크샵(SWCC 2011)
●사전등록기간
•2011년 8월 23일(화요일) 까지
●등록비
구분 사전등록 현장등록
정보과학회 회원
일반 250,000원 270,000원
학생 150,000원 170,000원
비회원일반 270,000원 290,000원
학생 170,000원 190,000원
●입금계좌
•입금은행 : 우리은행 •입금계좌 : 1002-535-765410•예 금 주 : 이상환 •담 당 자 : 김형진 (TEL : 02-910-5219 / HP : 010-3846-2553)
●등록방법
• 등록신청서 작성 후 메일로 송부 (김형진, [email protected])• 등록신청서는 (http://sigin.or.kr/swcc11)의 하단에서 다운로드• 세금 계산서 발행 가능 • 현금 결제 요망 (※카드 결제의 경우 등록신청서 작성 후 현장
에서 결제)
초대의 말씀
안녕하십니까? 올해에도 한국정보과학회 정보통신소사이어티에서는 SWCC(Summer Workshop on Computer Communication) 2011을 개최하고자 합니다. 올해에는 특히 여름의 마지막 아쉬움을 달랠 수 있는 부산에서 최근 화두로 등장하고 있는 미래 인터넷 기술, 컨텐츠 네트워킹 기술, 그린 네트워킹 기술, 및 Cyber Physical Systems에 대한 보안등의 초청강연과 연구 논문 발표 및 최신 연구결과를 깊이 있게 논의할 수 있는 신진연구자들의 연구발표를 준비하였습니다. 또한 관련 분야의 우수한 논문 40여편의 논문발표가 있을 예정입니다. 이번 행사를 준비하는데 많은 도움을 주신 여러 위원님들께 깊
은 감사의 말씀을 드립니다. 여러 위원님들의 수고가 없었다면 훌륭한 프로그램이 나올 수 없었을 것입니다.한국정보과학회 정보통신소사이어티에서 매년 개최하는 SWCC
2011에 소사이어티 회원님들은 물론 정보통신 관련 산업, 연구, 학계 등의 각 분야에 종사하시는 분과 이 분야에 관심있는 분들의 많은 참여를 통하여 좋은 정보교류의 장이 되기를 기원합니다.
2011년 8월정보통신소사이어티 회장 김 영 한SWCC 2011 운영위원장 김 기 천
SWCC 2011 프로그램위원장 정 송
http://www.pusan.ac.kr/KOR_PNUS/html/01_intro/intro_11_01.asp
SUMMER WORKSHOP ON COMPUTER COMMUNICATIONS
하계컴퓨터통신 워크샵(SWCC 2011) 프로그램SUMMER WORKSHOP ON COMPUTER COMMUNICATIONS
하계컴퓨터통신 워크샵(SWCC 2011) 프로그램SUMMER WORKSHOP ON COMPUTER COMMUNICATIONS
하계컴퓨터통신 워크샵(SWCC 2011) 프로그램
8월25일 (목요일)
12:00 – 13:00 등록
초청강연 1 (13:00-13:40)
인터넷의 미래기술과 그 대응방안방송통신위원회 이영희 미래인터넷 PM
신진 연구자 세션 1 (13:40-15:00)
13:40-14:20Personal Content Networking
이의진 교수 (KAIST)
14:20-15:00모션제어 시스템의 정밀도 분석 및 구현
김강희 교수 (숭실대)
휴식 (15:00 – 15:15)
신진 연구자 세션 2 (15:15 - 16:35)
15:15-15:55Robust Design of Modern Communication Networks
이향원 교수 (KAIST)
15:55-16:35WiFi 2.0: Next Generation WiFi Service over Spectrum Whitespaces
김효일 교수 (UNIST 울산과학기술대학교)
휴식 (16:35-16:50)
개회식 및 시상식 (16:50 ~ 17:20)
초청강연 2 (17:20-18:00)
Security for Cyber Physical Systems미네소타대 김용대 교수
8월 26일(금요일)
논문 세션 1 (09:30 – 11:30)
S1-1
미래 인터넷 페더레이션을 위한 개방형 어댑터 설계 및 구현장인선, 이기원, 백상헌 (고려대)
인터넷과 콘텐츠 기반 네트워크 연동 아키텍처 연구민성만, 고한얼, 김영현, 백상헌 (고려대)
A Spam SMS Filtering System Using Various Preprocessing and A-priori Algorithm
유환일, 채동규 (한양대)
실내 환경의 무선 데이터 채널 확보를 위한 TV White Space의 활용최기운, 최영준 (아주대)
Congestion-Dependent Pricing in Heterogeneous Wireless Access Networks
이주현, 이융, 정송 (KAIST)
Comparative Analysis of LISP Mapping Systems이재경, 김재인, 고석주 (경북대)
