One document matched: draft-liang-idr-aira-00.txt









    Inter-Domain Routing Working Group                       Weifang. Liang 
    Internet Draft                                                     NDSC 
    Intended status: Informational                               Yue. Zhang 
    Expires: May 2011                                  Zhengzhou University 
                                                            Dongnian. Cheng 
                                                                       NDSC 
                                                           November 9, 2010 
                                       
     
                                          
            An AS-based Inter-domain Routing Architecture for a Scalable 
                                  Internet (AIRA) 
                            draft-liang-idr-aira-00.txt 

    Abstract 

       This document describes a simple, incremental, autonomous system 
       (AS)-based inter-domain routing architecture to solve the scalability 
       problems of the Internet, by separation of Endpoint Identifiers (EIDs) 
       for identifying endpoints and Routing Identifiers (RIDs) for locating 
       endpoints in the Internet. The proposed architecture uses AS numbers 
       as RIDs so that global routing tables can grow only with of AS 
       numbers, thus avoiding the rapid growth driven by constantly 
       increasing IP prefixes. 

       This proposal was stimulated by the problem statement effort at the 
       Amsterdam IAB Routing and Addressing Workshop in October 2006. 

    Status of this Memo 

       This Internet-Draft is submitted to IETF in full conformance with the 
       provisions of BCP 78 and BCP 79. 

       Internet-Drafts are working documents of the Internet Engineering 
       Task Force (IETF), its areas, and its working groups.  Note that 
       other groups may also distribute working documents as Internet-Drafts. 

       Internet-Drafts are draft documents valid for a maximum of six months 
       and may be updated, replaced, or obsoleted by other documents at any 
       time.  It is inappropriate to use Internet-Drafts as reference 
       material or to cite them other than as "work in progress". 

       The list of current Internet-Drafts can be accessed at 
       http://www.ietf.org/ietf/1id-abstracts.txt 

       The list of Internet-Draft Shadow Directories can be accessed at 
       http://www.ietf.org/shadow.html 

       This Internet-Draft will expire on May, 2011. 

     
     
     
    Liang                     Expires May,2011                    [Page 1] 
     
    Internet-Draft                  AIRA                      November 2010 
     
    Copyright Notice 

       Copyright (c) 2010 IETF Trust and the persons identified as the 
       document authors.  All rights reserved.  

       This document is subject to BCP 78 and the IETF Trust's Legal 
       Provisions Relating to IETF Documents 
       (http://trustee.ietf.org/license-info) in effect on the date of 
       publication of this document.  Please review these documents 
       carefully, as they describe your rights and restrictions with respect 
       to this document.  Code Components extracted from this document must 
       include Simplified BSD License text as described in Section 4.e of 
       the Trust Legal Provisions and are provided without warranty as 
       described in the Simplified BSD License. 

    Table of Contents 

       1. Introduction.................................................2 
       2. Conventions used in this document............................5 
       3. Terminology..................................................5 
       4. Network Topology.............................................8 
       5. Routing.....................................................10 
          5.1. Intra-AS Routing.......................................10 
          5.2. Inter-AS Routing.......................................11 
          5.3. Basic Description......................................11 
       6. Format Considerations.......................................14 
          6.1. Intra-AS Packet Format.................................14 
          6.2. Inter-AS Packet Format.................................15 
       7. Evaluations of the Benefits.................................15 
          7.1. Improved Scalability...................................15 
          7.2. Improved Stability.....................................16 
          7.3. Incremental Deployment.................................17 
       8. Analysis of the Routing Tables Growth.......................18 
          8.1. AS Growth..............................................18 
          8.2. Multi-homing and Traffic Engineering...................20 
          8.3. Fragmented Address.....................................21 
       9. Security Considerations.....................................22 
       10. IANA Considerations........................................22 
       11. Conclusions................................................22 
       12. References.................................................22 
          12.1. Normative References..................................22 
          12.2. Informative References................................24 
       13. Acknowledgments............................................24 
         
    1. Introduction 

       The current Internet is a decentralized collection of ASes from all 
       around the world. Every one of these ASes usually uses one or more 
     
     
    Liang                     Expires May,2011                    [Page 2] 
        
    Internet-Draft                  AIRA                      November 2010 
     
       Interior Gateway Protocols (IGPs) such as Routing Information 
       Protocol (RIP), Open Shortest Path First (OSPF) or Intermediate 
       System (IS-IS) for exchange of routing information within an AS. On 
       the other hand, these ASes use inter-domain routing protocol to 
       exchange reachability information to establish routes among them and 
       to allow the transmission of packets accross different ASes. The 
       Border Gateway Protocol (BGP) is currently the de-facto standard of 
       inter-domain routing protocols in the Internet. 

       The initial design of BGP improves scalability by IP prefixes 
       aggregation. But with the development of the Internet, there are a 
       large number of unaggregated IP prefixes, resulting in the rapid 
       growth of global routing table at an alarming rate over recent years. 
       The rapid growth of global routing tables imposes a considerable 
       pressure on the scalability of the Internet. Studies show that the 
       main factors driving such a growth include multi-homing and traffic 
       engineering that advertised abundant unaggregated prefixes. In 
       addition, fragmented addresses are also account for the routing table 
       growth. 

       BGP advertises destination prefixes to create routes in the current 
       Internet. The routers use Routing Information Base (RIB) and 
       Forwarding Information Base (FIB) to store route information: The 
       former holds detailed information of routes to destination prefixes, 
       while the later performs packet forwarding. That is, for a given 
       destination prefix, the RIB stores all routes related to the 
       destination prefix, and the FIB maps the prefix to a local exit 
       address.  Because it stores routes of all reachable prefixes to local 
       exit addresses in the Internet, the FIB will grows with the growth of 
       IP prefixes in the current Internet. The number of IP prefixes has 
       increased at an amazing rate over recent years. As a result, the FIB?
       size has accordingly increased over recent years, and will exceed the 
       storage capability of routers [1]. 

       In order to solve the scalability problems caused by multi-homing and 
       traffic engineering, several separation and mapping architectures [2-
       8] have been developed, which separate the Endpoint Identifiers (EIDs, 
       which are used for identifying the endpoint) from the Routing 
       Identifiers (RIDs, which are used for locating the endpoint in the 
       Internet, also called RLOCs in some cases) and then place both into 
       two different name spaces respectively, and use a mapping system to 
       distribute EID-to-GRID mapping information through the Internet.  
       These separation and mapping architectures fall into two categories 
       [9]: Core-Edge Separation (CES) and Core-Edge Elimination (CEE). The 
       CES separates the transit ASes and stub ASes into two different 
       address spaces. The routing table maintains the reachability 
       information related to IP prefixes of transit ASes, removing stub AS 
       IP prefixes from the inter-domain routing architecture for reducing 
     
     
    Liang                     Expires May,2011                    [Page 3] 
        
    Internet-Draft                  AIRA                      November 2010 
     
       the global routing table size. The CEE assigns multiple addresses to 
       a multi-homing stub AS for aggressively aggregating addresses. A 
       multi-homing stub AS receives one Provider Aggregated (PA) address 
       from per provider, and each of its provider will aggregate the 
       address with own PA address for reducing the unaggregated addresses. 

