wdiff draft-ietf-nvo3-use-case-00.txt draft-ietf-nvo3-use-case-01.txt

Network working group Working Group L. Yong
Internet Draft Huawei
Category: Informational M. Toy
Comcast
A. Isaac
Bloomberg
V. Manral
Hewlett-Packard
L. Dunbar
Huawei

Expires: August November 2013 February 15, May 1, 2013

Use Cases for DC Network Virtualization Overlays

draft-ietf-nvo3-use-case-00

draft-ietf-nvo3-use-case-01

Abstract

This draft document describes the general DC NVO3 use cases. The work intention
is to help validate the NVO3 framework cases that may be
potentially deployed in various data centers and requirements as along
with the development apply to different
applications. An application in a DC may be a combination of the solutions. some
use cases described here.

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 August, November, 2013.

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.

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
1.1. Contributors..............................................4
1.2. Terminology...............................................4
2. Terminology....................................................4
3. Basic Virtual Networks in a Data Center........................4
4. Center........................5
3. Interconnecting DC Virtual Network and External Networks.......6
4.1.
3.1. DC Virtual Network Access via Internet....................7
4.2. Internet....................6
3.2. DC Virtual Network VN and WAN VPN Interconnection............7
5. Enterprise Sites interconnected via SP WAN......7
4. DC Applications Using NVO3.....................................9
5.1. NVO3.....................................8
4.1. Supporting Multi Technologies and Applications in a Data Center...........10
5.2. DC....9
4.2. Tenant Virtual Network with Bridging/Routing.............10
5.3. Multi-Subnets or across multi DCs.....9
4.3. Virtual Data Center (VDC)................................11
5.4. Federating NV03 Domains..................................13
6. (vDC)................................11
5. OAM Considerations............................................13
7.
6. Summary.......................................................13
8.
7. Security Considerations.......................................14
9.
8. IANA Considerations...........................................14
9. Acknowledgements..............................................14
10. Acknowledgements.............................................15
11. References...................................................15
11.1. References...................................................14
10.1. Normative References....................................15
11.2. References....................................14
10.2. Informative References..................................16 References..................................15
Authors' Addresses...............................................16 Addresses...............................................15

1. Introduction

Compute

Server Virtualization has dramatically and quickly changed IT industry in terms of efficiency,
cost, and the speed in providing a new applications and/or services.
However the problems in today's data center networks hinder the
support of an elastic cloud service and dynamic virtual tenant
networks [NVO3PRBM]. The goal of DC Network Virtualization Overlays,
i.e. NVO3, is to decouple tenant system the communication networking among tenant systems
from DC physical networks and to allow one physical network
infrastructure to provide: 1) traffic isolation among tenant virtual
networks over the same physical network; 2) independent address
space in each virtual network and address isolation from the
infrastructure's; 3) Flexible VM placement and move from one server
to another without any of the physical network
limitation. limitations. These
characteristics will help address the issues that hinder true
virtualization in the data centers [NVO3PRBM].

Although NVO3 may enable enables a true virtual environment where VMs and
network service appliances communicate, virtualization environment, the NVO3
solution has to address the communication between a virtual network
and one a physical network. This is because 1) many traditional DCs exist and that need to
provide network virtualization are currently running over physical
networks, the migration will not
disappear any time soon; be in steps; 2) a lot of DC
applications serve are served to Internet and/or cooperation users which run directly on
physical networks; 3) some applications are CPU bound like Big Data
analytics which are CPU bound and may not
want need the virtualization capability.

This document is to describe general NVO3 use cases that apply to
various data center networks to ensure nvo3 framework and solutions
can meet the demands. centers. Three types of the use cases described here
are:

o A virtual network connects many tenant systems within a Data
Center and form one L2 or L3 communication domain. A virtual
network segregates its traffic from others and allows the VMs in
the network moving from one server to another. The case may be
used for DC internal applications that constitute the DC East-
West traffic.

o A DC provider offers a secure DC service to an enterprise
customer and/or Internet users. In these cases, the enterprise
customer may use a traditional VPN provided by a carrier or an
IPsec tunnel over Internet connecting to an overlay virtual a NVO3 network offered by within a Data Center provider.
provider DC. This is mainly constitutes DC North-South traffic.

o A DC provider uses may use NVO3 to and other network technologies for a
tenant network, construct different topologies or zones for a
tenant network, and may design a variety of cloud applications
that make use of may require the network service appliance, virtual compute,
storage, and networking. In this case, the NVO3 provides the
virtual
networking functions for the applications.

The document uses the architecture reference model and terminologies defined in
[NVO3FRWK] to describe the use cases.

