wdiff draft-ietf-rtgwg-net2cloud-problem-statement-05.txt draft-ietf-rtgwg-net2cloud-problem-statement-06.txt

Network Working Group L. Dunbar
Internet Draft Futurewei
Intended status: Informational Andy Malis
Expires: March August 5, 2020 Independent
C. Jacquenet
Orange
M. Toy
Verizon
November 1, 2019
February 5, 2020

Dynamic Networks to Hybrid Cloud DCs Problem Statement
draft-ietf-rtgwg-net2cloud-problem-statement-05
draft-ietf-rtgwg-net2cloud-problem-statement-06

Abstract

This document describes the problems that enterprises face today
when interconnecting their branch offices with dynamic workloads in
third party data centers (a.k.a. Cloud DCs). There can be many
problems associated with network connecting to or among Clouds, many
of which probably are out of the IETF scope. The objective of this
document is to identify some of the problems that need additional
work in IETF Routing area. Other problems are out of the scope of
this document.

It examines some of the approaches interconnecting cloud DCs with
enterprises' on-premises DCs & branch offices. This document also
describes some of the network problems that many enterprises face
when they have workloads & applications & data split among different
data centers, especially for those enterprises with multiple sites
that are already interconnected by VPNs (e.g., MPLS L2VPN/L3VPN).

Current operational problems are examined to determine whether there
is a need to improve existing protocols or whether a new protocol is
necessary to solve them.

Status of this Memo

This Internet-Draft is submitted 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 April 1, 2009. August 5, 2020.

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

Table of Contents

1. Introduction...................................................3
1.1. On the evolution Key Characteristics of Cloud DC connectivity.................3 Services:....................3
1.2. Connecting to Cloud Services..............................3
1.3. The role of SD-WAN techniques in connecting to Cloud DC connectivity....4 Services........4
2. Definition of terms............................................4 terms............................................5
3. Interconnecting Enterprise Sites with Cloud DCs................5
3.1. Multiple connections High Level Issues of Connecting to workloads in a Cloud DC...........6 Multi-Cloud.................6
3.1. Security Issues...........................................6
3.2. Interconnect Private Authorization and Public Cloud DCs.................7 Identity Management.....................6
3.3. Desired Properties API abstraction...........................................7
3.4. DNS for Networks that interconnect Hybrid
Clouds.........................................................8 Cloud Resources...................................8
3.5. NAT for Cloud Services....................................8
3.6. Cloud Discovery...........................................9
4. Multiple Clouds Interconnection................................9 Interconnecting Enterprise Sites with Cloud DCs................9
4.1. Multi-Cloud Interconnection...............................9 Sites to Cloud DC........................................10
4.2. Desired Properties for Multi-Cloud Interconnection.......11 Inter-Cloud Interconnection..............................12
5. Problems with MPLS-based VPNs extending to Hybrid Cloud DCs...11 DCs...13
6. Problem with using IPsec tunnels to Cloud DCs.................13 DCs.................15
6.1. Complexity of multi-point any-to-any interconnection.....13 Scaling Issues with IPsec Tunnels........................15
6.2. Poor performance over long distance......................14
6.3. Scaling Issues with IPsec Tunnels........................14 distance......................15
7. Problems of Using SD-WAN to connect to Cloud DCs..............15 DCs..............16
7.1. SD-WAN among branch offices vs. interconnect More Complexity to Cloud DCs15 Edge Nodes............................16
7.2. Edge WAN Port Management.................................17
7.3. Forwarding based on Application..........................17
8. End-to-End Security Concerns for Data Flows...................18 Flows...................17
9. Requirements for Dynamic Cloud Data Center VPNs...............18 VPNs...............17
10. Security Considerations......................................19 Considerations......................................18
11. IANA Considerations..........................................19 Considerations..........................................18
12. References...................................................19 References...................................................18
12.1. Normative References....................................19 References....................................18
12.2. Informative References..................................19
13. Acknowledgments..............................................20 Acknowledgments..............................................19

1. Introduction

1.1. On the evolution Key Characteristics of Cloud DC connectivity

The ever-increasing use Services:

Key characteristics of cloud applications for communication
services change the way corporate business works Cloud Services are on-demand, scalable,
highly available, and shares
information. Such cloud applications use resources hosted in usage-based billing. Cloud Services, such as,
compute, storage, network functions (most likely virtual), third
party DCs that also host managed applications, etc. are usually hosted and managed by third parties Cloud Operators. Here are some examples of Cloud network
functions: Virtual Firewall services, Virtual private network
services, Virtual PBX services for other including voice and video
conferencing systems, etc. Cloud Data Center (DC) is shared
infrastructure that hosts the Cloud Services to many customers.