       Both of CES and CEE can alleviate the scalability pressure the 
       current inter-domain routing architecture is facing. However, they 
       essentially do not change the routing mechanisms and protocols of the 
       current Internet. The CES shrinks the address range to reduce the 
       size of a routing table, while the CEE aggregates the address space 
       to eliminate the stub AS addresses from the global routing area. If 
       using aggregated IP addresses to locate transit ASes, the global 
       routing table will still grow with prefixes due to multi-homing, 
       traffic engineering, and fragment addresses of transit ASes. 

       According to the data from the [10] from Mar. 2004 to Mar. 2008 and 
       the AS relationship from [11], the percent of multi-homing edge 
       transit ASes is about 71%-76%, while the percent of multi-homing stub 
       ASes is about 54%-57%. In addition, the average number of providers 
       of a multi-homing transit AS is 3, while it is 2 for a stub AS. 
       Therefore, if the RIDs use the IP addresses by aggregating prefixes 
       to scale the global routing system, the inter-domain routing 
       architecture still faces salability problems caused by multi-homing, 
       traffic engineering, and fragmented addresses of transit ASes. 

       The authors believe that IP prefixes are too fine routing 
       granularities to scale the routing system. Addressing the scalability 
       problems caused by using aggregated IP addresses in separation and 
       mapping architecture, this document proposes an AS-Based Inter-
       routing architecture (AIRA)which uses AS numbers (domain identifiers 
       for ASes) as GRIDs. The proposed architecture focuses on the 
       following issues: (1) By adopting 4 bytes AS numbers as GRIDs without 
       introducing a new name space, our proposal can be not only compatible 
       with the current IPv4 addresses but also available for future growth 
       of the Internet [12], so that routers can forward packets without 
       updating. (2) By using GMS(s) to support multi-homing and traffic 
       engineering for stub ASes, our proposal makes the multi-homing and 
       traffic engineering of stub ASes has no impact on the sizes of global 
       routing tables, and which makes the reachability of GRIDs completely 
       maintained in global routing tables. As a result, the global routing 
       table of the proposed architecture grows only with AS numbers, other 
       than IP prefixes. On the other hand, the proposed architecture makes 
       it easy to perform traffic engineering and to support routing 
       policies for all ASes and hosts. 

       Related work on AS-based solutions is given in Geographically 
       Informed Inter-domain Routing (GIRO) [13], Metro-base addressing [14], 
     
     
    Liang                     Expires May,2011                    [Page 4] 
        
    Internet-Draft                  AIRA                      November 2010 
     
       Geo-based addressing [15], Hybrid Link-state Path-vector (HLP) [16], 
       and Hierarchical Routing Architecture (HRA) [17].  Recently, some 
       schemes have been proposed to find how to design and assign GRID [18], 
       which is believed as a key of inter-domain routing based on ASes. 
       However, this document attempts to not compete or overlap with such 
       solutions and approaches. 

       After the introduction and related work given, the document is 
       proceeded as follows: Section 3 defines terms used in this document. 
       Section 4 describes network topology relevant to AIRA. Section 5 
       describes the routing mechanisms in the AIRA. Section 6 portrays the 
       format of packet being used in the AIRA. Section 7 evaluates the 
       benefits, and Section 8 analyzes the routing table growth, relevant 
       to AIRA in future individually. 

    2. Conventions used in this document 

       The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 
       "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 
       document are to be interpreted as described in RFC-2119 [RFC2119].  

    3. Terminology 

       This section defines terms relevant to AIRA and to be used in this 
       document. All of them is presented here for completeness, while some 
       of them are borrowed from other solutions [2-8]. 

       AIRA Stub Autonomous System (AIRA-Stub-AS): a network, similar to a 
       customer network in the current Internet, that is a source or 
       destination of packets. 

       AIRA Transit Autonomous System (AIRA-Transit-AS): an AS whose 
       business is to provide forwarding services for AIRA-Stub-ASes. All 
       the AIRA-Transit-ASes form the core network similar to Default Free 
       Zone in the current Internet. 

       Endpoint ID (EID): a 32-bit (for IPv4) or 128-bit (for IPv6) value 
       used in the source and destination address fields of an IP packet. A 
       host obtains a destination EID the same way it obtains a destination 
       address today, for example through a DNS lookup or SIP exchange.  The 
       source EID is obtained via existing mechanisms used to set a host's 
       "local" IP address.  An EID is allocated to a host from an EID-prefix 
       block associated with the AIRA-Stub-AS where the host is located.  An 
       EID can be used by a host to refer to other hosts (i.e. it is global 
       unique). EIDs MUST NOT be used neither as LRIDs nor as GRIDs. Note 
       that EID blocks may be assigned in a hierarchical manner, independent 
       of the network topology, to facilitate scaling of the Local Mapping 
       Server (LMS) database. In addition, an EID block assigned to an AIRA-
     
     
    Liang                     Expires May,2011                    [Page 5] 
        
    Internet-Draft                  AIRA                      November 2010 
     
       Stub-AS may have an AIRA-Stub-AS local structure (subnetting), and 
       this structure is not visible to the global routing system. 

       End-system: an IPv4 or IPv6 device that originates packets and 
       supplies an EID value for the destination address and the source 
       address field of the IP packets with IPv4 or IPv6 address when 
       communicating globally (i.e. outside of its AIRA-Stub-AS as usually). 
       An end-system can be a host computer, a switch or router device, or 
       any network appliance. 

       Local Routing Identifier (LRID): a 32-bit (for IPv4) or 128-bit (for 
       IPv6) value which is assigned to an interior router and is ONLY used 
       in the "LRID of Destination" fields in the intra-AS packet format 
       (see figure 2) for routing within the AS. A LRID is the output of a 
       EID-to-LRID mapping lookup. An EID MAY be mapped to one or more LRIDs. 
       Note that LRID MUST be locally unique within an AS, it MUST be 
       invisible out of the AS and all of the LRID(s) that locates within 
       the same AIRA-AS. 

       Global Routing Identifier (GRID): a 32-bit (4-bytes) value which is 
       assigned to an AS (no matter what the AS is an AIRA-Stub-AS or an 
       AIRA-Transit-AS). A GRID CAN appear either in the "LRID of 
       Destination" field in an intra-AS packet (see figure 2) or in an "AS 
       number of Destination" field in the inter-AS packet (see figure 3). 
       However, the GRID ONLY be used for inter-AS routing in the core 
       network. It is the output of an EID-to-GRID mapping lookup. An EID 
       MAY be mapped to one or more GRIDs. Because AIRA uses AS numbers as 
       GRIDs and an inter-domain routing table ONLY contains AS numbers, the 
       global routing table thus ONLY grows with AS numbers, not IP prefixes. 
       On the other hand, the AIRA adopts 4-byte AS numbers as GRIDs without 
       defining a new name space, which is not only compatible with current 
       IPv4 addresses but also available for the future growth of the 
       Internet.  Therefore, routers can forward packets without updating. 
       Note that GRID MUST be globally unique in the AIRA, that is, all of 
       the GRID(s) form the Global Forwarding Sysytem. 