1.1. Contributors

Vinay Bannai
PayPal
2211 N. First St,
San Jose, CA 95131
Phone: +1-408-967-7784
Email: vbannai@paypal.com

Ram Krishnan
Brocade Communications
San Jose, CA 95134
Phone: +1-408-406-7890
Email: ramk@brocade.com

1.2. Terminology

This document uses the terminologies defined in [NVO3FRWK],
[RFC4364]. Some additional terms used in the document are listed
here.

CUG: Closed User Group

L2 VNI: L2 Virtual Network Instance

L3 VNI: L3 Virtual Network Instance

ARP: Address Resolution Protocol

CPE: Customer Premise Equipment

DMZ: Demilitarized Zone

DNS: Domain Name Service

DMZ: DeMilitarized Zone

NAT: Network Address Translation

VNIF:

VIRB: Virtual Network Interconnection Interface

3. Integrated Routing/Bridging

Note that a virtual network in this document is a network
virtualization overlay instance.

2. Basic Virtual Networks in a Data Center

A virtual network may exist within a DC. The network enables a
communication among tenant systems Tenant Systems (TSs) that are in a Closed User
Group (CUG). A TS may be a physical server or virtual machine (VM)
on a server. The network virtual edge (NVE) may co-exist with Tenant
Systems, i.e. on an end-device, or exist on a different device, e.g.
a top of rack switch (ToR). A virtual network has a unique virtual
network identifier (may be local or global unique) for switches/routers an NVE to
properly differentiate it from other virtual networks.

The CUGs TSs attached to the same NVE are
formed not necessary in the same CUG,
i.e. in the same virtual network. The multiple CUGs can be
constructed in a way so that proper the policies can be applied are enforced when the TSs
in one CUG communicate with the TSs in other CUGs.

Figure 1 depicts this case by using An NVE provides
the framework model.[NVO3FRWK]
NVE1 and NVE2 are two network virtual edges and each may exist on reachbility for Tenant Systems in a
server or ToR. Each NVE CUG, and may be also have the member of one or more virtual
networks. Each virtual network may be L2 or L3 basis. In this
illustration, three virtual networks with VN context Ta, Tn, and Tm
are shown. The VN 'Ta' terminates on both NVE1 and NVE2; The VN 'Tn'
terminates on NVE1
policies and provide the VN 'Tm' at NVE2 only. If an NVE is a
member of reachbility for Tenant Systems in different
CUGs (See section 4.2). Furthermore in a VN, one or more virtual network instances (VNI) (i.e.
routing and forwarding table) exist on the NVE. Each NVE has one
overlay module to perform frame encapsulation/decapsulation and
tunneling initiation/termination. DC operators may construct
many tenant networks that have no communication at all. In this scenario, a tunnel between
NVE1 and NVE2 is necessary for the virtual
case, each tenant network Ta.

A TS attaches to a virtual may use its own address space. Note that
one tenant network (VN) via a virtual access point
(VAP) on an NVE. One TS may participate in contain one or more virtual
networks via VAPs; one NVE CUGs.

A Tenant System may also be configured with multiple VAPs for
a VN. Furthermore if individual addresses and
participate in multiple virtual networks networks, i.e. use different address spaces, the TS participating
in all of them will be
configured with multiple addresses as well. A TS as different virtual network. For examples, a gateway TS is an
example for this. In addition, multiple TSs may use one VAP to
attach to NAT GW; or a VN. For example, VMs are on TS
is a firewall server and NVE is on ToR,
then some VMs may attach to NVE for multiple CUGs.

Network Virtualization Overlay in this context means the virtual
networks over DC infrastructure network via a VLAN.

A VNI on an NVE is tunnel, i.e. a routing and forwarding table that caches and/or
maintains the mapping tunnel
between any pair of a NVEs. This architecture decouples tenant system
address schema from the infrastructure address space, which brings a
great flexibility for VM placement and its attached NVE. The
table entry may be updated by mobility. This also makes the control plane, data plane,
management plane, or
transit nodes in the combination infrastructure not aware of them.

+------- L3 Network ------+
| Tunnel Overlay |
+------------+--------+ +--------+-------------+
| +----------+------+ | | +------+----------+ |
| | Overlay Module | | | | Overlay Module | |
| +---+---------+---+ | | +--+----------+---+ |
| |Ta |Tn | | |Ta |Tm |
| +--+---+ +--+---+ | | +-+----+ +--+---+ |
| | VNIa |..| VNIn | | | | VNIa |..| VNIm | |
NVE1 | ++----++ ++----++ | | ++----++ ++----++ | NVE2
| |VAPs| |VAPs| | | |VAPs| |VAPs| |
+---+----+----+----+--+ +---+----+----+----+---+
| | | | | | | |
------+----+----+----+------ -----+----+----+----+-----
| .. | | .. | | .. | | .. |
| | | | | | | |
Tenant systems Tenant systems
Figure 1 NVO3 for Tenant System Networking the existence of
the virtual networks. One tunnel may carry the traffic belonging to
different virtual networks; a virtual network identifier is used for
traffic segregation in a tunnel.