1.2. Connecting to Cloud Services

With the advent of widely available third-party cloud DCs and
services in diverse geographic locations and the advancement of
tools for monitoring and predicting application behaviors, it is technically feasible
very attractive for enterprises to instantiate applications and
workloads in locations that are geographically closest to their end-users. end-
users. Such proximity
improves can improve end-to-end latency and overall
user experience. Conversely, an enterprise can easily shutdown
applications and workloads whenever end-users are in motion (thereby
modifying the networking connection of subsequently relocated
applications and workloads). In addition, an enterprise enterprises may wish to
take advantage of more and more business applications offered by third party private
cloud DCs.

Most of those enterprise branch offices & on-premises data centers
are already connected via VPNs, such as MPLS-based L2VPNs and
L3VPNs. Then connecting to the cloud-hosted resources may not be
straightforward if the provider of operators.

The networks that interconnect hybrid cloud DCs must address the VPN service does not have
direct connections
following requirements:
- High availability to access all workloads in the corresponding desired cloud
DCs. Under those
circumstances, the enterprise can upgrade the CPEs deployed in its
various premises to utilize SD-WAN techniques to reach
Many enterprises include cloud
resources (without any assistance from the VPN service provider), or
wait for in their VPN service provider to make new agreements with data
center providers to connect to disaster recovery
strategy, such as enforcing periodic backup policies within the cloud resources. Either way has
additional infrastructure and operational costs.

In addition, more enterprises are moving towards hybrid cloud DCs,
i.e. owned
cloud, or operated by running backup applications in the Cloud.

- Global reachability from different Cloud operators, to maximize geographical zones, thereby
facilitating the
benefits proximity of geographical proximity, elasticity applications as a function of the
end users' location, to improve latency.
- Elasticity: prompt connection to newly instantiated
applications at Cloud DCs when usages increase and special features
offered by different cloud DCs.

1.2. prompt
release of connection after applications at locations being
removed when demands change.
- Scalable security management.

1.3. The role of SD-WAN techniques in Cloud DC connectivity

This document discusses the issues associated with connecting
enterprise's workloads/applications instantiated in multiple third-
party data centers (a.k.a. Cloud DCs) and its on-prem data centers.
Very often, the actual to Cloud DCs that host Services

Some of the
workloads/applications can be transient.

SD-WAN, initially launched to maximize bandwidths between locations
by aggregating multiple paths managed by different service
providers, has expanded to include flexible, on-demand, application- characteristics of SD-WAN [SDWAN-BGP-USAGE], such as
network augmentation and forwarding based connections established over any networks on application IDs instead
of based on destination IP addresses, are very essential for
connecting to access dynamic
workloads in on-demand Cloud DCs.

Therefore, this document discusses the use of services.

Issues associated with using SD-WAN techniques for connecting to
improve enterprise-to-cloud DC and cloud DC-to-cloud DC
connectivity. Cloud services
are also discussed in this document.

2. Definition of terms

Cloud DC: Third party Data Centers that usually host applications
and workload owned by different organizations or
tenants.

Controller: Used interchangeably with SD-WAN controller to manage
SD-WAN overlay path creation/deletion and monitoring the
path conditions between two or more sites.

DSVPN: Dynamic Smart Virtual Private Network. DSVPN is a secure
network that exchanges data between sites without
needing to pass traffic through an organization's
headquarter virtual private network (VPN) server or
router.

Heterogeneous Cloud: applications and workloads split among Cloud
DCs owned or managed by different operators.

Hybrid Clouds: Hybrid Clouds refers to an enterprise using its own
on-premises DCs in addition to Cloud services provided
by one or more cloud operators. (e.g. AWS, Azure,
Google, Salesforces, SAP, etc).

SD-WAN: Software Defined Wide Area Network. In this document,
"SD-WAN" refers to the solutions of pooling WAN
bandwidth from multiple underlay networks to get better
WAN bandwidth management, visibility & control. When the
underlay networks are private networks, traffic can
traverse without additional encryption; when the
underlay networks are public, such as Internet, some
traffic needs to be encrypted when traversing through
(depending on user provided policies).