       AIRA-Transit-AS Border Router (AIRA-Transit-ASBR): an border router, 
       connected to ASBRs, which has the capability of rewriting addresses 
       and exchange inter-AS reachability information by BGP. Every AIRA-
       Transit-AS MUST deploy one or more AIRA-Transit-ASBRs in order to 
       connect with other ASes. An AIRA-Transit-ASBR at least has a LRID of 
       the AS that it accesses. The LRID is used to represent a location 
       within AS. Two AIRA-Transit-ASes can directly connect, and an AIRA-
       Stub-AS can connect one or more upstreaming AIRA-Transit-ASes for 
       multi-homing and traffic engineering purposes. 

       EID-to-LRID Cache: a short-lived, on-demand table in an access router 
       (AR) of an AS that stores, tracks, and is responsible for timing-out 
     
     
    Liang                     Expires May,2011                    [Page 6] 
        
    Internet-Draft                  AIRA                      November 2010 
     
       and otherwise validating EID-to-LRID mappings. This cache is distinct 
       from the full "EID-to-LRID Database" in a Local Mapping Server (LMS) 
       of the same AS. It is dynamic, local to the AR(s), and relatively 
       small while the database is distributed, relatively static, and much 
       more global in scope of the AS. 

       EID-to-LRID Database: a globally distributed database that contains 
       all known EID-to-LRID mappings. Each potential AR typically contains 
       a small portion of the database: the EID-to-LRID mappings for the EID 
       "behind" the router. These map an EID to a LRID of the AR's own and 
       locally-visible IP addresses. The same database mapping entries MUST 
       be configured on all ARs for a given AS. 

       EID-to-GRID Cache: a short-lived, on-demand table in an AIRA-Transit-
       ASBR of an AIRA-Transit-AS that stores, tracks, and is responsible 
       for timing-out and otherwise validating EID-to-GRID mappings. This 
       cache is distinct from the full "EID-to-GRID Database" in a Local 
       Mapping Server (LMS) of the same AS. It is dynamic, local to the 
       AIRA-Transit-ASBR(s), and relatively small while the database is 
       distributed, relatively static, and much more global in scope. 

       EID-to-LRID Database: a global distributed database that contains all 
       known EID-to-GRID mappings. Each potential AIRA-Transit-ASBR 
       typically contains a small portion of the database: the EID-to-GRID 
       mappings for the EID "behind" the AIRA-Transit-ASBR. These map an EID 
       to the GRID, the AIRA-Transit-ASBR's own and globally-visible AS 
       Number. The same database mapping entries MUST be configured on all 
       AIRA-Transit-ASBRs for a given AIRA-Transit-AS. 

       Local Mapping Server (LMS): an entity which locates within either one 
       AIRA-Stub-AS or one AIRA-Transit-AS, and which stores the EID-to-LRID 
       mapping information to resolve the LRID for Intra-AS routing. Note 
       that, an AIRA-Transit-AS could have one or more LMS(s); all of the 
       AIRA-Transit-AS' LMS(s) form the Local Routing System, together with 
       all of the EID-to-LRID cache(s) that locates within the different 
       AIRA-Transit-AS' ASBR(s), and together with all of its interior 
       router(s). In addition, the Local Routing System ONLY locates within 
       the AIRA-Transit-AS (i.e. it is invisible to the outside of the AIRA-
       Transit-AS).  Similar to the AIRA-Transit-AS' LMS(s), AIRA-Stub-AS' 
       LMS(s) also constitutes its Local Routing System, however, together 
       with all of the EID-to-LRID cache(s) that locates within the 
       different AIRA-Stub-AS' AR(s). 

       Global Mapping Server (GMS): an entity which ONLY locates within the 
       AIRA-Transit-AS(s), and which stores the EID-to-GRID mapping 
       information to resolve the GRID for Inter-AS routing. Several GMSs 
       have been proposed such as LISP-DHT [19], DHT-Map [20], and they are 
       also suitable for AIRA. Note that, an AIRA-Transit-AS could have one 
     
     
    Liang                     Expires May,2011                    [Page 7] 
        
    Internet-Draft                  AIRA                      November 2010 
     
       or more GMS(s). All of the AIRA-Transit-AS' GMSs form the Global 
       Routing System, together with all of the EID-to-GRID cache(s) that 
       locate in the different AIRA-Transit-AS' ASBR(s); at the same time, 
       the Global Routing System ONLY locates within the core network. 

    4. Network Topology 

       This section describes the network topology related to AIRA. 

       According to Section 3.1, the AIRA consists of provider networks and 
       customer networks as the situations in today's Internet. That is, the 
       combination of the provider networks and customer networks together 
       refers to the main part of the Internet. Such a network topology is 
       considered here: AIRA-Transit-ASes form the core network similar to 
       the Default Free Zone in the current Internet, which provides data 
       forwarding services for AIRA-Stub-ASes; while AIRA-Stub-ASes 
       constitute customer networks similar to the customer networks in the 
       current Internet, which MAY connect themselves to one or more 
       provider networks, and which ONLY be traffic sources or sinks of data 
       traffic (Figure 1 shows the network topology). 

       End system (hosts) are located in customer networks and are 
       identified by EIDs. An EID is assigned to an end host and does not 
       change during the lifetime of the end host. In addition, both the 
       source and the destination of a packet are represented by the EID of 
       the sending end host and the receiving end host, respectively. On the 
       one hand, when an end host which locates in one AIRA-Stub-AS sends an 
       original packet, the packet CAN be routed to its Customer Edge Router 
       (CER) by using mechanisms such as IGP (See details later in Section 
       4.1). On the other hand, the AIRA-Transit-AS which receives the 
       original packet, and to which the AIRA-Stub-AS immediately connects 
       through one of its ASBR(s), MAY use one of the ASBR's LRID(s) to 
       rewrite the EID of this packet (See details later in Figure 2) within 
       its local range (if the case is need). When the destination of the 
       rewroten packet does not belong to itself,the AIRA-Transit-AS MUST 
       rewrite the rewritten packet again (See details later in Figure 3) in 
       a hop-by-hop manner until the rewritten packet ultimately reaches the 
       destination of the original packet. Finally, the other AIRA-Stub-AS, 
       to which the end host's correspondent host belongs, SHOULD also route 
       any original packet to the end host's correspondent host by using 
       mechanisms such as IGP. 