A virtual network may have many be an L2 or L3 domain. An NVE members and interconnect
several thousands of TSs (as may be a matter of policy), the capability member
of
supporting a lot several virtual networks each of TSs per tenant instance and TS mobility which is
critical for NVO3 solution no matter where an in L2 or L3. A virtual
network may carry unicast traffic and/or broadcast/multicast/unknown
traffic from/to tenant systems. An NVE resides. may use p2p tunnels or a p2mp
tunnel to transport broadcast or multicast traffic, or may use other
mechanisms [NVO3MCAST].

It is worth to mention two distinct cases here. The first is when that TS
and NVE are co-located on a same physical end device, which means that the
NVE is can be made aware of the TS state at any time via internal API.

The second is when that TS and NVE are remotely connected, i.e. connected
via a switched network or point-to-point link. In this case, a
protocol is necessary for NVE to know TS state.

Note that if all NVEs are co-located with

One virtual network may have many NVE members each of which many TSs
may attach to. TS dynamic placement and mobility results in a CUG, the
communication frequent
changes in the CUG is in a true virtual environment. If a TS
connects to a and NVE remotely, the communication from this bindings. The TS reachbility update
mechanism MUST be fast enough to other
TSs in the CUG is not in a true virtual environment. cause any service interruption.
The packets
to/from this TS are directly carried over capability of supporting a physical network, i.e.
on the wire. This may require some necessary configuration on the
physical lot of TSs in a tenant network to facilitate the communication.

Individual virtual networks may use its own address space and the
space a
lot of tenant networks is isolated from DC infrastructure. This eliminates the route
reconfiguration in the DC underlying network when VMs move. Note
that the critical for NVO3 solutions still have to address VM move in the overlay
network, i.e. the TS/NVE association change when a VM moves. solution.

If a virtual network spans across multiple DC sites, one design is
to allow the corresponding NVO3 instance seamlessly span across
those sites without DC gateway routers' termination. In this case,
the tunnel between a pair of NVEs may in turn be tunneled over other
intermediate tunnels over the Internet or other WANs, or the intra
DC and inter DC tunnels are stitched together to form an end-to-end
tunnel between two NVEs in different
virtual network across DCs.

4. The latter is described in section 3.2.
Section 4.2 describes other options.

3. Interconnecting DC Virtual Network and External Networks

For customers (an enterprise or individuals) who want to utilize the
DC provider's compute and storage resources to run their
applications, they need to access their systems hosted in a DC
through Carrier WANs Internet or Internet. Service Providers' WANs. A DC provider may use
construct an NVO3
virtual network which all the resources designated for such a
customer connect to and allow the customer to access their systems; then it, systems
via the network. This, in turn, becomes the case of interconnecting
a DC virtual NVO3 network and external networks. networks via Internet or WANs. Two
cases are described here.

4.1.

3.1. DC Virtual Network Access via Internet

A user or an enterprise customer connects securely to a DC virtual
network via Internet but securely. Internet. Figure 2 1 illustrates this case. An L3 A virtual
network is configured on NVE1 and NVE2 and two NVEs are connected
via an L3 tunnel in the Data Center. A set of tenant systems attach are
attached to NVE1. NVE1 on a server. The NVE2 connects resides on a DC Gateway
device. NVE2 terminates the tunnel and uses the VNID on the packet
to one (may be more) TS
that runs pass the packet to the corresponding VN gateway and NAT applications (known as network
service appliance). GW entity on the DC GW. A
user or customer can access their systems systems, i.e. TS1 or TSn, in the
DC via Internet by using IPsec tunnel [RFC4301]. The encrypted IPsec tunnel is
established
between the VN GW and the user machine or CPE at enterprise edge location.
The VN GW provides IPsec functionality such as authentication scheme
and
encryption. encryption, as well as the mapping to the right virtual network
entity on the DC GW. Note that 1) some VN GW function functions such as
firewall and load balancer may also be performed by a locally attached
network
service appliance device or on a devices; 2) The virtual network in DC GW.

+--------------+ +----------+ may use
different address space than external users, then VN GW serves the
NAT function.

Server+---------------+
| TS1 TSn | +------+
| |...| | Firewall
| TS
+----+(OM)+L3 VNI+--+-+ NAT +-+---+-+ | External User
| | NVE1 | | +-----+
| (VN GW) +---+---+ | | +------+ PC | +----+-----+
+------+--------+ +--+--+
| *
L3 Tunnel +--------------+ ^ *
| NVE2 |IPsec Tunnel
+--------+---------+ *
DC GW +------+---------+ .--. .--.
| +------+-------+ +---+---+ | ( :' '* '.--.
| |Overlay Module| | NVE2 | | .-.' : * )
| +------+-------+ +---+---+ | ( * Internet )
| +-----+------+ +---+---+. | ( : * /
| | L3 VNI | | VNGW1 * * * * * * * * '-' : '-'
NVE1
| +-+--------+-+ +-------+ | \../+\.--/'
+----+--------+----+ | IPsec \../ \.--/'
| ... +--------+ | V
Tenant Systems User Access Tunnel
+----------------+

DC Provider Site

OM: Overlay Module;

Figure 2 1 DC Virtual Network Access via Internet

4.2.