VPC: Virtual Private Cloud is a virtual network dedicated to
one client account. It is logically isolated from other
virtual networks in a Cloud DC. Each client can launch
his/her desired resources, such as compute, storage, or
network functions into his/her VPC. Most Cloud
operators' VPCs only support private addresses, some
support IPv4 only, others support IPv4/IPv6 dual stack.

3. High Level Issues of Connecting to Multi-Cloud

There are many problems associated with connecting to hybrid Cloud
Services, many of which are out of the IETF scope. This section is
to identify some of the high level problems that can be addressed by
IETF, especially by Routing area. Other problems are out of the
scope of this document. By no means has this section covered all
problems for connecting to Hybrid Cloud Services, e.g. difficulty in
managing cloud spending is not discussed here.

3.1. Security Issues

Cloud Services is built upon shared infrastructure, therefore not
secure by nature. Security has been a primary, and valid, concern
from the start of cloud computing: you are unable to see the exact
location where your data is stored or being processed. Headlines
highlighting data breaches, compromised credentials, and broken
authentication, hacked interfaces and APIs, account hijacking
haven't helped alleviate concerns.

Secure user identity management, authentication, and access control
mechanisms are important. Developing appropriate security
measurements can enhance the confidence needed by enterprises to
fully take advantage of Cloud Services.

3.2. Authorization and Identity Management

One of the more prominent challenges for Cloud Services is Identity
Management and Authorization. The Authorization not only includes
user authorization, but also the authorization of API calls by
applications from different Cloud DCs managed by different Cloud
Operators. In addition, there are authorization for Workload
Migration, Data Migration, and Workload Management.

There are many types of users in cloud environments, e.g. end users
for accessing applications hosted in Cloud DCs, Cloud-resource users
who are responsible for setting permissions for the resources based
on roles, access lists, IP addresses, domains, etc.

There are many types of Cloud authorizations: including MAC
(Mandatory Access Control) - where each app owns individual access
permissions, DAC (Discretionary Access Control) - where each app
requests permissions from an external permissions app, RBAC (Role-
based Access Control) - where the authorization service owns roles
with different privileges on the cloud service, and ABAC (Attribute-
based Access Control) - where access is based on request attributes
and policies.

IETF hasn't yet developed comprehensive specification for Identity
management and data models for Cloud Authorizations.

3.3. API abstraction

Different Cloud Operators have different APIs to access their Cloud
resources, security functions, the NAT, etc.

It is difficult to move applications built by one Cloud operator's
APIs to another. However, it is highly desirable to have a single
and consistent way to manage the networks and respective security
policies for interconnecting applications hosted in different Cloud
DCs.

The difficulty of connecting applications in different Clouds might
be stemmed from the fact that they are direct competitors. Usually
traffic flow out of Cloud DCs incur charges. Therefore, direct
communications between applications in different Cloud DCs can be
more expensive than intra Cloud communications.

It is desirable to have a common API shim layer or abstraction for
different Cloud providers to make it easier to move applications
from one Cloud DC to another.

3.4. DNS for Cloud Resources

DNS name resolution is essential for on-premises and cloud-based
resources. For customers with hybrid workloads, which include on-
premises and cloud-based resources, extra steps are necessary to
configure DNS to work seamlessly across both environments.

Cloud operators have their own DNS to resolve resources within their
Cloud DCs and to well-known public domains. Cloud's DNS can be
configured to forward queries to customer managed authoritative DNS
servers hosted on-premises, and to respond to DNS queries forwarded
by on-premises DNS servers.

For enterprises utilizing Cloud services by different cloud
operators, it is necessary to establish policies and rules on
how/where to forward DNS queries to. When applications in one Cloud
need to communication with applications hosted in another Cloud,
there could be DNS queries from one Cloud DC being forwarded to the
enterprise's on premise DNS, which in turn be forwarded to the DNS
service in another Cloud. Needless to say, configuration can be
complex depending on the application communication patterns.

3.5. NAT for Cloud Services

Cloud resources, such as VM instances, are usually assigned with
private IP addresses. By configuration, some private subnets can
have the NAT function to reach out to external network and some
private subnets are internal to Cloud only.

Different Cloud operators support different levels of NAT functions.
For example, AWS NAT Gateway does not currently support connections
towards, or from VPC Endpoints, VPN, AWS Direct Connect, or VPC
Peering. https://docs.aws.amazon.com/AmazonVPC/latest/UserGuide/vpc-
nat-gateway.html#nat-gateway-other-services. AWS Direct
Connect/VPN/VPC Peering does not currently support any NAT
functionality.