     
     
    Liang                     Expires May,2011                    [Page 8] 
        
    Internet-Draft                  AIRA                      November 2010 
     
       +--------------------------------------------------------------+ 
       |                        ***************                       | 
       |                       ** ** ** ** ** **                      | 
       |                      * AIRA-Transit-AS1*                     | 
       |                       ** ** ** ** ** **                      | 
       |                        ***************                       | 
       |                           /       \                          | 
       |                    +------+       +------+                   | 
       |                    |BSAR11|       |BSAR12|                   | 
       |                    +------+       +------+                   | 
       |                        |             |                       | 
       |                        |             |                       | 
       |                    +------+       +------+                   | 
       |                    |BSAR21|       |BSAR31|                   | 
       |                    +------+       +------+                   | 
       |                    /                     \                   | 
       |             ***************      ***************             | 
       |            ** ** ** ** ** **    ** ** ** ** ** **            | 
       |           * AIRA-Transit-AS2*  * AIRA-Transit-AS3*           | 
       |            ** ** ** ** ** **    ** ** ** ** ** **            | 
       |             ***************      ***************             | 
       |                    /                    /   \                | 
       |             +------+             +------+   +------+         | 
       |             |BSAR22|             |BSAR32|   |BSAR33|         | 
       |             +------+             +------+   +------+         | 
       |              |    |                 |           |            | 
       |              |    |                 |           |            | 
       |     +---------+  +---------+   +---------+ +---------+       | 
       |     |AS1-CER11|  |AS1-CER21|   |AS1-CER22| |AS1-CER31|       | 
       |     +---------+  +---------+   +---------+ +---------+       | 
       |         /                  \    /               \            | 
       |  *************         *************          *************  | 
       | *AIRA-Stub-AS1*       *AIRA-Stub-AS2*        *AIRA-Stub-AS3* | 
       |  *************         *************          *************  | 
       |      /                      /   \                  \         | 
       |    +--------+      +--------+   +--------+     +--------+    | 
       |    |AS1-AR11|      |AS2-AR21|   |AS2-AR22|     |AS3-AR31|    | 
       |    +--------+      +--------+   +--------+     +--------+    | 
       |        |               |             |             |         | 
       |        |               |             |             |         | 
       |     +-----+         +-----+       +-----+       +-----+      | 
       |     |Host1|         |Host2|       |Host3|       |Host4|      | 
       |     +-----+         +-----+       +-----+       +-----+      | 
       +--------------------------------------------------------------+ 
        
                       Figure 1 The AIRA's Network Topology 


     
     
    Liang                     Expires May,2011                    [Page 9] 
        
    Internet-Draft                  AIRA                      November 2010 
     
    5. Routing 

       This section describes the routing in the AIRA, which includes intra-
       AS routing, inter-AS routing, and basic description. 

    5.1. Intra-AS Routing 

       Each AS in AIRA is an independent AS, and uses own interior routing 
       protocols. Each AS can independently select an interior routing 
       protocol. 

       The reachability information that an AS learns from the exterior 
       needs to be distributed within an AS so that the outside of the AS is 
       reachable from every router inside in the current Internet. As a 
       result, the FIB of DFZ routers maps all IP prefixes to local exit 
       addresses.  With the growth of prefixes, the all DFZ routers face the 
       scalability pressure. 

       The FIB of AIRA maps the all AS numbers in core network to local exit 
       addresses. The inter-AS routing table maintains the reachability 
       information of all the ASes based on AS numbers, and the intra-AS 
       routing table only stores local routing information. In order to 
       forward a packet to a local exit address, the router needs map the AS 
       number to a local exit address. AIRA defines a local neighbor table 
       to store AS numbers and the exit addresses of neighboring ASes. The 
       form of local neighboring table is (AS number, LRID, Weight), where 
       Weight is for the traffic engineering purpose. For example, if two 
       ASes directly connect each other by a link, Weight=1. If two ASes 
       directly link by i links, the sum of all i Weights is 1 and 
       1<=Weight<=1. 

       Endpoints are donated with EIDs in AIRA. To reach an endpoint in the 
       Internet, the routers need RIDs. The LMS provides the LRIDs for a 
       source endpoint host. The LRIDs of the source and destination hosts 
       are filled into packets addressed towards destination EID so that it 
       can be carried through the stub ASes and forwarded to the destination. 
       The packet format is showed in Figure 2. The Source EID and 
       Destination EID are the endpoint identifiers of the source and 
       destination hosts respectively, and the LRID of Destination and AS 
       number of Destination are the LRID and 4-byte AS numbers of the 
       destination respectively. 

       The interior routers forward packets according to intra-domain 
       routing information until the packets reach their destinations. 




     
     
    Liang                     Expires May,2011                   [Page 10] 
        
    Internet-Draft                  AIRA                      November 2010 
     
    5.2. Inter-AS Routing 

       The BGP advertise 4-byte AS numbers in AIRA. The FIB of ASBR maps all 
       AS numbers to local exit addresses through three tables. The ASBR 
       firstly maintains the Inter-AS table to store the reachbility 
       information of AS numbers for mapping the AS number to next AS number. 
       Secondly, the ASBR stores the local neighbor table for mapping the AS 
       numbers to local exit addresses. Thirdly, the ASBR stores the intra-
       routing table for forwarding packets to the exit address. 

       When an ASBR located in a transit AS receives a packet without local 
       AS number, it looks up its inter-AS routing table to decide the next 
       hop AS, and then looks up the local neighbor table to decide the 
       local exit address. According to the exit address, the ASBR forwards 
       packet according to ths intra-routing routing table. The format of a 
       packet is showed in figure 3.  The Source EID and Destination EID are 
       the endpoint identifiers of the source and destination hosts 
       respectively, and the LRID of ASBR and AS number of Destination are 
       the local exit addresses and 4-byte AS number of the destination 
       respectively. 

       In fact, when a packet reaches a transit AS, the ASBR needs to kook 
       up the inter-AS routing table for obtaining the next AS number, and 
       looks up the neighbor table for gaining the local exit address. 

    5.3. Basic Description 

       This portion provides an example of the unicast packet flow with the 
       following conditions (refers to Figure 1 above) 

       o Source host "host1" wants to communicate with the destination host 
          "host2", note that, "host1" is directly attached to the AIRA-Stub-
          AS1' "AS1-AR11", while "host2" is immediately linked to the AIRA-
          Stub-AS3' "AS3-AR32" 

       o Each AIRA-AS may be multi-homed, such as AIRA-Stub-AS2 is 
          connected both to AIRA-Transit-AS2 through the "AS2-CER21" and the 
          "BSAR22" and to AIRA-Transit-AS3 through the "AS2-CER22" and the 
          "BSAR32" 

       o Each BSAR may rewrite the packet that it receives from the other 
          AIRA-AS 

       o The details of looking up LMS and GMS are ignored, however, this 
          document assumes that every look-up is successful. 

        

     
     
    Liang                     Expires May,2011                   [Page 11] 
        
    Internet-Draft                  AIRA                      November 2010 
     
       When the Source host "host1" wants to communicate with the 
       destination host "host2", the packet sequence between them like the 
       followings: 

       1. The intra-domain routing within the AIRA-Stub-AS1: 

       (1.1)The "host1" sends the original packet, that tries to open a TCP 
       connection to "host2," and then reaches "AS1-AR11". Note that, the 
       original packet is a normal IP packet's (i.e. its "Source Field" is 
       the EID of the "host1", and its "Destination Field" is the EID of the 
       "host2"); 

       (1.2)AS1-AR11 makes a LMS (which locates in the AIRA-Stub-AS1) look-
       up on "host2". Because of the "host2" dose not belong to the AIRA-
       Stub-AS1, the look-up fails; 

       (1.3)Then, the AS1-AR11 looks up local neighbor table to decide the 
       local exit address.  The "LRIDs of Destination Field" (which is "AS1-
       CER11") are added to the packet addressed towards destination EID so 
       that it can be carried through the stub ASes and forwarded to the 
       destination and the "AS Number of Destination Field" keeps null 
       respectively (the packet format is shown in Figure 2); 

       2. The inter-domain routing between the AIRA-Stub-AS1 and AIRA-
          Transit-AS2: "AS1-AR11" directly forwards the normal IP packet to 
          "BSAR22". 