Device Proximity Analysis and Cooperative GPS Localization곽정호 (KAIST), 모정훈 (연세대), 정 송 (KAIST)
A Gossip Protocol for Churn Rate Problem in P2P Multimedia Streaming Networks
Thinh Nguyen, Duong Tuan Nguyen, 김영한 (숭실대)
S1-2
Semi-Automatically Generating Ontology and Auto Reflected for Creative User Opinion about the Creative Writing Support System Development for the Storywriter나성준, 김남진 (Mogencelab), 이상범 (단국대), 최이권 (Mogencelab,
신동렬 (성균관대)
MANET에서 경로 축적을 이용한 부분적 다중 경로 복구 개선 방안김진선, 최종원 (숙명여대)
무선 센서 네트워크에서 계층적 주소 기반 TCP ACK 병합 기법신동승, 이성원 (경북대), 이상국 (ADD), 김동균 (경북대)
모바일 컨텐츠 중심 네트워크에서의 접근 지연 시간 향상 기법김영현, 고한얼, 민성만, 백상헌 (고려대)
다중 커버러지와 연결성을 위한 최적의 센서 배치 패턴김용환, 김찬명, 한연희 (한국기술교육대)
The compressed packet transmission for efficiency of 6LoWPAN김민수, 김희수, 최대인, 강현국 (고려대)
전술이동통신체계를 위한 신뢰성 있는 환경정보 기반 인증 기법배병구, 고영배 (아주대)
유사 Proportional Fair 스케쥴링 기법안진수, 김영용 (연세대)
중식 (11:30 – 13:00)
논문 세션 2 (13:00 – 14:30)
S2-1
Caching Policy Considering One-time Content in CCN이연주, 김광수, 김혜림, 노병희 (아주대)
대용량 응용 데이터 분산 처리를 위한 클라우드 컴퓨팅 구축 연구윤준원, 곽재혁, 정용환, 김주현, 함재균, 박동인 (KISTI)
P2P based PTT Service over WiFi-DirectTuan-Hao Tieu, Duong-Tuan Nguyen, 김영한 (숭실대)
가시광 무선 통신 시스템의 실내 통신 채널 품질 분석황준호, Trong Hop Do, 유명식 (숭실대)
센서 네트워크에서 지역별 노드 수를 고려한 랜덤 키 사전 분배 방법유성재, 김성일 (숭실대)
Wi-Fi Direct 디바이스의 네트워크 구성시간에 관한 분석김정근, 이덕규 (경희대)
S2-2
수직자기기록 채널에서 RLL 부호를 접목한 LDPC 부호의 성능김진영, 이재진 (숭실대)
무선 Ad Hoc 네트워크에서 브로드캐스트 전달률 향상을 위한 중계 노드 선택 방법
안지형, 이태진 (성균관대)
D-OFDM 기술을 기반으로 다중 채널을 활용하는 인지 무선 네트워크에서의 MAC 프로토콜
이민규, 이태진 (성균관대)
모바일 VoIP 음성 보안통신 시스템 디자인조재만, 김형국 (광운대)
Mobile VoIP환경에서 음질향상을 위한 수신단측 음성패킷 손실보완 및 병합 알고리즘
서광덕, 김형국 (광운대)
Mobile IPTV 환경에서 KDC를 이용한 DCAS 인증기법조인희, 이희찬, 신용태 (숭실대)
휴식 (14:30 – 14:45)
논문 세션 3 (14:45 – 16:15)
S3-1
단일 인터페이스 기반 전투무선망에서의 채널 적응 기법서 윤, 고영배 (아주대)
스트리밍 서비스의 사용자 체감품질 향상을 위한 단계적 비디오 품질 조절 기법
김형준, 오승준, 정광수 (광운대)
A Procedure for Interoperability between OMA SCE and OMA SRM
이희찬, 조인희, 신용태 (숭실대)
소프트웨어 안전성분석을 강화하기 위한 법제도 개선방안 연구이상지, 이재용, 지정은, 신용태 (숭실대)
차세대 전술이동통신체계 메쉬 네트워크를 위한 다중서버 기반 상호 인증 기법
손유진, 손태식, 고영배 (아주대)
SMP8654 임베디드 보드 기반 디지털 사이니지 미들웨어 설계 및 구현남영진, 이종태, 박영균, 정순환 (대구대)
S3-2
ISP간의 Shapley-like 이득분배구조에 의한 협력과 분리장혜령, 이효정, 조정우, 이융 (KAIST)
중앙형 이동성 관리 구조와 분산형 이동성 관리 구조의 성능 비교 분석정성진, 장지원, 김영한 (숭실대)
분산형 이동성 관리 구조에서의 이동 멀티스크린 서비스 적용 방안장지원, 김영한 (숭실대)
IPv6 기반 센서 네트워크 구성 및 시험김재경, 홍연화, 김영한 (숭실대)
Data Fusion using Rao-Blackwellized Particle Filter김도형 (한양대)
Maneuvering Target Tracking using a Rao-Blackwellized Multiple Model Particle Filter
