draft-ietf-lsvr-applicability-01.xml

<?xml version="1.0"?>
<!DOCTYPE rfc SYSTEM "rfc2629.dtd" [
<!-- One method to get references from the online citation libraries.
     There has to be one entity for each item to be referenced.
     An alternate method (rfc include) is described in the references. -->
<!ENTITY RFC2119 SYSTEM
  "http://xml.resource.org/public/rfc/bibxml/reference.RFC.2119.xml">
<!ENTITY RFC2328 SYSTEM
  "http://xml.resource.org/public/rfc/bibxml/reference.RFC.2328.xml">
<!ENTITY RFC4271 SYSTEM
  "http://xml.resource.org/public/rfc/bibxml/reference.RFC.4271.xml">
<!ENTITY RFC2328 SYSTEM
  "http://xml.resource.org/public/rfc/bibxml/reference.RFC.2328.xml">
<!ENTITY RFC5580 SYSTEM
  "http://xml.resource.org/public/rfc/bibxml/reference.RFC.5580.xml">
<!ENTITY RFC4957 SYSTEM
  "http://xml.resource.org/public/rfc/bibxml/reference.RFC.4957.xml">
<!ENTITY RFC7752 SYSTEM
  "http://xml.resource.org/public/rfc/bibxml/reference.RFC.7752.xml">
<!ENTITY RFC7938 SYSTEM
  "http://xml.resource.org/public/rfc/bibxml/reference.RFC.7938.xml">
<!ENTITY RFC5925 SYSTEM
  "http://xml.resource.org/public/rfc/bibxml/reference.RFC.5925.xml">
<!ENTITY RFC7938 SYSTEM
  "http://xml.resource.org/public/rfc/bibxml/reference.RFC.7938.xml">
<!ENTITY RFC8174 SYSTEM
  "http://xml.resource.org/public/rfc/bibxml/reference.RFC.8174.xml">
<!ENTITY I-D.draft-ietf-lsvr-bgp-spf SYSTEM
  "http://xml.resource.org/public/rfc/bibxml3/reference.I-D.draft-ietf-lsvr-bgp-spf-03.xml">
<!ENTITY I-D.draft-acee-idr-lldp-peer-discovery SYSTEM
  "http://xml.resource.org/public/rfc/bibxml3/reference.I-D.draft-acee-idr-lldp-peer-discovery-03.xml">
<!ENTITY I-D.draft-xu-idr-neighbor-autodiscovery SYSTEM
  "http://xml.resource.org/public/rfc/bibxml3/reference.I-D.draft-xu-idr-neighbor-autodiscovery-10.xml">
<!ENTITY I-D.draft-ymbk-lsvr-lsoe SYSTEM
  "http://xml.resource.org/public/rfc/bibxml3/reference.I-D.draft-ymbk-lsvr-lsoe-01.xml">
<!ENTITY I-D.draft-li-lsr-dynamic-flooding SYSTEM
  "http://xml.resource.org/public/rfc/bibxml3/reference.I-D.draft-li-lsr-dynamic-flooding-01.xml">
]>

<?rfc comments="yes"?>
<?rfc compact="no"?>
<?rfc inline="yes"?>
<?rfc sortrefs="yes"?>
<?rfc subcompact="no"?>
<?rfc symrefs="yes"?>
<?rfc toc="yes"?>
<?rfc tocdepth="5"?>
<?rfc tocindent="yes"?>
<?rfc tocompact="yes"?>

<rfc category="info" docName="draft-ietf-lsvr-applicability-01.txt"
     ipr="trust200902">

  <front>

    <title abbrev="">Usage and Applicability of Link State Vector Routing in Data Centers </title>

    <author fullname="Keyur Patel" initials="K"
            surname="Patel">
      <organization>Arrcus, Inc.</organization>
      <address>
        <postal>
          <street>2077 Gateway Pl</street>
          <city>San Jose, CA</city>
          <country>USA</country>
          <code>95110</code>
        </postal>
        <phone></phone>
        <email>keyur@arrcus.com</email>
      </address>
    </author>