       3. The intra-domain routing within the AIRA-Transit-AS2: 

       (3.1) When "BSAR22" receives a packet without the local AS number, it 
       looks up the inter-AS routing table (GMS) to decide the next hop AS, 
       which is only "AIRA-Transit-AS3' GRID"; 

       (3.2) Then, "BSAR22" rewrites the "AS Number of Destination Field" 
       with the "AIRA-Transit-AS3' GRID"; 

       (3.3) Finally, "BSAR22" looks up its local neighbor table to decide 
       the local exit address. The "LRIDs of Destination Field" (which is 
       "BSAR21") are added to the packet addressed towards destination LRID 
       so that it can be carried through the "AIRA-Transit-AS2" and 
       forwarded to the destination. The format of packet is given in figure 
       3; 

     

       4. The inter-domain routing between the AIRA-Transit-AS2 and AIRA-
          Transit-AS1: "BSAR21" immediately forwards the rewritten packet to 
          "BSAR11". 
     
     
    Liang                     Expires May,2011                   [Page 12] 
        
    Internet-Draft                  AIRA                      November 2010 
     
       5. The intra-domain routing within the AIRA-Transit-AS1: 

       (5.1) When "BSAR11" receives the rewritten packet from "BSAR21", it 
       looks up the inter-AS routing table to decide the next hop AS, which 
       is only "AIRA-Transit-AS3' GRID."; 

       (5.2) Then, "BSAR11" looks up the local neighbor table to decide the 
       local exit address.  The "LRIDs of Destination Field" (which is 
       "BSAR12") are rewritten to the packet addressed towards destination 
       LRID so that it can be carried through the "AIRA-Transit-AS1" and 
       forwarded to the destination; 

       6. The inter-domain routing between the AIRA-Transit-AS1 and AIRA-
          Transit-AS3: "BSAR12" first forwards the rewritten packet to 
          "BSAR31". 

       7. The intra-domain routing within the AIRA-Transit-AS3: 

       (7.1) When "BSAR31" receives the rewritten packet from "BSAR12", it 
       finds that the "AS Number of Destination Field" in the packet is 
       itself; 

       (7.2) Then, it looks up the local neighbor table to decide the local 
       exit address. The "LRIDs of Destination Field" (which is "BSAR33") 
       are rewritten again to the packet addressed towards destination LRID 
       so that it can be carried through the "AIRA-Transit-AS3" and then is 
       forwarded to the destination; 

       8. The inter-domain routing between the AIRA-Transit-AS3 and AIRA-
          Stub-AS3: "BSAR33" directly forwards the rewritten packet to "AS3-
          CER31". 

       9. The intra-domain routing within the AIRA-Stub-AS3: 

       (9.1) When "AS3-CER31" receives the rewritten packet from "BSAR33", 
       it finds that the "AS Number of Destination Field" in the packet is 
       itself; 

       (9.2) Then, it looks up local neighbor table to decide the local exit 
       address. The "LRIDs of Destination Field" (which is "AS3-AR32") are 
       rewritten again to the packet addressed towards destination LRID so 
       that it can be carried through the "AIRA-Stub-AS3" and forwarded to 
       the destination; 

       (9.3) And then, when "AS3-AR32" receives the rewritten packet, it 
       finds that the "Destination EID" (which is "Host2") locates just 
       within itself, so that it immediately forwards the rewritten packet 
       to "Host2". The packet deliver is fulfilled. 
     
     
    Liang                     Expires May,2011                   [Page 13] 
        
    Internet-Draft                  AIRA                      November 2010 
     
    6. Format Considerations 

       This section portrays formats of packets being used in the AIRA, 
       which covers both the intra-AS packet format and the inter-AS packet 
       format. 

    6.1. Intra-AS Packet Format 

       0                   1                   2                   3 
       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 
       |Version|  IHL  |Type of Service|          Total Length       | 
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 
       |         Identification        |Flags|       Fragment        | 
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 
       |  Time to Live |    Protocol   |       Header Checksum       | 
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 
       |                         Source EID                          | 
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 
       |                       LRID of Domain                        | 
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 
       |                   Option                      |   Padding   | 
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 
       |                       Destination EID                       | 
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 
       |                   AS Number of Destination                  | 
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 
       |                           Data                              | 
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 
        
                         Figure 2 Intra-AS Packet Formart 

        














     
     
    Liang                     Expires May,2011                   [Page 14] 
        
    Internet-Draft                  AIRA                      November 2010 
     
    6.2. Inter-AS Packet Format 

       0                   1                   2                   3 
       0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 
       |Version|  IHL  |Type of Service|          Total Length       | 
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 
       |         Identification        |Flags|       Fragment        | 
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 
       |  Time to Live |    Protocol   |       Header Checksum       | 
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 
       |                         Source EID                          | 
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 
       |                       LRID of Egress ASBR                   | 
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 
       |                   Option                      |   Padding   | 
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 
       |                       Destination EID                       | 
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 
       |                   AS Number of Destination                  | 
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 
       |                           Data                              | 
       +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 
        
                         Figure 3 Inter-AS Packet Formart 

    7. Evaluations of the Benefits 

       This Section evaluates the benefits relevant to AIRA, which involves 
       in improved scalability, stability and incremental deployment. AIRA 
       separates RIDs from EIDs and in turn customer networks from provider 
       ones. As a result, AIRA shares the all benefits of separation and 
       mapping architectures such as supporting site multi-homing and 
       traffic engineering, and supporting host multi-homing and mobility. 
       This section mainly analyzes the other benefits of routing based on 
       AS number. 

    7.1. Improved Scalability 

       AIRA has three kinds of routing protocols: the intra-AS routing 
       protocol for routing within stub AS, the intra-AS routing protocol 
       for routing within transit AS, and the inter-AS routing protocol for 
       inter-AS routing. 

       The routing table maintains the local reachability information 
       related to accessing endpoints within stub ASes. The number of 
       customer networks is about 3*10^4 in Mar. 2010 [10], and the number 
       of endpoints is about 10^9 [21] in the current Internet. So, the 
     
     
    Liang                     Expires May,2011                   [Page 15] 
        
    Internet-Draft                  AIRA                      November 2010 
     
       average number of endpoints per customer network is about 3.3*10^4. 
       The number of routing table items is about 3.3*10^4. For the current 
       router, it is desirable. It is obvious that AIRA improves the 
       scalability of interior routers within stub ASes. 