3.2. DC Virtual Network VN and Enterprise Sites interconnected via SP WAN

An Enterprise company would lease some DC provider compute resources
to run some applications. For example, the company may run its web
applications at DC provider sites but run backend applications in
their own DCs. The Web applications and backend applications need to
communicate privately. DC provider may construct a NVO3 network to
connect all VMs running the Enterprise Web applications. The
enterprise company may buy a p2p private tunnel such as VPWS from a
SP to interconnect its site and the NVO3 network in provider DC site.
A protocol is necessary for exchanging the reachability between two
peering points and the traffic are carried over the tunnel. If an
enterprise has multiple sites, it may buy multiple p2p tunnels to
form a mesh interconnection among the sites and DC provider site.
This requires each site peering with all other sites for route
distribution.

Another way to achieve multi-site interconnection is to use Service
Provider (SP) VPN Interconnection services, in which each site only peers with SP PE
site. A DC Provider and Carrier VPN SP may build a VN NVO3 network (VN) and VPN independently
independently. The VN provides the networking for all the related
TSes within the provider DC. The VPN interconnects several
enterprise sites, i.e. VPN sites. The DC provider and
interconnect VPN SP further
connect the VN and VPN at the DC GW GW/ASBR and PE SP PE/ASBR. Several
options for an enterprise
customer. Figure 3 depicts this case in an L3 overlay (L2 overlay is the same). The DC provider constructs an L3 VN between interconnection of the NVE1 on a
server VN and the NVE2 on the DC GW VPN are described in the
RFC4364 [RFC4364]. In Option A with VRF-LITE [VRF-LITE], both DC site; GW
and SP PE maintain the carrier
constructs an L3VPN between PE1 routing/forwarding table, and PE2 in its IP/MPLS network. An
Ethernet Interface physically connects perform the
table lookup in forwarding. In Option B, DC GW and PE2 devices.
The local VLAN over SP PE do not
maintain the Ethernet interface [VRF-LITE] is configured
to connect forwarding table, it only maintains the L3VNI/NVE2 VN and VRF, which makes VPN
identifier mapping, and exchange the interconnection
between identifier on the L3 VN packet in the
forwarding process. In option C, DC GW and SP PE use the same
identifier for VN and VPN, and just perform the tunnel stitching,
i.e. change the L3VPN tunnel end points. Each option has pros/cons (see
RFC4364) and has been deployed in IP/MPLS network. An
Ethernet Interface SP networks depending on the
applications. The BGP protocols may be used between PE1 and CE to connect in these options for
route distribution. Note that if the
L3VPN provider DC is the SP Data
Center, the DC GW and enterprise physical networks. PE in this case may be on one device.

This configuration allows the enterprise networks communicating to
the tenant systems attached to the L3 VN in a provider DC without
interfering with DC provider underlying physical networks and other overlay
virtual networks in the DC. The enterprise may use its own address
space on the tenant systems attached to the L3 VN. The DC provider can
manage the VMs and storage attached attachment to the L3 VN for the enterprise
customer. The enterprise customer can determine and run their
applications on the VMs. From the L3 VN perspective, an end point in the enterprise
location appears as the end point associating to the NVE2. The NVE2
on the DC GW has to perform both the GRE tunnel termination [RFC4797]
and the local VLAN termination and forward the packets in between.
The DC provider and Carrier negotiate the local VLAN ID used on the
Ethernet interface.

This configuration makes the L3VPN over the WANs only has the
reachbility to the TS in the L3 VN. It does not have the
reachability of DC physical networks and other VNs in the DC.
However, the L3VPN has the reachbility of enterprise networks. Note
that both the DC provider and enterprise may have multiple network
locations connecting to the L3VPN.

The eBGP protocol can be used between DC GW and PE2 for the route
population in between. In fact, this is like the Option A in
[RFC4364]. This configuration can work with any NVO3 solution. The
eBGP, OSPF, or other can be used between PE1 and CE See section 4 for the route
population.