    <author fullname="Acee Lindem" initials="A"
            surname="Lindem">
      <organization>Cisco Systems</organization>
      <address>
        <postal>
          <street>301 Midenhall Way</street>
          <city>Cary, NC</city>
          <country>USA</country>
          <code>95110</code>
        </postal>
        <phone></phone>
        <email>acee@cisco.com</email>
      </address>
    </author>
        <author fullname="Shawn Zandi" initials="S"
            surname="Zandi">

      <organization>Linkedin</organization>

      <address>
        <postal>
          <street>222 2nd Street</street>

          <city>San Francisco</city>

          <region>CA</region>

          <code>94105</code>

          <country>USA</country>
        </postal>

        <email>szandi@linkedin.com</email>

      </address>
</author>
        <author fullname="Gaurav Dawra" initials="G"
            surname="Dawra">

      <organization>Linkedin</organization>

      <address>
        <postal>
          <street>222 2nd Street</street>

          <city>San Francisco</city>

          <region>CA</region>

          <code>94105</code>

          <country>USA</country>
        </postal>

        <email>gdawra@linkedin.com</email>

      </address>
</author>
    <date/>

    <area>Routing</area>

    <workgroup>LSVR</workgroup>

    <abstract>
      <t>
	This document discusses the usage and applicability of Link State Vector Routing
	(LSVR) extensions in the CLOS architecture of Data Center Networks. The document is
	intended to provide a simplified guide for the deployment of LSVR extensions.
      </t>
    </abstract>

  </front>

  <middle>
    <section anchor="intro" title="Introduction">

      <t>
	This document complements <xref target="I-D.ietf-lsvr-bgp-spf"/> by
	discussing the applicability of the technology in a simple and fairly common
	deployment scenario, which is described in <xref target="usecases"/>.
      </t>

      <t>
	After describing the deployment scenario, <xref target="motivation"/>
	will describe the reasons for BGP modifications for such deployments.
      </t>

      <t>
	Once the control plane routing protocol requirements are described,
	<xref target="bgpspf"/> will cover the LSVR protocol enhancements
	to BGP to meet these requirements and their applicability to Data Center
        CLOS networks.
      </t>

    </section>


    <section title="Requirements Language">
     <t>The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL
      NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED",
      "MAY", and "OPTIONAL" in this document are to be interpreted as
      described in BCP 14 <xref target="RFC2119"/> <xref target="RFC8174"/> 
      when, and only when, they appear in all capitals, as shown here.</t>
    </section>


    <section anchor="reco" title="Recommended Reading">
      <t>
	This document assumes knowledge of existing data center networks and data center
	network topologies <xref target="CLOS"/>. This document also assumes
	knowledge of data center routing protocols like
	BGP <xref target="RFC4271"/>, BGP-SPF <xref target="I-D.ietf-lsvr-bgp-spf"/>,
	OSPF <xref target="RFC2328"/>, as well as,
	data center OAM protocols like LLDP <xref target="RFC4957"/> and
        BFD <xref target="RFC5580"/>.
      </t>


    </section>


    <section anchor="usecases" title="Common Deployment Scenario">
      <t>
	Within a Data Center, a common network design to interconnect servers is done using the CLOS
	topology <xref target="CLOS"/>. The CLOS topology is fully non-blocking and the topology is
	realized using Equal Cost Multipath (ECMP). In a CLOS topology, the minimum number
        of parallel paths between two servers is determined by the width of a tier-1 stage as shown in
        the figure 1.
      </t>

      <t>
	The following example illustrates multistage CLOS topology.
      </t>

      <figure align="center" anchor="fig1" title="Illustration of the basic CLOS">
	<artwork align="left"><![CDATA[