김도형 (한양대)
8월 27일(토요일)
워크샵요약 및 전체토의 (10:00 - 12:00)
2011년도 하계컴퓨터통신 워크샵
- 127 -
Comparative Analysis of LISP Mapping Systems
Jae-Gyeong Lee, Ji-In Kim and Seok-Joo Koh
School of Computer Science and Engineering, Kyungpook National University
Abstract The Locator Identifier Separation Protocol (LISP) has recently been made for enhancement of routing scalability. One of
the critical issues on LISP is to design an effective mapping system for Endpoint Identifier (EID) and Routing Locator
(RLOC). Until now, several schemes have been proposed for LISP mapping system, each of which has its own features.
This paper compares those schemes for LISP mapping system by numerical analysis in terms of the mapping operation
cost. From the numerical results, it seems that hybrid push/pull approaches, such as LISP-ALT and LISP-CONS, provide
better performance than the only pull-based or push-based schemes.
1. Introduction
The Locator-Identifier Separation Protocol (LISP) [1] was
proposed for enhancement of routing scalability. LISP gives a
lot of benefits by separating the IP address into separate spaces
for Endpoint Identifiers (EIDs) and Routing Locators (RLOCs).
One of the primary issues on LISP is to design an effective
mapping system to manage and distribute the mappings between
EIDs and RLOCs. Several schemes for mapping system have
been proposed so far, and each system has its own feature. In
this paper, we review the existing schemes for LISP mapping
systems and then compare them in terms of the operational costs
associated with mapping query by numerical analysis
The paper is organized as follows. Section 2 gives a brief
overview of existing LISP mapping systems. In Section 3, we
analyze the candidate schemes for comparison. Section 4 shows
the numerical results. We conclude this paper in Section 5.
2. Overview of Candidate Mapping Schemes
Depending on how to manage the EID-RLOC mappings, the
LISP mapping system can be classified into pull-based, push-
based, and hybrid schemes. In the pull-based scheme, an Ingress
Tunnel Router (ITR) gets the RLOC information for an EID by
sending a Map Request to a mapping database. On the contrast,
in the „push-based‟ approach, the mapping system distributes all
mapping information to TRs.
Table 1 compares the existing mapping schemes. Only the
LISP-NERD [2] can be seen as the push-based mapping system.