+-----------------+ +-------------+
| +----------+ | | +-------+ |
NVE2 | | L3 VNI +---+===========+-+ VRF | |
| +----+-----+ | VLAN | +---+---+ | PE2
| | | | | |
| +-----+-------+ | /+-----+-------+--\
| |Overly Module| | ( : '
| +-------------+ | { : }
+--------+--------+ { : LSP Tunnel }
| ; : ;
|IP Tunnel { IP/MPLS Network }
| \ : /
+--------+---------+ +----+------+ -
| +------+-------+ | | +--+---+ | '
| |Overlay Module| | | | VRF | |
| +------+-------+ | | +--+---+ | PE1
| |Ta | | | |
| +-----+------+ | +----+------+
| | L3 VNI | | |
NVE1 | +-+--------+-+ | |
| | VAPs | | CE Site
+----+--------+----+
| ... | Enterprise Site
Tenant systems

DC Provider Site

Figure 3 L3 VNI and L3VPN interconnection across multi networks

If an enterprise only has one location, it may use P2P VPWS [RFC4664]
or L2TP [RFC5641] to connect one DC provider site. In this case, one
edge connects to a physical network and another edge connects to an
overlay network.

Various alternatives can be configured between DC GW and SP PE to
achieve the same capability. Option B, C, or D in RFC4364 [RFC4364]
can be used and the characteristics of each option are described
there. more.

The interesting feature in this use case is that the L3 VN and compute
resource are managed by the DC provider. The DC operator can place
them at any location without notifying the enterprise and
carrier WAN SP
because the DC physical network is completely isolated from the
carrier and enterprise network. Furthermore, the DC operator may
move the VMs assigned to the enterprise from one sever to another in
the DC without the enterprise customer awareness, i.e. no impact on
the enterprise 'live' applications running these resources. Such
advanced feature brings features bring DC providers great benefits in serving these
kinds of applications but also add some requirements for NVO3
[NVO3PRBM].

4. DC Applications Using NVO3

NVO3 brings DC operators the flexibility to design in designing and deploying
different applications in a true virtual environment (or nearly true) without
worrying an end-to-end virtualization environment,
where the operators not need worry about the constraints of the
physical network configuration in the Data Center. DC
operators provider may build several virtual networks and interconnect them
directly to form a tenant virtual network
use NVO3 in various ways and implement also use it in the
communication rules, i.e. policy between different virtual networks;
or may allocate some VMs to run tenant applications and some to run
network service application such as Firewall and DNS conjunction with
physical networks in DC for the tenant.
Several many reasons. This section highlights
some use cases are given in this section.

5.1. but not limits to.

4.1. Supporting Multi Technologies and Applications in a Data Center DC

Most likely servers deployed in a large data center are rolled in at
different times and may have different capacities/features. Some
servers may be virtualized, some may not; some may be equipped with
virtual switches, some may not. For the ones equipped with
hypervisor based virtual switches, some may support VxLAN [VXLAN]
encapsulation, some may support NVGRE encapsulation [NVGRE], and
some may not support any types of encapsulation. To construct a
tenant virtual network among these servers and the ToRs, ToR switches, it
may use construct one virtual network overlay and one virtual network
w/o overlay, or two virtual networks and a gateway to allow overlay with different
implementations working together.
implementations. For example, one virtual network overlay uses VxLAN
encapsulation and another virtual network w/o overlay uses
traditional VLAN. VLAN or another virtual network overlay uses NVGRE.

The gateway entity, either on VMs device or standalone one, virtual gateway on a device may be used. The
gateway participates in to both virtual networks, networks. It performs the
packet encapsulation/decapsulation and may also perform address
mapping or translation, and etc.

A data center may be also constructed with multi-tier zones. Each
zone has different access permissions and maps run different applications.
For example, the services three-tier zone design has a front zone (Web tier)
with Web applications, a mid zone (application tier) with service
applications such as payment and identifiers booking, and
changes a back zone (database
tier) with Data. External users are only able to communicate with
the packet encapsulations.

5.2. web application in the front zone. In this case, the
communication between the zones MUST pass through the security
GW/firewall. The network virtualization may be used in each zone. If
individual zones use the different implementations, the GW needs to
support these implementations as well.

4.2. Tenant Virtual Network with Bridging/Routing Multi-Subnets or across multi DCs

A tenant network may contain multiple subnets. DC operators may
construct multiple tenant networks. The access policy for inter-
subnets is often necessary. To benefit the policy management, the
policies may be placed at some designated gateway devices only. Such
design requires the inter-subnet traffic MUST be sent to one of the
gateways first for the policy checking. However this may cause
traffic hairpin on the gateway in a DC. It is desirable that an NVE
can hold some policy and be able to forward inter-subnet traffic
directly. To reduce NVE burden, the hybrid design may be deployed,
i.e. an NVE can perform forwarding for the selected inter-subnets
and the designated GW performs for the rest. For example, each NVE
performs inter-subnet forwarding for a tenant, and the designated GW
is used for inter-subnet traffic from/to the different tenant
networks.