                                   Tier-1
                                  +-----+
                                  |NODE |
                               +->| 12  |--+
                               |  +-----+  |
                       Tier-2  |           |   Tier-2
                      +-----+  |  +-----+  |  +-----+
        +------------>|NODE |--+->|NODE |--+--|NODE |-------------+
        |       +-----|  9  |--+  | 10  |  +--| 11  |-----+       |
        |       |     +-----+     +-----+     +-----+     |       |
        |       |                                         |       |
        |       |     +-----+     +-----+     +-----+     |       |
        | +-----+---->|NODE |--+  |NODE |  +--|NODE |-----+-----+ |
        | |     | +---|  6  |--+->|  7  |--+--|  8  |---+ |     | |
        | |     | |   +-----+  |  +-----+  |  +-----+   | |     | |
        | |     | |            |           |            | |     | |
      +-----+ +-----+          |  +-----+  |          +-----+ +-----+
      |NODE | |NODE | Tier-3   +->|NODE |--+   Tier-3 |NODE | |NODE |
      |  1  | |  2  |             |  3  |             |  4  | |  5  |
      +-----+ +-----+             +-----+             +-----+ +-----+
        | |     | |                                     | |     | |
        A O     B O            <- Servers ->            Z O     O O

        ]]></artwork>
      </figure>

    </section>


    <section anchor="motivation" title="Justification for BGP SPF Extension">
      <t>
	In order to simplify layer-3 routing and operations <xref target="RFC7938"/>, many data centers
        use BGP as a routing protocol to create an overlay as well as an underlay network for their CLOS
        Topologies.
        However, BGP is a path-vector routing protocol. Since it
        does not create a fabric topology, it uses hop-by-hop EBGP peering to
        facilitate hop-by-hop routing to create the underlay network and to resolve any
        overlay next hops.
        The hop-by-hop BGP peering paradigm imposes several
        restrictions within a CLOS. It severely prohibits a deployment of Route Reflectors/Route
        Controllers as the EBGP sessions are congruent with the data path. The BGP best path algorithm is
	prefix-based and it prevents announcements of prefixes to other BGP speakers until the best
        path decision process is performed for the prefix at each intermediate hop.
        These restrictions significantly delay the overall convergence of the underlay
        network within a CLOS.
      </t>

      <t>
        The LSVR SPF modifications allow BGP to overcome these limitations. Furthermore, using the BGP-LS NLRI
        format <xref target="RFC7752"/> allows the LSVR data to be advertised for nodes, links, and prefixes in the
        BGP routing domain and used for SPF computations.
      </t>
    </section>

    <section anchor="bgpspf" title="LSVR Applicability to CLOS Networks">
      <t>
        With the BGP SPF extensions <xref target="I-D.ietf-lsvr-bgp-spf"/>, the BGP
        best path computation and route computation are replaced with OSPF-like
        algorithms <xref target="RFC2328"/> both to determine whether an BGP-LS NLRI
        has changed and needs to be re-advertised and to compute the routing table.
        These modifications will significantly improve convergence of the underlay
        while affording the operational benefits of a single routing protocol
        <xref target="RFC7938"/>.
      </t>
      <t>
        Data center controllers typically require visibility to the BGP topology to compute
        traffic-engineered paths. These controllers learn the topology and other relevant
        information via the BGP-LS address family <xref target="RFC7752"/> which is totally
        independent of the underlay address families (usually IPv4/IPv6 unicast). Furthermore,
        in traditional BGP underlays, all the BGP routers will need to advertise their BGP-LS
        information independently. With the BGP
        SPF extensions, controllers can learn the topology using the same BGP advertisements
        used to compute the underlay routes. Furthermore, these data center controllers can
        avail the convergence advantages of the BGP SPF extensions. The placement of
        controllers can be outside of the forwarding path or within the forwarding path.
      </t>
       <t>
	Alternatively, as each and every router in the BGP SPF domain will have a complete view
	of the topology, the operator can also choose to configure BGP sessions in hop-by-hop
	peering model described in <xref target="RFC7938"/> along with BFD <xref target="RFC5580"/>.
	In doing so, while the hop-by-hop peering model lacks inherent benefits of the
        controller-based model, BGP updates need not be serialized by BGP best path algorithm in 
        either of these models. This helps overall network convergence.
      </t>
	    