       The routing table maintains the local routing information within 
       transit AS. The transit AS only stores the reachability information 
       related to routers within a local AS. The number of routers is small 
       within most transit ASes. The current number of ISP is about 5*10^3 
       [10], and the number of routers is about 5*10^5 [22]. The average 
       number of routers per ISP is about 10^2. It is also obvious that AIRA 
       improves the scalability of interior routers within transit ASes. For 
       a large ISP, the intra-AS routing table can contain between 50,000 
       and 150,000 routes. If the router stores both intra-AS and inter-AS 
       routing tables, it will have significant impact on the scalability. 
       The routers in AIRA only store the intra-domain routing table, which 
       can alleviate scalability pressure on routers in large ISP. 

       The inter-AS routing table maintains the routing information related 
       to AS numbers. As a result the number of routing table items is the 
       number of AS numbers of all transit ASes. If each transit AS has one 
       AS number, the number of inter-AS routing items is the number of 
       transit ASes in the Internet. But the multi-homing transit ASes need 
       more AS numbers for the traffic engineering purpose. The average 
       number of ISPs is 3. This result shows: (1) the estimated number of 
       AS numbers in AIAR is comparable to the number of IP prefixes in the 
       current Internet. From Mar. 2006 to Mar. 2010, the number of IP 
       prefixes increased from 200,000 to 320,000. (2) the number of AS 
       number is two orders of magnitude fewer than the size of current IP 
       prefixes, and is likely to remain more or less constant over time. 

       In the current Internet, the number of routing table entries is 
       about3.2*10^5, while the number of transit ASes is about 3*10^4. Even 
       assuming that every transit AS advertises 3 AS numbers, the routing 
       table size of AIRA is about 9.3% of the average size of a current 
       routing table. 

    7.2. Improved Stability 

       The ASes are dependent entries in the current Internet, and use BGP 
       to advertise reachability information of ASes. However, the route 
       failure in one AS will be flooded in the whole Internet, which 
       results in instability of inter-domain routing. In addition, there 
       are many routers and links between routers inside an AS. The topology 
       changes caused by instable links consume more resource for processing 
       the route updates related to interior routes. The studies on Tire-1 
       ISP show that a router with rich links processes hundreds of routing 
       updates within a minute, the majority of which are caused by the 
     
     
    Liang                     Expires May,2011                   [Page 16] 
        
    Internet-Draft                  AIRA                      November 2010 
     
       interactivity between intra-routers and inter-routers [23]. Some 
       studies show BGP can churn because of wrong configurations inside an 
       AS [24], and can persistently loop because of topology changes inside 
       an AS [25]. The Internet routes today can be overwhelmed due to 
       frequent routing updates. 

       ASes is independent entities in AIRA, and BGP only advertises 
       reachability information of AS numbers without distributing exterior 
       reachability information within ASes. The routing dynamics occurring 
       inside ASes will have no impact on the inter-AS routing stability. 
       Furthermore, one AS is usually connected multiple other ASes by 
       multiple links for the fault-tolerance purpose. So, when one link 
       failures, AIRA can deal with the failure by local neighbor table, 
       instead of triggering routing updates. 

    7.3. Incremental Deployment 

       On the Internet, one can not simply set a flag day when all 
       participators will switch to a new design, no matter how great an 
       advantage the design offers. Those designs that want to be deployed 
       should follow an incremental pattern. And incremental deployment 
       scheme should be backward compatible for communicating with legacy 
       networks and has incentives so that different participators adopt the 
       scheme. 

       The EID and LRID both adopt IPv4 addresses which makes the AIRA is 
       backward compatible with the current intra-domain routing protocols. 
       AIRA uses 4-byte AS number for the backward compatibility with the 
       current inter-domain routing because current BGP can only advertise 
       IPv4 addresses. 

       The design of AIRA follows the backward compatibility, but AIRA needs 
       a mechanism for communicating with legacy networks. AIRA can provide 
       backward compatibility by ASBR.  The IP prefixes of legacy networks 
       are still advertised throughout the Internet. When a source endpoint 
       in the new network wants to communicate with a destination endpoint 
       in legacy network, the source endpoint fills the Source EID and LRID 
       of ASBR by using its own EID and the exit address respectively. The 
       Destination EID and AS number of Destination are filled using the 
       IPv4 address of the destination endpoint. The routers of the Internet 
       forwards packets without changing the destination addresses until the 
       packets reach the legacy network.  Then, the legacy network forwards 
       the packets to destination endpoint based on IPv4 address. 

       When a source endpoint in a legacy network wants to send packets to a 
       destination endpoint in a new network, the source endpoint fills the 
       Source and Destination address fields using its own IPv4 address and 
       the EID of destination endpoint respectively. After forwarding 
     
     
    Liang                     Expires May,2011                   [Page 17] 
        
    Internet-Draft                  AIRA                      November 2010 
     
       packets to a legacy network, the ASBR of the new network replaces the 
       Destination EID with AS number of the Destination. The routers in the 
       legacy network normally forward packets.  When the ASBRs of new 
       networks receive a packet from a legacy network, the ASBRs 
       differentiated processing packets according to the packet format. If 
       the packet is a normal IPv4 one, the ASBRes add the LRID of ASBR and 
       AS number of the destination endpoint. If the packet contains the AS 
       number of the Destination endpoint, the ASBRs replace the Destination 
       EID with the AS number of the Destination. After rewriting the 
       addresses, the ASBRs forward packets. 

       There are currently three kinds of participators: endpoint host, stub 
       AS and transit AS.  AIRA provides incentives for early different 
       adoptions. AIRA allows host to deploy multi-homing and traffic 
       engineering. AIRA separates the endpoint identifier from routing 
       identifier so that the host can both access multiple stub ASes 
       without changing own identifier and control traffic by the mapping 
       system. Furthermore, the separation of identity identifier from 
       routing identifier makes host mobility easier. 

       The stub AS in AIRA can use Provider-independent addresses, which 
       eliminates the renumbering problem caused by changing providers, and 
       thus multi-homing easier.  Furthermore, the stub AS can easily 
       control traffic by mapping system. 

       The routers inside a transit AS only maintain the local routing 
       information, which alleviate the scalability pressure due to 
       processing only the inter-AS routing information. The ASBRs maintain 
       the routing information related to AS numbers, which is far less than 
       the IP prefixes in the current Internet. All routers of a transit AS 
       store less routing information, and process less routing updates than 
       the routers in the current Internet.  

    8. Analysis of the Routing Tables Growth 

       This Section analysis the routing table growth of the AIRA, including 
       AS growth, multi-homing, traffic engineering, and fragmented 
       addresses. The rapid growth of inter-domain routing table size is 
       mainly resulted in the increasing number of ASes, multi-homing, 
       traffic engineering, and fragment addresses [18]. If the growth of AS 
       numbers exceeds the growth of IP prefixes, the AIRA will be not scale. 
       We will analysis the impact of the main factors that cause the rapid 
       growth of IP prefix on the routing table size of AIRA. 