A tenant virtual network may span across multiple Data Centers. Centers in distance.
DC
operator operators may want to use an L2VN within a each DC and L3VN outside between DCs
for a tenant network. This is very similar to today's DC physical network
configuration. L2 bridging has the simplicity and endpoint
awareness while L3 routing has advantages in policy based routing,
aggregation, and scalability. For this configuration, the virtual
L2/L3 gateway
function is necessary to interconnect L2VN and L3VN in each DC. can be implemented on DC GW device. Figure 4 2
illustrates this configuration.

Figure 4 2 depicts two DC sites. The site A constructs an L2VN that
terminates on with
NVE1, NVE2, and GW1. NVE3. NVE1 and NVE2 reside on the servers where the
tenant systems are created. NVE3 resides on the DC GW device. The
site Z has similar configuration with NVE3 and NVE4 on the servers
and NVE6 on the DC GW. An L3VN is configured between the
GW1 NVE5 at
site A and the GW2 NVE6 at site Z. An internal Virtual Network
Interconnection Interface (VNIF) connects to Integrated
Routing and Bridging (VIRB) is used between L2VNI and L3VNI on GW1.
Thus the GW1 is the members of the L2VN NVE5
and L3VN. NVE6. The L2VNI is the MAC/NVE mapping table and the L3VNI is
the IP prefix/NVE mapping table.
Note that a VNI also has the mapping of TS and VAP at the local NVE.
The site Z has the similar configuration. A packet coming to the GW1 NVE5 from L2VN will
be descapulated decapsulated and converted into an IP packet and then
encapsulated and sent to the site Z. The Gateway uses ARP
protocol to obtain MAC/IP address mapping.

Note that both the L2VN L2VNs and L3VN in the figure Figure 2 are encapsulated and
carried by the
tunnels supported by the underlying networks which are not shown in
the figure.

+------------+ +-----------+
GW1| over within DC and across WAN networks, respectively.

NVE5/DCGW+------------+ +-----------+NVE6/DCGW
| +-----+ | '''''''''''''''' | +-----+ |GW2 |
| |L3VNI+----+' L3VN '+---+L3VNI| |
| +--+--+ | '''''''''''''''' | +--+--+ |
| |VNIF |VIRB | | VNIF| VIRB| |
| +--+--+ +--+---+ | | +--+--+ +---+--+ |
| |L2VNI| |L2VNIs| | | |L2VNI| |L2VNIs| |
| +--+--+ +--+---+ | | +--+--+ +---+--+ |
+----+-------+ +------+----+
''''|'''''''''' ''''''|'''''''
' L2VN ' ' L2VN '
NVE1
NVE1/S ''/'''''''''\'' NVE2 NVE3 '''/'''''''\'' NVE4 NVE2/S NVE3/S'''/'''''''\'' NVE4/S
+-----+---+ +----+----+ +------+--+ +----+----+
| +--+--+ | | +--+--+ | | +---+-+ | | +--+--+ |
| |L2VNI| | | |L2VNI| | | |L2VNI| | | |L2VNI| |
| ++---++ | | ++---++ | | ++---++ | | ++---++ |
+--+---+--+ +--+---+--+ +--+---+--+ +--+---+--+
|...| |...| |...| |...|

Tenant Systems Tenant Systems

DC Site A DC Site Z

Figure 4 2 Tenant Virtual Network with Bridging/Routing

5.3.

4.3. Virtual Data Center (VDC) (vDC)

Enterprise DC's today may often use several routers, switches, and
service
network appliance devices to construct its internal network, DMZ,
and external network access. A DC Provider may offer a virtual DC
service to an enterprise customer to and run enterprise applications
such as
website/emails. website/emails as well. Instead of using many hardware devices,
devices to do it, with the
overlay and network virtualization technology of NVO3, overlay
technology, DC operators can may build them such vDCs on top of a common
network infrastructure for many such customers and run network
service applications per customer a vDC basis. The net service applications may include
such as firewall, gateway, DNS, load
balancer, NAT, etc. balancer can be designed per vDC. The
network virtualization overlay further enables potential for vDC
mobility when customer moves to different locations because tenant
systems and net appliances configuration can be completely decouple
from the infrastructure network.

Figure 5 3 below illustrates this one scenario. For the simple
illustration, it only shows the L3VN or L2VN as virtual and overlay
routers or switches. In this case, DC operators construct several L2
VNs (L2VNx, L2VNy, L2VNz L2VNz) in Figure 5) 3 to group the end tenant
systems together per application basis, create an L3VNa for the
internal routing. A server or net device (may be a VM or server) runs
firewall/gateway applications and connects to the L3VNa and
Internet. A VPN load Balancer (LB) is used in L2VNx. A VPWS p2p tunnel
is also built between the gateway and enterprise router. The design
runs Enterprise Web/Mail/Voice applications at the provider DC site;
lets the users at Enterprise site to access the applications via the
VPN tunnel and Internet via a gateway at the Enterprise site; let
Internet users access the applications via the gateway in the
provider DC.