      <section anchor="lsvrsafi" title="Usage of BGP-LS SAFI">
      <t>
        The BGP SPF extensions <xref target="I-D.ietf-lsvr-bgp-spf"/> define a new BGP-LS SAFI
	for announcement of BGP SPF link-state. The NLRI format and its associated attributes follow
	the format of BGP-LS for node, link, and prefix announcements. Whether the peering
	model within a CLOS follows hop-by-hop peering described in <xref target="RFC7938"/> or any
	controller-based or route-reflector peering, an operator can exchange BGP SPF SAFI routes over
        the BGP peering	by simply configuring BGP SPF SAFI between the necessary BGP speakers.
      </t>
       <t>
	The BGP-LS SPF SAFI can also co-exist with BGP IP Unicast SAFI which could exchange overlapping IP routes. 
        The routes received by these SAFIs
	are evaluated, stored, and announced separately according to the rules of <xref target="RFC4760"/>. The
        tie-breaking of route installation is a matter of the local policies and preferences of the network
        operator.
       </t>
       <t>
	Finally, as the BGP SPF peering is done following the procedures described in <xref target="RFC4271"/>, all the
	existing transport security mechanisms including <xref target="RFC5925"/> are available for the BGP-LS SPF SAFI.
       </t>
       <section anchor="other-safi" title="Relationship to Other BGP AFI/SAFI Tuples">
       <t>Normally, the BGP-LS AFI/SAFI is used solely to compute the underlay and is given preference over other 
          AFI/SAFIs. Other BGP SAFIs, e.g., IPv6/IPv6 Unicast VPN would use the BGP-SPF computed routes for next hop 
          resolution. However, if BGP-LS NLRI is also being advertised for controller consumption, there is no need
          to replicate the Node, Link, and Prefix NLRI in BGP-NLRI. Rather, additional NLRI attributes can be advertised
          in the BGP-LS SPF AFI/SAFI as required.
       </t>
      </section>
    </section>
    <section anchor="peering" title="Peering Models">
      <t>
        As previously stated, BGP SPF can be deployed using the existing peering model 
        where there is a single hop BGP session on each and every link in the data center 
        fabric <xref target="RFC7938"/>. This provides for both the advertisement of routes
        and the determination of link and neighboring switch availability. With BGP SPF, the underlay will converge 
        faster due to changes in the decision process which will allow NLRI changes to be advertised
        faster after detecting a change.
       </t>
      <section anchor="sparse-peering" title="Sparse Peering Model">
         <t>
          Alternately, BFD <xref target="RFC5580"/> can be used to swiftly determine the availability of links and
          the BGP peering model can be significantly sparser than the data center fabric. BGP SPF sessions then only 
          be established with enough peers to provide a bi-connected graph. If IEBGP is used, then the BGP routers at
          tier N-1 will act as route-reflectors for the routers at tier N. 
         </t>
         <t>
          The obvious usage of sparse peering is to avoid parallel sessions between the same two BGP speakers in the
          data center fabric. However, this use case is not very useful since parallel layer-3 links between
          the same two BGP routers are rare in CLOS or Fat-Tree topologies. Two more interesting scenarios are described
          below. 
         </t>
         <t>
          In current Data Center topologies, there is often a very dense mesh of links between levels, e.g.,
          leaf and spine, providing 32-way, 64-way, or more Equal-Cost Multi-Path (ECMP) paths. In 
          these topologies, it is desirable not to have a BGP session on every link and techniques such as the one
          described below <xref target="bi-connected"/> can be used establish sessions on some subset of northbound
          links. 
         </t>
         <t>
          Alternately, controller-based data center topologies are envisioned where BGP speakers within the data
          center only establish BGP sessions with two or more controllers. In these topologies, fabric nodes below the
          first tier (using <xref target="RFC7938"/> hierarchy) will establish BGP multi-hop sessions with the 
          controllers. For the multi-hop sessions, determining the route to the controllers without depending on BGP
          would need to be through some other means beyond the scope of this document. However, the BGP discovery 
          mechanisms <xref target="bgp-discovery"/> would be one possibility.
         </t>
       </section>  
      <section anchor="bi-connected" title="Bi-Connected Graph Heuristic">
      <t>With this heuristic,  discovery of BGP peers is assumed <xref target="bgp-discovery"/>. Additionally, it 
      assumed that the direction of the peering can be ascertained. In the context of a data center fabric, direction
      is either northbound (toward the spine), southbound (toward the Top-Of-Rack (TOR) switches) or east-west (same
      level in hierarchy. The determination of the direction is beyond the scope of this document. However, it would 
      be reasonable to assume a technique where the TOR switches can be identified and the number of hops to the TOR 