LISP-EMACS [3] and LISP-DHT [4, 5], LISP-TREE [6] are the
examples of the pull-based scheme. The „hybrid‟ scheme is to
combine the push and pull operations, which includes LISP-
CONS [7] and LISP-ALT [8, 9].
Table 1. Summary of Existing Mapping Systems
System Distribution Model Propagated Information
LISP-NERD Push Entire Mapping
LISP-EMACS Pull -
LISP-DHT Pull -
LISP-TREE Pull -
LISP-CONS Hybrid Push/Pull EID-Prefix
LISP-ALT Hybrid Push/Pull EID-Prefix
A. LISP-NERD
The NERD (Not-so-novel EID RLOC Database) scheme is the
simplest mapping system, which is based on a monolithic
database containing all EID-RLOC mappings and managed by a
central authority. This mapping system is a push-based approach,
since all TRs receive the entire mappings from the authority. In
this aspect, NERD offers the advantage of reducing signaling
overhead for map query operations. However, each ITR needs
to store all of the existing mappings, even though many of them
will not be used. For this reason, the size of mapping table may
gets large, which can deteriorate the burden of TR routers. It can
also take a long time in the bootstrapping operation.
B. LISP-EMACS
The EMACS (EID Mappings Multicast Across Cooperation
Systems) scheme is based on a complex multi-tree infrastructure
and the bidirectional PIM multicast protocol [10]. An ITR joins
an appropriate multicast group to get the mapping information.
The Map Request is multicast to the group, and the concerned
ETR will respond directly to the requesting ITR.
C. LISP-DHT
The Distributed Hash Table (DHT) scheme uses an overlay
network based on „Chord.‟ In this system, each Chord node has
a unique k-bit Chord ID, and the whole Chord IDs are organized
as a ring. Each Chord node must maintain correct successor and
predecessor node pointers on the ring. If an ITR sends a Map
Request to a DHT node, the DHT node first looks up its DHT
mapping table. If the mapping is not found, the Map Request is
forwarded to another node, as per the specified DHT rule. When
the mapping is found, the concerned ETR will respond back to
the originating ITR.
D. LISP-TREE
TREE is a hierarchical mapping system that separates the
storage of mappings and their discovery. The mapping storage is
managed by ETRs, while the discovery mechanism with a Map
Resolver (MR) is installed on top of the DNS protocol. The
main role of discovery mechanism is to provide the list of ETRs
associated with the requested EID. Once MR obtains the
corresponding list of authoritative ETRs, it will send a Map
Request to one of them and receives the Map Reply. Then, it
2011년도 하계컴퓨터통신 워크샵
- 128 -
forwards the Map Reply to the requesting ITR.
E. LISP-CONS
CONS (Content distribution Overlay Network Service) uses the
mapping system with a hierarchical database for the ETRs. This
is a hybrid push/pull approach. In the push operation, the EID
prefixes are delivered to the root of a tree. In the pull operation,
when an ITR needs the RLOC of a specific EID, it sends a
query for the EID. The query message is forwarded up over the
tree hierarchy, until the mapping query is resolved. Once the
ETR is found, it responds to the requesting ITR.
F. LISP-ALT
The ALT (Alternative Logical Topology) uses an overlay
network, which is based on the General Routing Encapsulation
(GRE) tunnels among BGP routers to advertise the EID-Prefixes
of ETR. Each ITR does not have a full mapping table. Instead, it
sends a Map Request message to ALT over GRE tunnel, and the
message is delivered the correspondent ETR via BGP routers.
Then, ETR delivers the EID-RLOC mapping directly to ITR.
3. Cost Analysis
For performance analysis, the total costs (TC) for mapping
system are divided into Delivery Cost (DC) and Processing Cost
(PC). DC is the cost for exchanging sending and receiving the
control between entities so as to get the mappings, whereas PC
is the cost for searching the associated mapping table to find
corresponding mapping entry in the database.