The enterprise operators can also use the VPN tunnel or
IPsec over Internet to access the vDC for the management purpose.
The firewall/gateway provides application-level and packet-level
gateway function and/or NAT function.

The Enterprise customer decides which applications are accessed by
intranet only and which by both intranet and extranet; DC operators
then design and configure the proper security policy and gateway
function. Furthermore DC operators may further use multi-zones in a vDC for
the security and/or set different QoS levels for the different
applications for a customer. based on customer applications.

This application use case requires the NVO3 solution to provide the DC operator
an easy way to create NVEs a VN and VNIs NVEs for any design and to quickly
assign TSs to a VNI, VNI on a NVE they attach to, easily place to set up
virtual topology and place or configure policies on an NVE, NVE or VMs
that run net services, and support VM mobility. Furthermore, DC
operator needs to view the tenant network topology and know the
tenant node capability and is able to configure a net service on the
tenant node. DC provider may further let a tenant to manage the vDC
itself.

Internet ^ Internet
|
^ +-+----+
| | GW |
| +--+---+
| |
+-------+--------+ +-+----+
|FireWall/Gateway+---VPN Tunnel---+Router|
|FireWall/Gateway+--- VPWS/MPLS---+Router|
+-------+--------+ +-+--+-+
| | |
...+... |..|
+-----: L3VNa :--------+ LANs
|
+-+-+ ....... |
|LB | | | Enterprise Site
+-+-+ | |
...+... ...+... ...+...
: L2VNx : : L2VNy : : L2VNz :
....... ....... .......
|..| |..| |..|
| | | | | |
Web Apps Mail Apps VoIP Apps

Provider DC Site

firewall/gateway and Load Balancer (LB) may run on a server or VMs

Figure 5 3 Virtual Data Center by Using NVO3

5.4. Federating NV03 Domains

Two general cases are 1) Federating AS managed by a single operator;
2) Federating AS managed by different Operators. The detail will be
described in next version.

5. OAM Considerations

NVO3 brings the ability for a DC provider to segregate tenant
traffic. A DC provider needs to manage and maintain NVO3 instances.
Similarly, the tenant needs to be informed about tunnel underlying network
failures impacting tenant applications. applications or the tenant network is able
to detect both overlay and underlay network failures and builds some
resiliency mechanisms.

Various OAM and SOAM tools and procedures are defined in [IEEE
802.1ag], [ITU-T Y.1731], [RFC4378], [RFC5880], [ITU-T Y.1564] for
L2 and L3 networks, and for user, including continuity check,
loopback, link trace, testing, alarms such as AIS/RDI, and on-demand
and periodic measurements. These procedures may apply to tenant
overlay networks and tenants not only for proactive maintenance, but
also to ensure support of Service Level Agreements (SLAs).

As the tunnel traverses different networks, OAM messages need to be
translated at the edge of each network to ensure end-to-end OAM.

It is important that failures at lower layers which do not affect
NVo3 instance are to be suppressed.

6. Summary

The document describes some basic general potential use cases of NVO3. NVO3 in
DCs. The combination of these cases should give operators
flexibility and capability to design more sophisticated cases for
various purposes.

The key requirements for NVO3 are 1) traffic segregation; 2)
supporting a large scale number of virtual networks in a common
infrastructure; 3) supporting highly distributed virtual network
with sparse memberships 3) VM mobility 4) auto or easy to construct
a NVE and its associated TS; 5) Security 6) NVO3 Management
[NVO3PRBM].

Difference between other overlay network technologies and NVO3 is
that the client edges of the NVO3 network are individual and
virtualized hosts, not network sites or LANs. NVO3 enables these
virtual hosts communicating in a true virtual environment without
constraints in physical networks.

NVO3 allows individual tenant virtual networks to use their own
address space and isolates the space from the network infrastructure.
The approach not only segregates the traffic from multi tenants on a
common infrastructure but also makes VM placement and move easier.

DC services may vary from infrastructure as a service (IaaS),
platform as a service (PaaS), to software as a service (SaaS), in
which the network virtual virtualization overlay is just a portion of an
application service. NVO3 decouples the services service
construction/configurations from the DC network infrastructure configuration.
configuration, and helps deployment of higher level services over
the application.

NVO3's underlying network provides the tunneling between NVEs so
that two NVEs appear as one hop to each other. Many tunneling
technologies can serve this function. The tunneling may in turn be
tunneled over other intermediate tunnels over the Internet or other
WANs. It is also possible that intra DC and inter DC tunnels are
stitched together to form an end-to-end tunnel between two NVEs.

A DC virtual network may be accessed via an external network in a
secure way. Many existing technologies can help achieve this.

