Network working group X. Xu Internet Draft S. Hares Category: Informational Huawei Technologies Y. Fan China Telecom Expires: January 2013 July 5, 2012 Virtual Subnet: A Host Route based Subnet Extension Solution draft-xu-virtual-subnet-08 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 January 5, 2012. Copyright Notice Copyright (c) 2009 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. Xu, et al. Expires January 5, 2013 [Page 1] Internet-Draft Virtual Subnet July 2012 Abstract This document describes a host route based subnet extension solution referred to as Virtual Subnet, which mainly reuses existing BGP/MPLS IP VPN [RFC4364] and ARP proxy [RFC925][RFC1027] technologies. Virtual Subnet provides a scalable approach for interconnecting geographically dispersed cloud data centers. 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]. Table of Contents 1. Introduction ................................................ 3 2. Terminology ................................................. 5 3. Solution Description......................................... 5 3.1. Unicast ................................................ 5 3.1.1. Intra-subnet Unicast .............................. 5 3.1.2. Inter-subnet Unicast .............................. 6 3.2. Multicast/Broadcast .................................... 9 3.3. CE Host Discovery ..................................... 10 3.4. ARP Proxy ............................................. 10 3.5. CE Host Mobility ...................................... 10 3.6. Forwarding Table Scalability .......................... 11 3.6.1. MAC Table Reduction on Data Center Switches ...... 11 3.6.2. FIB Reduction on PE Routers ...................... 11 3.6.3. RIB Reduction on PE Routers ...................... 13 3.7. ARP Table Scalability on Default Gateways ............. 14 3.8. ARP/Unknown Uncast Flood Avoidance .................... 15 3.9. Active-active Multi-homing ............................ 15 3.10. Path Optimization .................................... 15 4. Future Work ................................................ 15 5. Security Considerations .................................... 16 6. IANA Considerations ........................................ 16 7. Acknowledgements ........................................... 16 8. References ................................................. 16 8.1. Normative References .................................. 16 8.2. Informative References ................................ 16 Authors' Addresses ............................................ 17 Xu, et al. Expires January 5, 2013 [Page 2] Internet-Draft Virtual Subnet July 2012 1. Introduction For business continuity purposes, Virtual Machine (VM) migration across data centers is commonly used in those situations such as data center maintenance, data center migration, data center consolidation, data center expansion, and data center disaster avoidance. It's obvious that IP renumbering of servers (i.e., VMs) after the migration is usually complex and costly and therefore would prolong the business downtime during the process of migration. To allow the seamless migration of a VM from one data center to another without IP renumbering, the subnet on which the VM resides needs to be extended across these data centers. In the Infrastructure-as-a-Service (IaaS) cloud data center environments, to achieve subnet extension across multiple data centers in a scalable way, the following requirements SHOULD be considered for any data center interconnect solution: 1) VPN Instance Scalability In a modern cloud data center environment, thousands or even tens of thousands of tenants could be hosted over a shared network infrastructure. For security and performance isolation considerations, these tenants need to be isolated from one another. Hence, the data center interconnect solution SHOULD be capable of providing a large enough VPN space for tenant isolation. 