Network Virtualization using Extreme Fabric Connect

Table Of Contents
Network Virtualization Using Extreme Fabric Connect
© 2019 Extreme Networks, Inc. All rights reserved. 51
of IPVPNs or L3 VSNs, a VRF is thus a repository of all the IP routes that belong to a given virtualized
routing domain.
A VRF consists of an IP routing table, a forwarding table, and interfaces assigned to it. Common routing
protocols such as OSPF, RIP, BGP, and IS-IS can be used to advertise and learn routes within the VRFs. In a
scalable virtualized multi-tenant architecture one single backbone routing protocol is used to advertise and
learn routes across all VRFs within their respective virtual domain (IPVPN or L3 VSN).
Tip
In an SPB fabric this role is performed by IS-IS using I-SID-IPv4 (or I-SID-IPv6) routes,
which then associate those IP routes with the respective L3 VSN I-SID.
In an MPLS backbone, this role is performed by MP-BGP using VPN-IPv4 (or VPN-IPv6)
routes over iBGP peerings which have to render those IP routes unique to BGP via addition
of the Route Distinguisher (RD) and specific to the VPN-id via addition of Export Route
Targets (RT).
A proper implementation of VRFs must be able to support overlapping IP routes across different VRFs/L3
VSNs, which can be essential when integrating networking infrastructure from different entities following
mergers and acquisitions.
Note
A given router will support a maximum number of VRFs and a maximum number of IP
routes which can be held across all VRFs as well as in the non-virtualized GRT.
Tip
In the Extreme Networks implementation of Fabric Connect using SPB, a VRF can be
software limited to a defined maximum number of IP routes such that it cannot consume
all available record entries at the expense of other VRFs operating on the same router.
Global Router Table
The Global Router Table (GRT) constitutes the non-virtualized IP routing instance that all routers operate in
by default. It is often also referred to as VRF-0, to differentiate it from VRFs that are numbered starting
from VRF-1.
In an SPB fabric there are no underlying dependencies on the GRT, and the IGP used within it, and it is
conceptually treated just like any other VRF. IP routing is seen as a service provided by SPB, whether it is
performed inside the GRT or a VRF.
This contrasts with MPLS, where the GRT and its IGP take on a foundational role without which neither
MPLS’s label distribution protocols (LDP or RSVP) nor BGP (with its full mesh of iBGP peerings) would be
able to operate. While this is never an issue in carrier networks (where the GRT and its IGP is restricted to
the MPLS backbone alone), it becomes less ideal in Enterprise networks where some or most user traffic
naturally resides in the Global Routing Table.
There are however, even with SPB, some important properties that distinguish the GRT from VRFs. In the
Extreme Networks SPB implementation, the GRT can equally be extended above the Ethernet fabric (as a
service); however, this is done via explicit configuration rather than via assigning an I-SID and the resulting
domain is referred to as GRT IP Shortcut Routing (rather than as an L3 VSN).
Also in the Extreme Networks product implementation, all in-band management traffic (Telnet, SSH, SNMP,
etc.) is only processed if received on an IP interface belonging to the GRT. So in an SPB fabric deployment,
the IP Shortcuts are always enabled for management purposes.