In this paper, we do not analyze the LISP-NERD scheme,
since this scheme uses a totally different approach and tends to
give the worst performance. LISP-TREE has the two different
modes: iterative and recursive. In this paper, we consider only
the recursive mode, since it is reported that the recursive mode
gives better performance than the iterative mode [6]. In
particular, we consider the only two-level tree hierarchy for
equal comparison in LISP-TREE.
We define the parameters used for analysis in Table 2.
Table 2. Parameters used for costs analysis
Parameter Description
HMS-Search Hop counts required for searching in Mapping System
HITR-ETR Hop counts between ITR and ETR
HTR-MS Hop counts between TR and Mapping System (MS)
Nhost Number of host managed by ETR
Nprefix Number of EID prefixes managed by MS router
α Unit lookup cost for EID prefix by MS router
β Unit lookup cost for EID-RLOC mapping by ETR
τ Unit transmission cost per hop
Figure 1 shows the network model for cost analysis, in which
each cost parameter is indicated, as described in Table 2.
Figure 1. Network model for cost analysis
1) LISP-EMACS
In LISP-EMACS, when an ITR needs the mapping information
for an EID, it sends a Map Request to its nearest MS router that
manages the associated multicast group. After that, the router
advertises the Map Request to all the ETRs within the multicast
group. For analysis, we assume that the multicast delivery cost
is proportional to HTR-MS x HMS-Search. The mapping information
is found, ETR will respond to the ITR with a Map Reply. Thus,
we can represent the DC of LISP-EMACS as follows.
DCLISP-EMACS = Scontrol x τ (HTR-MS+HTR-MS x HMS-Search + HITR-ETR).
The PC of LISP-EMACS depends on the number of hosts
managed by ETR and the number of ETRs in MS. Accordingly,
the PC of LISP-EMACS can be represented as follows
PCLISP-EMACS = HMS-Search x β log (Nhost).
So, we can get TCLISP-EMACS = DCLISP-EMACS + PCLISP-EMACS.
2) LISP-DHT
In LISP-DHT, when ITR needs the mapping information, it
sends Map Request to its nearest Chord node, which takes HTR-
MS. Then, the Map Request should be delivered to the Chord
node managing the destination node, which depends on the
DHT searching algorithm and approximately takes 2 x HMS-Search.
After that, the receiving Chord node returns a Map Reply
directly to the originating ITR, which is HITR-ETR. Then, we can
obtain the DC of LISP-DHT as follows
DCLISP-DHT = Scontrol x τ (HTR-MS + 2 x HMS-Search + HITR-ETR).
The PC of LISP-DHT is only required for destination Chord
node to look up the mapping table. Thus, PC can be
PCLISP-DHT = β log (Nhost).
Now, we get TCLISP-DHT = DCLISP-DHT + PCLISP-DHT.
3) LISP-TREE
In LISP-TREE, we should consider the costs for the discovery
operation in the tree hierarchy. The DC of LISP-TREE consists
of the following factors. When ITR needs the mapping
information, it sends a Map Request to the nearest MS router
with a Map Resolver, which takes HTR-MS routers. If the MS
2011년도 하계컴퓨터통신 워크샵
- 129 -
router does not have the corresponding EID-RLOC mapping, it
will send a Map Request to the LISP-TREE Sever (LTS), which
takes HMS-Search. We consider the only two-level tree hierarchy.
That is, the hop count between LTS and MS is one for simple
comparison. If the corresponding EID-RLOC mapping is found,
the LTS will respond with a Map Reply to ITR via MS over the
tree hierarchy. Thus, the DC of LISP-TREE is represented as
DCLISP-TREE = 2 x Scontrol x τ (HTR-MS + HMS-Search + 1).
The PC of LISP-TREE is required for the lookup operation
by LTS and the mapping storage of TR. So, we can derive the
PC of LISP-TREE as follows
PC LISP-TREE = α log (Nprefix) + 2 β log (Nhost).
Thus, we can get TCLISP-TREE = DCLISP-TREE + PCLISP-TREE.