2) Forwarding Table Scalability With the development of virtualization technologies, a single cloud data center containing millions of VMs is not uncommon today. This number already implies a big challenge for data center switches, especially for core/aggregation switches, from the perspective of forwarding table scalability. Provided that multiple data centers of such scale were interconnected at layer2, this challenge would be even worse. Hence an ideal data center interconnect solution SHOULD prevent the forwarding table size of data center switches from growing by folds as the number of data centers to be interconnected increases. Furthermore, if any kind of L2VPN or L3VPN technologies is used for interconnecting data centers, the scale of forwarding tables on PE routers SHOULD be taken into consideration as well. 3) ARP Table Scalability on Default Gateways Xu, et al. Expires January 5, 2013 [Page 3] Internet-Draft Virtual Subnet July 2012 [NARTEN-ARMD] notes that the ARP tables maintained by data center default gateways in cloud data centers can raise both scalability and security issues. Therefore, an ideal data center interconnect solution SHOULD prevent the ARP table size from growing by multiples as the number of data centers to be connected increases. 4) ARP/Unknown Unicast Flood Suppression or Avoidance It's well-known that the flooding of ARP broadcast and unknown unicast traffic within a large Layer2 network will lead to certain performance impact on both networks and hosts. As multiple data centers each containing millions of VMs are interconnected together across the Wide Area Network (WAN) at layer2, the impact of flooding as mentioned above will become even worse. As such, how to suppress or even avoid the flooding of ARP broadcast and unknown unicast traffic across data centers becomes increasingly desirable for the purpose of avoiding the unnecessary consumption of network bandwidth resources and service CPU resources. 5) Active-active Multi-homing In order to utilize the bandwidth of all available paths between the data center and the transport network in addition to providing resilient connectivity between them, active-active multi-homing is increasingly advocated by data center operators as a replacement of the traditional active-standby multi-homing approach. 6) Path Optimization A subnet usually indicates a location in the network. However, when a subnet has been extended across multiple geographically dispersed data center locations, the location semantics of such subnet is not retained any longer. As a result, the traffic from a cloud user (i.e., a VPN user) which is destined for a given server located at one data center location of such extended subnet may arrive at another data center location firstly according to the subnet route, and then be forwarded to the location where the service is actually located. This suboptimal routing would obviously result in the unnecessary consumption of the bandwidth resources which are intended for data center interconnection. Furthermore, in the case where the traditional VPLS technology [RFC4761, RFC4762] is used for data center interconnect and default gateways of different data center locations are configured within the same virtual router redundancy group, the returning traffic from that server to the Xu, et al. Expires January 5, 2013 [Page 4] Internet-Draft Virtual Subnet July 2012 cloud user may be forwarded at layer2 to a default gateway at one of remote data center locations, rather than the one at the local data center location. This suboptimal routing would also unnecessarily consume the bandwidth resources which are intended for data center interconnect. This document describes a host route based subnet extension solution referred to as Virtual Subnet (VS), which can meet all of the requirements of cloud data center interconnect as described above. Since VS mainly reuses existing technologies including BGP/MPLS IP VPN [RFC4364] and ARP proxy [RFC925][RFC1027], it allows service providers who are offering IaaS cloud services to the public to interconnect their geographically dispersed data centers in a much scalable may, and more importantly, they can accomplish this interconnection on basis of their existing MPLS/BGP IP VPN infrastructures and their years of experience in the operation and provisioning of MPLS/BGP IP VPN services. Please note that VS is targeted at scenarios where the traffic across data centers is routable IP traffic. In such scenario, data center operators who are implementing data center interconnect could benefit from the advantages that such host route based subnet extension solution exclusively has, such as MAC table reduction on data center switches, ARP table reduction on data center default gateways, path optimization for inter-subnet traffic, and so on. 2. Terminology This memo makes use of the terms defined in [RFC4364], [RFC2338] [MVPN] and [VA-AUTO]. 