      is used to determine the direction.</t>
      <t>In this heuristic, BGP speakers allow passive session establishment for southbound BGP sessions. For northbound
      sessions, BGP speakers will attempt to maintain two northbound BGP sessions with different switches (in data center
      fabrics there is normally a single layer-3 connection anyway). For east-west sessions, passive BGP session
      establishment is allowed. However, BGP speaker will never actively establish an east-west BGP session unless 
      it can't establish two northbound BGP sessions.</t>
      </section>
    </section>
    <section anchor="bgp-discovery" title="BGP Peer Discovery">
    <section anchor="bgp-discovery-req" title="BGP Peer Discovery Requirements">
    <t>The most basic requirement is to be able to discover the address of a single-hop peer without
    pre-configuration. This is being accomplished today with using IPv6 Router Advertisements (RA)
    <xref target="RFC4861"/> and assuming that a BGP sessions is desired with any discovered peer. 
    Beyond the basic requirement, it is useful to have to
    following information relating to the BGP session:
     <list style="symbols"> 
         <t>Autonomous System (AS) and BGP Identifier of a potential peer. The latter can be used for debugging
            and to decrease the likelihood of BGP session establishment collisions.</t>
         <t>Security capabilities supported and for cryptographic authentication, the security capabilities and
            possibly a key-chain <xref target="RFC8177"/> to be used.</t>
         <t>Session Policy Identifier - A group number or name used to associate common session parameters with
            the peer. For example, in a data center, BGP sessions with a Top of Rack (ToR) device could have
            parameters than BGP sessions between leaf and spine.</t>
     </list></t>
    <t>In a data center fabric, it is often useful to know whether a peer is southbound (towards the servers) or
       northbound (towards the spine or super-spine) <xref target="bi-connected"/>. A potential requirement
       would also be to determine this dynamically. One mechanism, without specifying all the details, might be
       for the ToRs to be identified when installed and for the others switches in the fabric to determine their
       level based on the distance from the closest ToR.</t> 
    <t>If there are multiple links between BGP speakers or the links between BGP speakers are unnumbered,
       it is also useful to be able to establish multi-hop sessions using the loopback addresses. This will often 
       require the discovery protocol to install route(s) toward the potential peer loopback addresses prior to 
       BGP session establishment.</t>
     <t>
      Finally, a simple BGP discovery protocol could also be used to establish a multi-hop session with one or
      more controllers by advertising connectivity to one or more controllers. However, once the multi-hop 
      session actually traverses multiple nodes, it is bordering a distance-vector routing protocol and
      possibly this is not a good requirement for the discovery protocol.
     </t>
    </section> 
    <section anchor="bgp-discovery-alt" title="BGP Peer Discovery Alternatives">
    <t>
    While BGP peer discovery is not part of  <xref target="I-D.ietf-lsvr-bgp-spf"/>, there are, at least,
    three proposals for BGP peer discovery. At least one of these mechanisms will be adopted and will
    be applicable to deployments other than the data center. It is strongly RECOMMENDED that the accepted 
    mechanism be used in conjunction with BGP SPF in data centers. The BGP discovery mechanism should discovery 
    both peer addresses and endpoints for BFD discovery. Additionally, it would be great if there were a 
    heuristic for determining whether the peer is at a tier above or below the discovering BGP speaker (refer to 
    <xref target="bi-connected"/>).
    </t>
    <t>
    The BGP discovery mechanisms under consideration are <xref target="I-D.acee-idr-lldp-peer-discovery"/>,
    <xref target="I-D.xu-idr-neighbor-autodiscovery"/>, and <xref target="I-D.ymbk-lsvr-lsoe"/>.
    </t>
    </section>
    <section anchor="dci" title="Data Center Interconnect (DCI) Applicability">
    <t>
    Since BGP SPF is to be used for the routing underlay and DCI gateway boxes typically have direct
    or very simple connectivity, BGP external sessions would typically not include the BGP SPF SAFI. 
    </t>
    </section> 
    </section>
    <section anchor="other-env" title="Non-CLOS/FAT Tree Topology Applicability">
    <t>
    The BGP SPF extensions <xref target="I-D.ietf-lsvr-bgp-spf"/> can be used in other topologies and avail
    the inherent convergence improvements. Additionally, sparse peering techniques may be utilized 
    <xref target="peering"/>. However, determining whether or to establish a BGP session is more complex and 
    the heuristic described in <xref target="bi-connected"/> cannot be used. In such topologies, other
    techniques such as those described in <xref target="I-D.li-lsr-dynamic-flooding"/> may be employed. 
    One potential deployment would be the underlay for a Service Provider (SP) backbone where usage of a 
    single protocol, i.e., BGP, is desired.
    </t>
    </section>
    </section>