4) LISP-CONS
In LISP-CONS scheme, we consider the only two-level tree
hierarchy for equal comparison with the other schemes. In the
LISP-CONS scheme, a qCAR (querying Content Access
Resource) generates a request message on behalf of ITR, and
aCAR (answering Content Access Resource) stores the EID-
RLOC mapping information and gives a Map Reply to the peer
qCAR. In addition, a CDR (Content Discovery Resource) stores
the EID-prefixes information. Then, the DC of LISP-CONS
consist of the following components: 1) the delivery cost of
Map Request between TRs and MS, which is HTR-MS; 2) the
searching cost within MS, which is HMS-Search; 3) the delivery
cost of Map Reply from aCAR to qCAR, which is HITR-ETR.
Thus, we can get the DC of LISP-CONS as
DCLISP-CONS = Scontrol x τ (2 x HTR-MS + HMS-Search + HITR-ETR).
The PC of CONS is proportional to the number of nodes
within MS, excluding the two CARs. Each CDR of MS contains
the EID-Prefixes information that was aggregated from aCAR.
So, we can represent the PC of LISP-CONS as follows.
PCLISP-CONS = α (HMS-Search-2)log(NprefixxHMS-Search)+2βlog(Nhost).
So, we get TCLISP-CONS = DCLISP-CONS + PCLISP-CONS.
5) LISP-ALT
In LISP-ALT scheme, we assume that both ITR and ETR
function as the nodes of Mapping System (MS) routers. Thus,
the DC of LISP-ALT consists of the following components: 1)
the delivery time of Map Request from ITR to ETR over MS,
which takes HMS-Search; 2) the delivery time of Map Reply from
ETR to ITR, which is HITR-ETR. By considering the control
message of Scontrol and the unit transmission cost of τ, the DC of
LISP-ALT can be represented as
DCLISP-ALT = Scontrol x τ (HMS-Search + HITR-ETR).
The PC of LISP-ALT can be obtained as follows. In this
scheme, we need to look up the mapping table of intermediate
ALT routers on the path to ETR in MS, which takes α log (Nprefix
x HMS-Search) by using a tree-based data structure in the mapping
table. Then, the final ETR will look up its EID-RLOC mapping
table, which takes β log (Nhost). Thus, the PC of LISP-ALT can
be represented as
PCLISP-ALT = HMS-Search x α log (Nprefix x HMS-Search) + β log (Nhost).
Accordingly, we get TCLISP-ALT = DCLISP-ALT + PCLISP-ALT.
4. Numerical Results
For numerical analysis, we set the parameter values, as shown
in Table 3. The values in the table are the same or similar to
those given in [11]. Among these parameter values, HMS-Search,
Nhost, τ may depend on a variety of network conditions. Thus,
we need to compare the total cost for different values of those
parameters.
Table 3. Parameter values used for cost analysis
Parameter Default Min Max
HMS-Search 10 5 20
HITR-ETR 2
HTR-MS 1
Nhost 10 10 100
Nprefix 3
α 2
β 3
τ 2 1 20
Figure 2 compares the total costs of candidate schemes for
different transmission cost over wired link. In this figure, it is
shown that the total costs increase as the unit of transmission
cost of wired link (τ) gets larger. We can see that LISP-TREE
scheme has the worst performance. This is because LISP-TREE
has to perform the discovery operations along the tree hierarchy.
On the other hand, we can also see that LISP-ALT provides the
best performance for mapping systems. This is because LISP-
ALT uses the hybrid scheme of push-based and pull-based
operations effectively. The gap of performances between LISP-
ALT and the other schemes tend to get larger, as τ increases.
Figure 2. Total cost for different wired transmission cost
Figure 3 shows the impacts of the hop counts in the Mapping
System. From the figure, we can also see that the total cost of
LISP-ALT is the lowest among candidate schemes. It is also
noted that the hybrid schemes, LISP-ALT and LISP-CONS,
2011년도 하계컴퓨터통신 워크샵
- 130 -
give better performance than the other pull-based schemes,
since those hybrid schemes propagate the EID prefixes of hosts
to the adjacent routers in advance, which is helpful to reduce the
overhead of the subsequent Map Request operations. Other than
those two hybrid schemes, LISP-EMACS gives relatively better
performance, since in this scheme a Map Request tends to take
less hop counts from the root router to ETRs. On the contrast,
LISP-TREE and LISP-DHT show relatively worse performance.