    8.1. AS Growth 

       As more networks connected to the Internet, more AS numbers will be 
       assigned to identify these networks accordingly. If the number of AS 
     
     
    Liang                     Expires May,2011                   [Page 18] 
        
    Internet-Draft                  AIRA                      November 2010 
     
       number grows, the inter-AS routing table will grows accordingly. With 
       the development of the Internet the growth of AS number will be 
       inevitable. We use the data from Route Views Project [11] between 
       2007 and 2009 to evaluate the growth of AS number and IP prefix in 
       the future. Table 1 below shows the increasing rate of AS numbers and 
       IP prefixes in the recent three years respectively. 

       Table1: Increasing rate of IP prefixes and transit ASes 
       +----+------+---------+----------+ 
       |Year|Iterms|IP Prefix|Transit AS| 
       +----+------+---------+----------+ 
       |    | Jan. | 216,190 |  3,794   | 
       |    |------|---------|----------| 
       |2007| Dec. | 244,886 |  4,282   | 
       |    |------|---------|----------| 
       |    | Rate |  13.3%  |  12.9%   | 
       +----+------+---------+----------+ 
       |    | Jan. | 247,204 |  4,271   | 
       |    |------|--------------------+ 
       |2008| Dec. | 284,795 |  4,685   | 
       |    |------|--------------------+ 
       |    | Rate |  15.2%  |  9.70%   | 
       +----+------+---------+----------+ 
       |    | Jan. | 289,185 |  4,732   | 
       |    |------|--------------------+ 
       |2009| Dec. | 312,244 |  5,126   | 
       |    |------|---------|----------| 
       |    | Rate |   8.0%  |   8.3%   | 
       +----+------+---------+----------+ 
        
        

       The results show the number of transit ASes is two orders of 
       magnitude fewer than the number of IP prefixes in current Internet. 
       The number of transit ASes are 1.75%, 1.65% and 1.64% of the number 
       of IP prefixes respectively. Even every multi-homing transit AS 
       advertises 3 AS numbers, the average size of AIRA's routing table is 
       less than the current BGP routing table size by 6%. 

       Assuming the increasing rate of transit ASes is denoted by lambda, 
       the number of transit ASes in Dec. 2009 is denoted by k, and the 
       number of transit ASes in year N is denoted by S_T , we have: 

        S_T = k*(lambda+1)^(N-2009) 

       Assuming the increasing rate of IP prefixes is denoted by mu, the 
       number of IP prefixes in Dec. 2009 is denoted by s, and the number of 
       IP prefixes in year N denoted by S_IP , we have: 
     
     
    Liang                     Expires May,2011                   [Page 19] 
        
    Internet-Draft                  AIRA                      November 2010 
     
       S_IP = s*(mu+1)^(N-2009) 

       Let k be 5,126 and s 312,244 in Dec.2009, repectively. During three 
       years, the average value of lambda is 10.3% and the average value of 
       mu is 12.12%. 

       The results show that the number of transit ASes is two orders of 
       magnitude fewer than the number of IP prefixes in future years. The 
       number of transit AS is less than the number of IP prefix in current 
       Internet within the future 50 years. 

    8.2. Multi-homing and Traffic Engineering 

       Both transit and stub ASes can support multi-homing and traffic 
       engineering, which will result in the growth of inter-domain routing 
       table size in the current Internet. The multi-homing and traffic 
       engineering of stub ASes has no impact on the inter-AS routing table 
       size in AIRA. The stub ASes use mapping system for supporting multi-
       homing and traffic engineering, avoiding the growth of routing table 
       size due to multi-homing and traffic engineering of stub ASes. 

       The multi-homing transit AS does not need additional AS numbers for 
       multi-homing, but it needs additional AS numbers for traffic 
       engineering. The inter-AS routing table contains the AS numbers. If 
       there are multiple routes to a common destination AS, the routing 
       table needs multiple entries to present the different routes for 
       traffic engineering. 

       We analyze the multi-homing of transit AS and stub AS according to 
       the data from Route Views Project between 2006 and 2008. Table 2 
       shows the result related to multi-homing percent of transit ASes and 
       stub ASes, and the average number of upstreaming ISPs of transit ASes 
       and stub ASes. 














     
     
    Liang                     Expires May,2011                   [Page 20] 
        
    Internet-Draft                  AIRA                      November 2010 
     
       Table2: Percent and average number of ISPs 
       +----+-------------+-------------+----------+ 
       |Year|    Iterms   | Transit ASes| Stub ASes| 
       +----+-------------+-------------+----------+ 
       |    |   Percent   |   74.84     |  56.45   | 
       |2006|-------------+-------------|----------| 
       |    |Average ISPs |    3.07     |   2.26   | 
       +----+-------------+-------------+----------+ 
       |    |   Percent   |   75.14     |  55.14   | 
       |2007|-------------+-------------|----------| 
       |    |Average ISPs |    3.18     |   2.28   | 
       +----+-------------+-------------+----------+ 
       |    |   Percent   |   76.27     |  54.41   | 
       |2008|-------------+-------------|----------| 
       |    |Average ISPs |    3.30     |   2.32   | 
       +----+-------------+-------------+----------+ 
        
        

       The results show that about 75% of transit ASes deploy multi-homing, 
       while 55% of stub ASes deploy multi-homing. For multi-homing ASes, 
       the number of transit ASes is greater than that of stub ASes. 
       Furthermore, the average number of ISPs of a transit AS is 3, while 
       that of ISPs of a stub AS is 2, which shows a transit AS connects 
       more ISPs than a stub AS.  

       If a transit AS wants to use multiple distinct paths for traffic 
       engineering, it must advertise multiple AS numbers, which will cause 
       the routing table growth. Set the percent of multi-homing transit 
       ASes be alpha, the average number of upstreaming ISPs be beta, and 
       the number of AS numbers advertised be S-AS in year N. We have: S-AS 
       =alpha*beta*k*(lambda+1)^(N-2009)+(1-alpha)*k*(lambda+1)^(N-2009). 

       If the average number of AS numbers advertised by multi-homing 
       transit ASes equals the average number of upstreaming ISPs, the 
       routing table size is one order of magnitude fewer than the number of 
       IP prefixes in the current Internet. The number of IP prefixes in the 
       Internet is 312,224 in Dec. 2009. The number of AS numbers in AIRA is 
       12,815 accounting for traffic engineering, which is 4.1% of the IP 
       prefixes in the Internet. If AIRA is deployed, the routing table size 
       will not exceed the routing table size of the current Internet until 
       2043. 

    8.3. Fragmented Address 

       Current ASes can not aggregate multiple fragment addresses, causing 
       the growth of routing table. Many factors can cause fragment 
       addresses. If an organization needs additional address space to 
     
     
    Liang                     Expires May,2011                   [Page 21] 
        
    Internet-Draft                  AIRA                      November 2010 
     
       identify its endpoints, the new address space can not be aggregated 
       with the previous address space, thus resulting in fragment addresses. 
       When an organization merges with other organization, the two address 
       spaces of two organizations are commonly not continuous, also 
       resulting in fragment addresses. The fragment addresses are 
       advertised in the Internet. 

       However, the fragment addresses of an AS are aggregated into AS 
       number in AIRA. FIBs of inter-AS routers map the AS number to next 
       hops, instead of mapping IP prefixes to next hops.  As a result, the 
       routing table needs an entry in order to forward the packets to an AS 
       even if the AS has multiple fragment addresses. The intra-routing 
       table decides how to forward packets within AS. The fragment 
       addresses do not cause the growth of routing table. 