    <section anchor="IANA" title="IANA Considerations">

      <t>
	No IANA updates are requested by this document.
      </t>

    </section>

    <section anchor="Security" title="Security Considerations">
      <t>
	This document introduces no new security considerations above
	and beyond those already specified in the <xref target="RFC4271"/> and
        <xref target="I-D.ietf-lsvr-bgp-spf"/>.
      </t>

    </section>

    <section anchor="Acknowledgements" title="Acknowledgements">
      <t>
       The authors would like to thank Alvaro Retana and Yan Filyurin for the review and comments.
      </t>
    </section>
  </middle>

  <back>

    <references title="Normative References">
      <?rfc include="reference.RFC.2119"?>
      <?rfc include="reference.RFC.8174"?>
      &I-D.draft-ietf-lsvr-bgp-spf;


   </references>

    <references title="Informative References">
      <?rfc include="reference.RFC.2328"?>
      <?rfc include="reference.RFC.4271"?>
      <?rfc include="reference.RFC.4760"?>
      <?rfc include="reference.RFC.4861"?>
      <?rfc include="reference.RFC.4957"?>
      <?rfc include="reference.RFC.5580"?>
      <?rfc include="reference.RFC.5925"?>
      <?rfc include="reference.RFC.7752"?>
      <?rfc include="reference.RFC.7938"?>
      <?rfc include="reference.RFC.8177"?>
      &I-D.draft-acee-idr-lldp-peer-discovery;
      &I-D.draft-xu-idr-neighbor-autodiscovery;
      &I-D.draft-ymbk-lsvr-lsoe;
      &I-D.draft-li-lsr-dynamic-flooding;
      <reference anchor="CLOS" target="">
        <front>
          <title>A Study of Non-Blocking Switching Networks</title>

	  <author initials="" surname="">
            <organization></organization>
          </author>
          <date month='March' year='1953' />
        </front>
        <seriesInfo name='' value='The Bell System Technical Journal, Vol. 32(2),
				   DOI 10.1002/j.1538-7305.1953.tb01433.x' />
      </reference>

    </references>
  </back>
</rfc>