The performance gaps between these two schemes and the other
three schemes get larger, as the hop count within the Mapping
System increases. This is because LISP-TREE and LISP-DHT
depend on the transmission delays in the Mapping System.
Figure 3. Total cost for different number of hops in MS
Figure 4 describes the impacts of the number of hosts
managed by ETR. From this figure, we can see that the total
costs are not nearly affected by the number of hosts for all the
schemes. In the figure, it is shown that LISP-ALT and LISP-
CONS are very effective schemes. On the one hand, LISP-
TREE and LISP-DHT provide relatively worse performance,
compared to other schemes.
Figure 4. Total cost for different number of host in ETR
5. Conclusions
In this paper, we analyze and compare several candidate
schemes for LISP mapping system. The existing mapping
schemes are categorized into the three approaches: push-based,
pull-based, and hybrid schemes. LISP-NERD is the only pull-
based approach. LISP-EMACS, LISP-DHT and LISP-TREE are
the pull models in which the Map Request for query is used to
find the EID-RLOC mapping information. On the other hand,
LISP-CONS and LISP-ALT can be seen as the hybrid schemes
with both push and pull operations.
From numerical analysis, it is shown that the hybrid approach,
LISP-ALT and LISP-CONS, provide better performance than
the only push-based or pull-based schemes. This is mainly
because the hybrid schemes propagate the EID prefixes of hosts
to the mapping systems in advance (push-based operation) and
this is very helpful to reduce the overhead of the subsequent
Map Request operations (pull-based operation).
Acknowledgment
This research was supported by Basic Science Research
Program of NRF(2010-0020926),
ITRC program of NIPA(NIPA-2011-C1090-1121-0002), and
IT R&D support program of KCA(KCA-2011-10913-05004).
References
[1] D. Farinacci, et al., Locator/ID Separation Protocol (LISP),
IETF Internet-Draft, draft-ietf-lisp-14. June 2011
[2] E. Lear, NERD: A Not-so-novel EID to RLOC database,
IETF Internet-Draft, draft-ietf-lisp-nerd-08, Marh 2010.
[3] S. Brim, et al., EID Mappings Multicast Across Cooperating
Systems for LISP, IETF Internet-Draft, draft-curran-lisp-
emacs-00, November 2007.
[4] L. Mathy, et al., “LISP-DHT: Towards a DHT to map
identifiers onto locators,” ACM CoNEXT Conference,
December 2008.
[5] H. Luo, et al., “A DHT-based Identifier to locator Mapping
Approach for a Scalable Internet,” IEEE Transactions on
Parallel and Distributed Systems, Vol. 20, No. 12, December
2009.
[6] L. Jakab, et al., “LISP-TREE: A DNS Hierarchy to Support
the LISP Mapping System,” IEEE JSAC, Vol. 28, No. 8,
pp1332-1343, August 2010.
[7] S. Brim, et al., LISP-CONS: A Content distribution Overlay
Network Service for LISP, IETF Internet-Draft, draft-meyer-
lisp-cons-04, April 2008.
[8] D. Farinacci, et al., LISP Alternative Topology (LISP-ALT),
IETF Internet-Draft, draft-ietf-lisp-alt-06, March 2011.
[9] D. Farinacci, et al., LISP Map Server, IETF Internet-Draft,
draft-ietf-lisp-ms-08, April 2011
[10] M. David, et al., PIM Group-to-Rendezvous-Point
Mapping, IETF RFC 4601, May 2011.
[11] Y.H.Choi, et al., “Using Correspondent Information for
Route Optimization Scheme on Proxy Mobile IPv6,”
Journal of Networks, Vol. 5, NO. 8, pp. 984-989, August
2010.