    9. Security Considerations 

       This document does not introduce any security considerations. 

    10. IANA Considerations 

       This document makes no request of IANA. 

       Note to RFC Editor: this section may be removed on publication as an 
       RFC. 

    11. Conclusions 

       A key to separation and mapping architecture is how to select routing 
       identifier. Aiming at the scalability problem caused by topology 
       aggregated IP prefixes, this paper proposes an AS-based routing 
       architecture for improving the scalability of separation and mapping 
       architecture. The proposed architecture adopts AS number as routing 
       identifier, and inter-AS routing only contains AS numbers. The 
       routing table grows with the growth of AS numbers advertised, and the 
       growth of IP prefixes has no impact on the routing table. The 
       benefits and scalability of proposed architecture show that the 
       architecture can effectively alleviate the scalability pressure the 
       current Internet is facing, and has good scalability in the future. 

    12. References 

    12.1. Normative References 

       [1]  Meyer, D., Zhang, L., and K. Fall, "Report from the IAB 
            Workshop on Routing and Addressing", RFC 4984, September 2007. 


     
     
    Liang                     Expires May,2011                   [Page 22] 
        
    Internet-Draft                  AIRA                      November 2010 
     
       [2]  D.Farinacci, V.Fuller, D.Meyer, D.Lewis. Locator/ID Separation 
            Protocol(LISP)[EB/OL],draft-ietf-lisp-07.txt, 2010. 

       [3]  D.Jen, M.Meisel, D Massey., L.Wang, B.Zhang, L.Zhang. APT: A 
            Practical Transit Mapping Service [EB/OL], draft-jen-apt-01.txt, 
            2007. 

       [4]  R.Whittle. Ivip(Internet Vastly Improved Plumbing) Architecture. 
            Internet Draft, draft-whittle-ivip-arch-04, Mar.2010. 

       [5]  Juan Jose Adan. Tunneled Inter-doamin Routing(TIDR). Internet 
            Draft, draft-adan-idr-tidr-01, Nov. 2006. 

       [6]  R.Atkinson. ILNP Concept of Operations. Internet Draft, Draft-
            rja-ilnp-intro-03.txt, Feb. 2010. 

       [7]  R.Moskowitz, P.Nikander. Host Identify Protocol(HIP) 
            Architecture. RFC 4423, May 2006. 

       [8]  X.Xu. Routing Architecture for the Next Generation Internet 
            (RANGI), Internet Draft, draft-xu-rangi-03, Feb. 2010. 

       [9]  D.Jen, M.meisel, H.Yan, D.massey, L.wang. Towards a Future 
            Internet Architecture: Arguments for Separating Edges from                          
            Transit Core. in 7  ACM Workshop on Hot Topics in Networks 
            (HotNets), Cal.gary, Alberta, Canada, Oct.2008. 

       [10] The RouteViews Project [EB/OL]. http://www.routeviews.org/. 

       [11] The CAIDA AS Relationships Dataset [EB/OL]. <2004.03-2008.03>. 
            http://www. caida.org/data/active/as-relationships/. 

       [12] Potaroo. The 32-bit AS number report, 2008. 
            http://www.potaroo.net/tools/asn32/index.html. 

       [13] R.Oliveira, M.Lad, B.Zhang, L.Zhang. Geographically Informed 
            Inter-Domain Routing[A]. In: IEEE International Conference on 
            Network Protocols 2007 (ICNP2007) [C], Beijing, China, 2007: 
            103-112. 

       [14] S.Deering. Metro-based Addressing: A Proposed Addressing Scheme 
            for the IPv6. Presentation, Xerox PARC, July, 1995. 

       [15] T.Hain. An IPv6 Provider-Independent Global Unicast Address 
            Format [EB/OL]. draft-hain-ipv6-pi-addr-10.txt, Aug.2006. 



     
     
    Liang                     Expires May,2011                   [Page 23] 
        
    Internet-Draft                  AIRA                      November 2010 
     
       [16] L.Subramanian, M.Caesar, C.T.Ee, M.handley, M.Mao, S.Shenker, 
            I.Stoica. HLP: A Next Generation Inter-domain Routing Protocol 
            [A]. In ACM SIGCOMM[C], Philadelphia, Pennsylvania, USA, 2005: 
            13-24. 

       [17] X.Xu, D.Guo. Hierarchical Routing Architecture (HRA) [EB/OL], 
            draft-xu-rrg-hra-00.txt, Feb.2008. 

       [18] Preliminary Recommendation for a Routing Architecture. draft-
            irtf-rrg-recommendation-02. 2009.03 

    12.2. Informative References 

       [19] L.Mathy, L.Iannone. LISP-DHT: Towards a DHT to map Identifiers 
            onto Locators, In ACM ReArch, Madrid, Spain, 2008. 

       [20] H.Luo, Y.Qin, H.Zhang. A DHT-based Identifier-to-Locator 
            Mapping Approach for a Scalable Internet. IEEE Transaction on 
            Parallel and Distribution Systems. Vol.20, No.10.2009. 

       [21] Internet System Consortium. The ISC Domain Survey. 
            http://isc.org/solutions/survey, 2009. 

       [22] Lumeta. Internet Map Project. 
            http://www.lumeta.com/internetmapping/. 

       [23] R.Teixeira, A.Shaikh,T.Griffin, J.Rexford. Dynamics of hot-
            potato routing in IP networks. In Proc. of ACM SIGMETRICS`04, 
            2004. 

       [24] T. G .Griffin, G. Wilfong. On the correctness of IBGP 
            Configuration. In Proc. ACM SIGCOMM,  2002. 

       [25] R.DuBE, A comparison of scaling techniques for BGP. ACM 
            Computer Communications Review 29 ,3(July 1999):44-46. 

    13. Acknowledgments 

       Funding for the RFC Editor function is currently provided by the 
       Internet Society. 

        






     
     
    Liang                     Expires May,2011                   [Page 24] 
        
    Internet-Draft                  AIRA                      November 2010 
     
    Authors' Addresses 

       Weifang Liang 
       National Digital Switching System Engineering and Technological R&D 
       Center of China 
       No. 5, Jianxue Street, Zhengzhou, HeNan 450002, P. R. China 
        
       Phone: +8637181632830 
       Email: lwf@ndsc.com.cn 
        

       Yue Zhang 
       Zhengzhou University, School of Information & Engineering 
       No.100, Science Avenue, ZhengZhou, HeNan 450052, P. R. China 
        
       Phone: +8637167767757 
       Email: ieyzhang@zzu.edu.cn 
        

       Dongnian Cheng 
       National Digital Switching System Engineering and Technological R&D 
       Center of China 
       No. 5, Jianxue Street, Zhengzhou, HeNan 450002, P. R. China 
        
       Phone: +8637181632743 
       Email: cdn@ndsc.com.cn 
        




















     
     
    Liang                     Expires May,2011                   [Page 25] 
        


PAFTECH AB 2003-20262026-04-24 09:08:08