3. Solution Description 3.1. Unicast 3.1.1. Intra-subnet Unicast As shown in Figure 1, two CE hosts (i.e., Host A and B) which are configured within the same subnet (i.e., 1.1.1.0/24) are located in two different data centers (i.e., DC West and DC East) respectively. PE routers (i.e., PE-1 and PE-2) which are used for interconnecting the above two data centers create host routes for their local CE hosts respectively and then redistribute these routes into BGP. Meanwhile, ARP proxy is enabled on the VRF attachment circuits of these PE routers. Xu, et al. Expires January 5, 2013 [Page 5] Internet-Draft Virtual Subnet July 2012 +--------------------+ +-----------------+ | | +-----------------+ |VPN_A:1.1.1.1/24 | | | |VPN_A:1.1.1.1/24 | | \ | | | | / | | +------+ \++---+-+ +-+---++/ +------+ | | |Host A+----+ PE-1 | | PE-2 +----+Host B| | | +------+\ ++-+-+-+ +-+-+-++ /+------+ | | 1.1.1.2/24 | | | | | | 1.1.1.3/24 | | | | | | | | | | DC West | | | IP/MPLS Backbone | | | DC East | +-----------------+ | | | | +-----------------+ | +--------------------+ | | | VRF_A : V VRF_A : V +------------+---------+--------+ +------------+---------+--------+ | Prefix | Nexthop |Protocol| | Prefix | Nexthop |Protocol| +------------+---------+--------+ +------------+---------+--------+ | 1.1.1.1/32 |127.0.0.1| Direct | | 1.1.1.1/32 |127.0.0.1| Direct | +------------+---------+--------+ +------------+---------+--------+ | 1.1.1.2/32 | 1.1.1.2 | Direct | | 1.1.1.2/32 | PE-1 | IBGP | +------------+---------+--------+ +------------+---------+--------+ | 1.1.1.3/32 | PE-2 | IBGP | | 1.1.1.3/32 | 1.1.1.3 | Direct | +------------+---------+--------+ +------------+---------+--------+ | 1.1.1.0/24 | 1.1.1.1 | Direct | | 1.1.1.0/24 | 1.1.1.1 | Direct | +------------+---------+--------+ +------------+---------+--------+ Figure 1: Intra-subnet Unicast Example Now assume host A sends an ARP request for host B before communicating to host B. Upon receiving the ARP request, PE-1 as an ARP proxy returns its own MAC address as a response. Host A then sends IP packets for host B to PE-1. Strictly according to the normal L3VPN forwarding procedure, PE-1 tunnels such packets towards PE-2 which in turn forwards them to host B. In this way, host A and B could communicate with each other as if they were located within the same subnet or Local Area Network (LAN). In fact, such subnet is a virtual subnet which is emulated by using host routes, rather than a real subnet. 3.1.2. Inter-subnet Unicast As shown in Figure 2, only one data center (i.e., DC East) is deployed with a default gateway (i.e., GW). PE-2 which is connected to GW would either be configured with or learn from GW a default route with its next-hop being pointed to GW, and this route is distributed to other PE routers (i.e., PE-1) as per normal [RFC4364] operation. Assume host A sends an ARP request for its default gateway (i.e., 1.1.1.4) prior to communicating with a destination host outside of its subnet (i.e., 1.1.1.0/24). Upon receiving this Xu, et al. Expires January 5, 2013 [Page 6] Internet-Draft Virtual Subnet July 2012 ARP request, PE-1 as an ARP proxy returns its own MAC address as a response. Host A then sends a packet for the destination host to PE- 1. PE-1 forwards such packet towards PE-2 according to the default route learnt from PE-2, which in turn forwards that packet to GW according to the default route as well. In contrast, if host B sends an ARP request for its default gateway (i.e., 1.1.1.4) prior to communicate with a destination host outside of its subnet, it will receive an ARP response from GW. As such, the packet destined for the destination host will be forwarded directly to GW. Note that since the outgoing interface of the best-match route for the target host (i.e., 1.1.1.4) is the same as the one over which the ARP packet arrived, PE-2 would not respond to this ARP request. +--------------------+ +-----------------+ | | +-----------------+ |VPN_A:1.1.1.1/24 | | | |VPN_A:1.1.1.1/24 | | \ | | | | / | | +------+ \++---+-+ +-+---++/ +------+ | | |Host A+------+ PE-1 | | PE-2 +-+----+Host B| | | +------+\ ++-+-+-+ +-+-+-++ | /+------+ | | 1.1.1.2/24 | | | | | | | 1.1.1.3/24 | | GW=1.1.1.4 | | | | | | | GW=1.1.1.4 | | | | | | | | | +------+ | | | | | | | | +----+ GW +--| | | | | | | | /+------+ | | | | | | | | 1.1.1.4/24 | | | | | | | | | | DC West | | | IP/MPLS Backbone | | | DC East | +-----------------+ | | | | +-----------------+ | +--------------------+ | | | VRF_A : V VRF_A : V +------------+---------+--------+ +------------+---------+--------+ | Prefix | Nexthop |Protocol| | Prefix | Nexthop |Protocol| +------------+---------+--------+ +------------+---------+--------+ | 1.1.1.1/32 |127.0.0.1| Direct | | 1.1.1.1/32 |127.0.0.1| Direct | +------------+---------+--------+ +------------+---------+--------+ | 1.1.1.2/32 | 1.1.1.2 | Direct | | 1.1.1.2/32 | PE-1 | IBGP | +------------+---------+--------+ +------------+---------+--------+ | 1.1.1.3/32 | PE-2 | IBGP | | 1.1.1.3/32 | 1.1.1.3 | Direct | +------------+---------+--------+ +------------+---------+-------- | 1.1.1.4/32 | PE-2 | IBGP | | 1.1.1.4/32 | 1.1.1.4 | Direct | +------------+---------+--------+ +------------+---------+--------+ | 1.1.1.0/24 | 1.1.1.1 | Direct | | 1.1.1.0/24 | 1.1.1.1 | Direct | +------------+---------+--------+ +------------+---------+--------+ | 0.0.0.0/0 | PE-2 | IBGP | | 0.0.0.0/0 | 1.1.1.4 | Static | +------------+---------+--------+ +------------+---------+--------+ Figure 2: Inter-subnet Unicast Example (1) Xu, et al. Expires January 5, 2013 [Page 7] Internet-Draft Virtual Subnet July 2012 As shown in Figure 3, in this case where each data center is deployed with a default gateway, CE hosts will get ARP responses from their local default gateways, rather than from their local PE routers when sending ARP requests for their default gateways. +--------------------+ +-----------------+ | | +-----------------+ |VPN_A:1.1.1.1/24 | | | |VPN_A:1.1.1.1/24 | | \ | | | | / | | +------+ \++---+-+ +-+---++/ +------+ | | |Host A+----+-+ PE-1 | | PE-2 +-+----+Host B| | | +------+\ | ++-+-+-+ +-+-+-++ | /+------+ | | 1.1.1.2/24 | | | | | | | | 1.1.1.3/24 | | GW=1.1.1.4 | | | | | | | | GW=1.1.1.4 | | +------+ | | | | | | | | +------+ | |--+ GW-1 +----+ | | | | | | +----+ GW-2 +--| | +------+\ | | | | | | /+------+ | | 1.1.1.4/24 | | | | | | 1.1.1.4/24 | | | | | | | | | | DC West | | | IP/MPLS Backbone | | | DC East | +-----------------+ | | | | +-----------------+ | +--------------------+ | | | VRF_A : V VRF_A : V +------------+---------+--------+ +------------+---------+--------+ | Prefix | Nexthop |Protocol| | Prefix | Nexthop |Protocol| +------------+---------+--------+ +------------+---------+--------+ | 1.1.1.1/32 |127.0.0.1| Direct | | 1.1.1.1/32 |127.0.0.1| Direct | +------------+---------+--------+ +------------+---------+--------+ | 1.1.1.2/32 | 1.1.1.2 | Direct | | 1.1.1.2/32 | PE-1 | IBGP | +------------+---------+--------+ +------------+---------+--------+ | 1.1.1.3/32 | PE-2 | IBGP | | 1.1.1.3/32 | 1.1.1.3 | Direct | +------------+---------+--------+ +------------+---------+-------- | 1.1.1.4/32 | 1.1.1.4 | Direct | | 1.1.1.4/32 | 1.1.1.4 | Direct | +------------+---------+--------+ +------------+---------+--------+ | 1.1.1.0/24 | 1.1.1.1 | Direct | | 1.1.1.0/24 | 1.1.1.1 | Direct | +------------+---------+--------+ +------------+---------+--------+ | 0.0.0.0/0 | 1.1.1.4 | Static | | 0.0.0.0/0 | 1.1.1.4 | Static | +------------+---------+--------+ +------------+---------+--------+ Figure 3: Inter-subnet Unicast Example (2) Alternatively, as shown in Figure 4, PE routers themselves could be directly configured as the default gateways of their locally connected CE hosts as long as these PE routers have routes for the outside networks. Xu, et al. Expires January 5, 2013 [Page 8] Internet-Draft Virtual Subnet July 2012 +------+ +------+ PE-3 +------+ +-----------------+ | +------+ | +-----------------+ |VPN_A:1.1.1.1/24 | | | |VPN_A:1.1.1.1/24 | | \ | | | | / | | +------+ \++---+-+ +-+---++/ +------+ | | |Host A+------+ PE-1 | | PE-2 +------+Host B| | | +------+\ ++-+-+-+ +-+-+-++ /+------+ | | 1.1.1.2/24 | | | | | | 1.1.1.3/24 | | GW=1.1.1.1 | | | | | | GW=1.1.1.1 | | | | | | | | | | DC West | | | IP/MPLS Backbone | | | DC East | +-----------------+ | | | | +-----------------+ | +--------------------+ | | | VRF_A : V VRF_A : V +------------+---------+--------+ +------------+---------+--------+ | Prefix | Nexthop |Protocol| | Prefix | Nexthop |Protocol| +------------+---------+--------+ +------------+---------+--------+ | 1.1.1.1/32 |127.0.0.1| Direct | | 1.1.1.1/32 |127.0.0.1| Direct | +------------+---------+--------+ +------------+---------+--------+ | 1.1.1.2/32 | 1.1.1.2 | Direct | | 1.1.1.2/32 | PE-1 | IBGP | +------------+---------+--------+ +------------+---------+--------+ | 1.1.1.3/32 | PE-2 | IBGP | | 1.1.1.3/32 | 1.1.1.3 | Direct | +------------+---------+--------+ +------------+---------+--------+ | 1.1.1.0/24 | 1.1.1.1 | Direct | | 1.1.1.0/24 | 1.1.1.1 | Direct | +------------+---------+--------+ +------------+---------+--------+ | 0.0.0.0/0 | PE-3 | IBGP | | 0.0.0.0/0 | PE-3 | IBGP | +------------+---------+--------+ +------------+---------+--------+ Figure 4: Inter-subnet Unicast Example (3) 3.2. Multicast/Broadcast To support IP multicast and broadcast between CE hosts of the same virtual subnet, the MVPN technology [MVPN] could be directly reused. For example, PE routers attached to a given VPN join a default provider multicast distribution tree which is dedicated for that VPN. Ingress PE routers, upon receiving multicast or broadcast packets from their local CE hosts, forward them towards remote PE routers through the corresponding default provider multicast distribution tree. More details about how to support multicast and broadcast in VS will be explored in a later version of this document. Xu, et al. Expires January 5, 2013 [Page 9] Internet-Draft Virtual Subnet July 2012 3.3. CE Host Discovery PE routers MUST be able to discovery their local CE hosts in time, especially after rebooting up, and meanwhile keep the list of local CE hosts up to date in a timely manner so as to ensure the availability of the host route information. PE routers could accomplish local CE host discovery by some traditional host discovery means such as ARP scan and/or ICMP scan. Furthermore, Link Layer Discovery Protocol (LLDP) described in [802.1AB] or VSI Discovery and Configuration Protocol (VDP) described in [802.1Qbg], or even interaction with the data center orchestration system could also be considered as a means of local CE host discovery. More details about local CE host discovery in VS will be explored in a later version of this document. 3.4. ARP Proxy Acting as an ARP proxy, PE router SHOULD only respond to an ARP request for the target host for which there is a route in the associated VRF and the outgoing interface of the route is different from the one over which the ARP request arrived. Otherwise, PE router would not respond. In the scenario where a given VPN site (i.e., a data center) is multi-homed to more than one PE router via an Ethernet switch or an Ethernet network, VRRP is usually enabled on these PE routers for router redundancy purposes. In this case, only the PE router which has been elected as the VRRP master is entitled to perform the ARP proxy function and furthermore it SHOULD respond with the virtual IP address, rather than its physical IP address. 3.5. CE Host Mobility After moving from one VPN site to another, a CE host (e.g., a VM) will send a gratuitous ARP packet. Upon receiving that packet, PE router attached to the new site will create a host route for that CE host and then advertise it to remote PE routers. PE router which that CE host was previously attached to, upon learning such route, would immediately check whether that CE host is still connected to it by some means (e.g., ARP PING and/or ICMP PING). If not, the PE router would withdraw the corresponding host route which has been advertised before. Meanwhile, the PE router would broadcast a gratuitous ARP packet on behalf of that CE host. As such, the ARP entry of that CE host which was cached on any local CE host would be updated accordingly. Xu, et al. Expires January 5, 2013 [Page 10] Internet-Draft Virtual Subnet July 2012 3.6. Forwarding Table Scalability 3.6.1. MAC Table Reduction on Data Center Switches In VS, the MAC learning domain associated with a given virtual subnet which has been extended across multiple data centers is partitioned into segments and each of the segments is confined within a single data center. Therefore data center switches only needs to learn local MAC addresses, rather than learning both local and remote MAC addresses as required in the case where the traditional VPLS technology [RFC4761, RFC4762] is used for data center interconnect. 3.6.2. FIB Reduction on PE Routers +------+ +------+RR/APR+------+ +-----------------+ | +------+ | +-----------------+ |VPN_A:1.1.1.1/24 | | | |VPN_A:1.1.1.1/24 | | \ | | | | / | | +------+ \++---+-+ +-+---++/ +------+ | | |Host A+------+ PE-1 | | PE-2 +------+Host B| | | +------+\ ++-+-+-+ +-+-+-++ /+------+ | | 1.1.1.2/24 | | | | | | 1.1.1.3/24 | | | | | | | | | | DC West | | | IP/MPLS Backbone | | | DC East | +-----------------+ | | | | +-----------------+ | +--------------------+ | | | VRF_A : V VRF_A : V +------------+---------+--------+------+ +------------+---------+--------+------+ | Prefix | Nexthop |Protocol|In_FIB| | Prefix | Nexthop |Protocol|In_FIB| +------------+---------+--------+------+ +------------+---------+--------+------+ | 1.1.1.1/32 |127.0.0.1| Direct | Yes | | 1.1.1.1/32 |127.0.0.1| Direct | Yes | +------------+---------+--------+------+ +------------+---------+--------+------+ | 1.1.1.2/32 | 1.1.1.2 | Direct | Yes | | 1.1.1.2/32 | PE-1 | IBGP | No | +------------+---------+--------+------+ +------------+---------+--------+------+ | 1.1.1.3/32 | PE-2 | IBGP | No | | 1.1.1.3/32 | 1.1.1.3 | Direct | Yes | +------------+---------+--------+------+ +------------+---------+--------+------+ | 1.1.1.0/25 | RR | IBGP | Yes | | 1.1.1.0/25 | RR | IBGP | Yes | +------------+---------+--------+------+ +------------+---------+--------+------+ |1.1.1.128/25| RR | IBGP | Yes | |1.1.1.128/25| RR | IBGP | Yes | +------------+---------+--------+------+ +------------+---------+--------+------+ | 1.1.1.0/24 | 1.1.1.1 | Direct | Yes | | 1.1.1.0/24 | 1.1.1.1 | Direct | Yes | +------------+---------+--------+------+ +------------+---------+--------+------+ Figure 5: FIB Reduction Example Xu, et al. Expires January 5, 2013 [Page 11] Internet-Draft Virtual Subnet July 2012 To reduce the FIB size of PE routers, Virtual Aggregation (VA) [VA- AUTO] technology can be used here. Take the VPN instance A shown in Figure 5 as an example, the procedures of FIB reduction are as follows: 1) Multiple more specific prefixes (e.g., 1.1.1.0/25 and 1.1.1.128/25) equivalent to the prefix of virtual subnet (i.e., 1.1.1.0/24) are configured as Virtual Prefixes (VPs) and a Route- Reflector (RR) is configured as an Aggregation Point Router (APR) for these VPs. PE routers as RR clients advertise host routes for their own local CE hosts to the RR which in turn, as an APR, installs those host routes into FIB and then attach the ''can- suppress'' tag to those host routes before reflecting them to its clients. Those host routes which have been attached with that tag would not be installed into FIB by clients who are VA-aware since they are not APRs for those host routes. In addition, the RR as an APR would advertise the corresponding VP routes to all of its clients, and those of which who are VA-aware in turn would install these VP routes into FIB. Upon receiving a packet from a local CE host, if no matching host route found, the ingress PE router will forward the packet to the RR according to one of the VP routes learnt from the RR, which in turn forwards the packet to the egress PE router according to the host route learnt from that egress PE router. In a word, the FIB table size of PE routers can be greatly reduced at the cost of path stretch. Note that in the case where the RR is not available for transferring L3VPN traffic between PE routers due to some reason, the APR function could actually be performed by a given PE router other than the RR as long as that PE router has installed all host routes belonging to the virtual subnet into its FIB. In this way, the RR only needs to attach a ''can-suppress'' tag to the host routes learnt from its clients before reflecting them to the other clients. Furthermore, PE routers themselves could directly attach the ''can-suppress'' tag to the host routes for their local CE hosts before distributing them to remote peers. 2) Provided a given local CE host sends an ARP request for a remote CE host, ingress PE router receiving such request will immediately install the host route for that remote CE host into FIB, in case there is a host route for that CE host in RIB and which has not yet been installed into FIB. Therefore, the subsequent packets destined for that remote CE host will be forwarded directly to the egress PE router. Note that the FIB entries corresponding to remote host routes would expire if they have not been used for routing packets for a certain period of time. Xu, et al. Expires January 5, 2013 [Page 12] Internet-Draft Virtual Subnet July 2012 3.6.3. RIB Reduction on PE Routers +------+ +------+ RR +------+ +-----------------+ | +------+ | +-----------------+ |VPN_A:1.1.1.1/24 | | | |VPN_A:1.1.1.1/24 | | \ | | | | / | | +------+ \++---+-+ +-+---++/ +------+ | | |Host A+------+ PE-1 | | PE-2 +------+Host B| | | +------+\ ++-+-+-+ +-+-+-++ /+------+ | | 1.1.1.2/24 | | | | | | 1.1.1.3/24 | | | | | | | | | | DC West | | | IP/MPLS Backbone | | | DC East | +-----------------+ | | | | +-----------------+ | +--------------------+ | | | VRF_A : V VRF_A : V +------------+---------+--------+ +------------+---------+--------+ | Prefix | Nexthop |Protocol| | Prefix | Nexthop |Protocol| +------------+---------+--------+ +------------+---------+--------+ | 1.1.1.1/32 |127.0.0.1| Direct | | 1.1.1.1/32 |127.0.0.1| Direct | +------------+---------+--------+ +------------+---------+--------+ | 1.1.1.2/32 | 1.1.1.2 | Direct | | 1.1.1.3/32 | 1.1.1.3 | Direct | +------------+---------+--------+ +------------+---------+--------+ | 1.1.1.0/25 | RR | IBGP | | 1.1.1.0/25 | RR | IBGP | +------------+---------+--------+ +------------+---------+--------+ |1.1.1.128/25| RR | IBGP | |1.1.1.128/25| RR | IBGP | +------------+---------+--------+ +------------+---------+--------+ | 1.1.1.0/24 | 1.1.1.1 | Direct | | 1.1.1.0/24 | 1.1.1.1 | Direct | +------------+---------+--------+ +------------+---------+--------+ Figure 6: RIB Reduction Example To reduce the RIB size of PE routers, BGP Outbound Route Filtering (ORF) mechanism is used to realize on-demand route announcement. Take the VPN instance A shown in Figure 6 as an example, the procedures of RIB reduction are as follows: 1) PE routers as RR clients advertise host routes for their local CE hosts to a RR which in turn, however doesn't reflect these host routes by default unless it receives explicit ORF requests for them from its clients. The RR is configured with routes for more specific subnets (e.g., 1.1.1.0/25 and 1.1.1.128/25) equivalent to the virtual subnet (i.e., 1.1.1.0/24) with next-hop being pointed to Null0 and then advertises these routes to its clients via BGP. Upon receiving a packet from a local CE host, if no matching host route found, ingress PE router will forward the packet to the RR according to one of the subnet routes learnt from the RR, which in turn forwards the packet to the egress PE router according to the Xu, et al. Expires January 5, 2013 [Page 13] Internet-Draft Virtual Subnet July 2012 host route learnt from that egress PE router. In a word, the RIB table size of PE routers can be greatly reduced at the cost of path stretch. Just as the approach mentioned in section 3.6.2, in the case where the RR is not available for transferring L3VPN traffic between PE routers due to some reason, a PE router other than the RR could advertise the more specific subnet routes as long as that PE router has installed all host routes belonging to that virtual subnet into its FIB. 2) Provided a given local CE host sends an ARP request for a remote CE host, ingress PE router receiving such request will request the corresponding host route from its RR by using ORF (e.g., a group ORF containing Route-Target (RT) and prefix information) in case there is no host route for that CE host yet in its RIB. Once the host route for the remote CE host is learnt from the RR, the subsequent packets destined for that CE host would be forwarded directly to the egress PE router. Note that the RIB entries of remote host routes could expire if they have not been used for routing packets for a certain period of time. Once the expiration time for a given RIB entry is approaching, the PE router would notice its RR to withdraw the corresponding host route by sending an ORF message. Upon receiving the corresponding withdraw message from its RR, the PE router will delete that host route from its RIB accordingly. 3.7. ARP Table Scalability on Default Gateways In the case where data center default gateway functions are implemented on PE routers of the VS as shown in Figure 4, since the ARP table on each PE router only needs to contain ARP entries of local CE hosts, the ARP table size will not grow accordingly as the number of data centers to be connected increases. Alternatively, if dedicated default gateways are directly connected to PE routers of the VS as shown in Figure 3. Due to the use of ARP proxy on PE routers, all remote CE hosts of a given virtual subnet share the same MAC address (i.e., the MAC address of the local PE router) from the point of view of default gateways. Therefore, ARP entries of those remote CE hosts could be aggregated into one ARP entry (i.e., 1.1.1.0/24-> the MAC address of the PE router). Accordingly, default gateways are required to use the longest- matching algorithm for ARP cache lookup instead of the existing exact-matching algorithm. In this way, the ARP table size of DC gateways can be reduced greatly as well. Xu, et al. Expires January 5, 2013 [Page 14] Internet-Draft Virtual Subnet July 2012 3.8. ARP/Unknown Uncast Flood Avoidance In VS, the flooding domain associated with a given virtual subnet that has been extended across multiple data centers, has been partitioned into segments and each of the segments is confined within a single data center. Therefore, the performance impact on networks and servers caused by the flooding of ARP broadcast and unknown unicast traffic is alleviated. 