NVO3 implementation may vary. Some DC operators prefer to use
centralized controller to manage tenant system reachbility in a
tenant network, other prefer to use distributed protocols to
advertise the tenant system location, i.e. attached NVEs. For the
migration and special requirement, the different solutions may apply
to one tenant network in a DC. When a tenant network spans across
multiple DCs and WANs, each network administration domain may use
different methods to distribute the tenant system locations. Both
control plane and data plane interworking are necessary.

7. Security Considerations

Security is a concern. DC operators need to provide a tenant a
secured virtual network, which means one tenant's traffic isolated
from the other tenant's traffic and non-tenant's traffic; they also
need to prevent DC underlying network from any tenant application
attacking through the tenant virtual network or one tenant
application attacking another tenant application via DC networks.
For example, a tenant application attempts to generate a large
volume of traffic to overload DC underlying network. The NVO3
solution has to address these issues.

8. IANA Considerations

This document does not request any action from IANA.

10.

9. Acknowledgements

Authors like to thank Sue Hares, Young Lee, David Black, Pedro
Marques, Mike McBride, David McDysan, Randy Bush, and Uma Chunduri
for the review, comments, and suggestions.

11.

10. References

11.1.

10.1. Normative References

[RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
Requirement Levels", BCP 14, RFC 2119, March 1997

[RFC4364] Rosen, E. and Y. Rekhter, "BGP/MPLS IP Virtual Private
Networks (VPNs)", RFC 4364, February 2006.

[IEEE 802.1ag] "Virtual Bridged Local Area Networks - Amendment 5:
Connectivity Fault Management", December 2007.

[ITU-T G.8013/Y.1731] OAM Functions and Mechanisms for Ethernet
based Networks, 2011.

[ITU-T Y.1564] "Ethernet service activation test methodology", 2011.

[RFC4378] Allan, D., Nadeau, T., "A Framework for Multi-Protocol
Label Switching (MPLS) Operations and Management (OAM)",
RFC4378, February 2006

[RFC4301] Kent, S., "Security Architecture for the Internet
Protocol", rfc4301, December 2005

[RFC4664] Andersson, L., "Framework for Layer 2 Virtual Private
Networks (L2VPNs)", rfc4664, September 2006

[RFC4797] Rekhter, Y., et al, "Use of Provider Edge to Provider Edge
(PE-PE) Generic Routing Encapsulation (GRE) or IP in
BGP/MPLS IP Virtual Private Networks", RFC4797, January
2007

[RFC5641] McGill, N., "Layer 2 Tunneling Protocol Version 3 (L2TPv3)
Extended Circuit Status Values", rfc5641, April 2009.

[RFC5880] Katz, D. and Ward, D., "Bidirectional Forwarding Detection
(BFD)", rfc5880, June 2010.

11.2.

10.2. Informative References

[NVGRE] Sridharan, M., "NVGRE: Network Virtualization using Generic
Routing Encapsulation", draft-sridharan-virtualization-
nvgre-01,
nvgre-02, work in progress.

[NVO3PRBM] Narten, T., etc "Problem Statement: Overlays for Network
Virtualization", draft-ietf-nvo3-overlay-problem-
statement-02, work in progress.

[NVO3FRWK] Lasserre, M., Motin, T., and etc, "Framework for DC
Network Virtualization", draft-ietf-nvo3-framework-02,
work in progress.

[NVO3MCAST] Ghanwani, A., "Multicast Issues in Networks Using NVO3",
draft-ghanwani-nvo3-mcast-issues-00, work in progress.

[VRF-LITE] Cisco, "Configuring VRF-lite", http://www.cisco.com

[VXLAN] Mahalingam,M., Dutt, D., etc "VXLAN: A Framework for
Overlaying Virtualized Layer 2 Networks over Layer 3
Networks", draft-mahalingam-dutt-dcops-vxlan-02.txt, draft-mahalingam-dutt-dcops-vxlan-03.txt, work
in progress.

Authors' Addresses

Lucy Yong
Huawei Technologies,
4320
5340 Legacy Dr.
Plano, Tx75025 US TX 75025

Phone: +1-469-277-5837
Email: lucy.yong@huawei.com

Mehmet Toy
Comcast
1800 Bishops Gate Blvd.,
Mount Laurel, NJ 08054

Phone : +1-856-792-2801
E-mail : mehmet_toy@cable.comcast.com

Aldrin Isaac
Bloomberg
E-mail: aldrin.isaac@gmail.com

Vishwas Manral
Hewlett-Packard Corp.
191111 Pruneridge Ave.

Cupertino,
3000 Hanover Street, Building 20C
Palo Alto, CA 95014

Phone: 408-447-1497 650-857-5501
Email: vishwas.manral@hp.com

Linda Dunbar
Huawei Technologies,
4320
5340 Legacy Dr.
Plano, Tx75025 TX 75025 US

Phone: +1-469-277-5840
Email: linda.dunbar@huawei.com