3.9. Active-active Multi-homing For the PE router redundancy purpose, a VPN site could be multi- homed to more than one PE router. In this case, VRRP [RFC2338] SHOULD be enabled on these PE routers and only the PE router which has been elected as the VRRP master could perform the ARP proxy functionality. However, all PE routers, either as a VRRP master or a VRRP slave, are allowed to advertise host routes for their local CE hosts. Hence, from the perspective of remote PE routers, there will be multiple host routes for a given CE host located within that multi-homed site. In other words, active-active multi-homing is available for the inbound traffic of a given multi-homed site. 3.10. Path Optimization Take the scenario shown in Figure 4 as an example, to optimize the forwarding path for traffic between enterprise sites (e.g., cloud users) and cloud data centers, PE routers located at cloud data centers (i.e., PE-1 and PE-2), which also perform the role of data center default gateway, could propagate host routes for their local CE hosts respectively to remote PE routers which are attached to enterprise sites (i.e., PE-3). As such, the traffic from enterprise sites to a given server on the virtual subnet which has been extended across data centers would be forwarded directly to the data center location at which that server is actually located, since the traffic is now forwarded on basis of the host route for that server, rather than the subnet route. Furthermore, for the traffic from the cloud data center to enterprise sites, since each PE router acting as an default gateway would forward the traffic received from its local CE hosts directly to the remote PE routers (i.e., PE-3) according to the best-match route in the corresponding VRF, and as a result, the traffic from data centers to enterprise sites is forwarded along the optimal path without consuming the data center interconnect bandwidth resources. 4. Future Work How to support IPv6 CE hosts in VS is for future study. Xu, et al. Expires January 5, 2013 [Page 15] Internet-Draft Virtual Subnet July 2012 5. Security Considerations TBD. 6. IANA Considerations There is no requirement for IANA. 7. Acknowledgements Thanks to Dino Farinacci, Himanshu Shah, Nabil Bitar, Giles Heron, Ronald Bonica, Monique Morrow and Christian Jacquenet for their valuable comments and suggestions on this document. 8. References 8.1. Normative References [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, March 1997. 8.2. Informative References [RFC4364] Rosen. E and Y. Rekhter, "BGP/MPLS IP Virtual Private Networks (VPNs)", RFC 4364, February 2006. [MVPN] Rosen. E and Aggarwal. R, "Multicast in MPLS/BGP IP VPNs", draft-ietf-l3vpn-2547bis-mcast-10.txt, Work in Progress, Janurary 2010. [VA-AUTO] Francis, P., Xu, X., Ballani, H., Jen, D., Raszuk, R., and L. Zhang, "Auto-Configuration in Virtual Aggregation", draft-ietf-grow-va-auto-05.txt, Work in Progress, December 2011. [RFC925] Postel, J., "Multi-LAN Address Resolution", RFC-925, USC Information Sciences Institute, October 1984. [RFC1027] Smoot Carl-Mitchell, John S. Quarterman, ''Using ARP to Implement Transparent Subnet Gateways'', RFC 1027, October 1987. [RFC2338] Knight, S., et al., "Virtual Router Redundancy Protocol", RFC 2338, April 1998. Xu, et al. Expires January 5, 2013 [Page 16] Internet-Draft Virtual Subnet July 2012 [RFC4761] Kompella, K. and Y. Rekhter, "Virtual Private LAN Service (VPLS) Using BGP for Auto-Discovery and Signaling", RFC 4761, January 2007. [RFC4762] Lasserre, M. and V. Kompella, "Virtual Private LAN Service (VPLS) Using Label Distribution Protocol (LDP) Signaling", RFC 4762, January 2007. [802.1AB] IEEE Standard 802.1AB-2009, "Station and Media Access Control Connectivity Discovery", September 17, 2009. [802.1Qbg] IEEE Draft Standard P802.1Qbg/D2.0, "Virtual Bridged Local Area Networks -Amendment XX: Edge Virtual Bridging", Work in Progress, December 1, 2011. [NARTEN-ARMD] Narten, T., Karir, M., and I. Foo, "Problem Statement for ARMD", draft-ietf-armd-problem-statement-01.txt, Work in Progress, February 2012. Authors' Addresses Xiaohu Xu Huawei Technologies, Beijing, China. Phone: +86 10 60610041 Email: xuxiaohu@huawei.com Susan Hares Huawei Technologies (FutureWei group) 2330 Central Expressway Santa Clara, CA 95050 Phone: +1-734-604-0332 Email: Susan.Hares@huawei.com Yongbing Fan Guangzhou Institute,China Telecom Guangzhou, China. Phone: +86 20 38639121 Email: fanyb@gsta.com Xu, et al. Expires January 5, 2013 [Page 17]