Google's Cloud NAT allows Google Cloud virtual machine (VM)
instances without external IP addresses and private Google
Kubernetes Engine (GKE) clusters to connect to the Internet. Cloud
NAT implements outbound NAT in conjunction with a default route to
allow instances to reach the Internet. It does not implement inbound
NAT. Hosts outside of VPC network can only respond to established
connections initiated by instances inside the Google Cloud; they
cannot initiate their own, new connections to Cloud instances via
NAT.

For enterprises with applications running in different Cloud DCs,
proper configuration of NAT have to be performed in Cloud DC and in
their own on-premise DC.

3.6. Cloud Discovery

One of the concerns of using Cloud services is not aware where the
resource is actually located, especially Cloud operators can move
application instances from one place to another. When applications
in Cloud communicate with on-premise applications, it may not be
clear where the Cloud applications are located or to which VPCs they
belong.

It is highly desirable to have tools to discover cloud services in
much the same way as you would discover your on-premises
infrastructure. A significant difference is that cloud discovery
uses the cloud vendor's API to extract data on your cloud services,
rather than the direct access used in scanning your on-premises
infrastructure.

Standard data models, APIs or tools can alleviate concerns of
enterprise utilizing Cloud Resources, e.g. having a Cloud service
scan that connects to the API of the cloud provider and collects
information directly.

4. Interconnecting Enterprise Sites with Cloud DCs
3.1. Multiple

Considering that many enterprises already have existing VPNs (e.g.
MPLS based L2VPN or L3VPN) interconnecting branch offices & on-
premises data centers, connecting to Cloud services will be mixed of
different types of networks. When an enterprise's existing VPN
service providers do not have direct connections to workloads in a the
corresponding cloud DCs that the enterprise prefers to use, the
enterprise has to face additional infrastructure and operational
costs to utilize Cloud services.

4.1. Sites to Cloud DC

Most Cloud operators offer some type of network gateway through
which an enterprise can reach their workloads hosted in the Cloud
DCs. For example, AWS (Amazon Web Services) offers the following options to reach
workloads in AWS Cloud DCs:

- AWS Internet gateway allows communication between instances in
AWS VPC and the internet.
- AWS Virtual gateway (vGW) where IPsec tunnels [RFC6071] are
established between an enterprise's own gateway and AWS vGW, so
that the communications between those gateways can be secured
from the underlay (which might be the public Internet).
- AWS Direct Connect, which allows enterprises to purchase direct
connect from network service providers to get a private leased
line interconnecting the enterprises gateway(s) and the AWS
Direct Connect routers. In addition, an AWS Transit Gateway can
be used to interconnect multiple VPCs in different Availability
Zones. AWS Transit Gateway acts as a hub that controls how
traffic is forwarded among all the connected networks which act
like spokes.

As an example, some branch offices

Microsoft's ExpressRoute allows extension of an enterprise can connect to
over the Internet a private network to reach AWS's vGW via IPsec tunnels. Other branch
offices
any of the same enterprise Microsoft cloud services, including Azure and Office365.
ExpressRoute is configured using Layer 3 routing. Customers can connect opt
for redundancy by provisioning dual links from their location to AWS DirectConnect via two
Microsoft Enterprise edge routers (MSEEs) located within a private network (without any encryption). ). It third-
party ExpressRoute peering location. The BGP routing protocol is important for
enterprises
then setup over WAN links to be able provide redundancy to observe the specific behaviors when
connected cloud. This
redundancy is maintained from the peering data center into
Microsoft's cloud network.

Google's Cloud Dedicated Interconnect offers similar network
connectivity options as AWS and Microsoft. One distinct difference,
however, is that Google's service allows customers access to the
entire global cloud network by default. It does this by connecting
your on-premises network with the Google Cloud using BGP and Google
Cloud Routers to provide optimal paths to the different connections. regions of
the global cloud infrastructure.

Figure below shows an example of some tenants' of a tenant's workloads are
accessible via a virtual router connected by AWS Internet Gateway;
some are accessible via AWS vGW, and others are accessible via AWS
Direct Connect. vR1 uses IPsec to establish secure tunnels over the
Internet

Different types of access require different level of security
functions. Sometimes it is not visible to avoid paying extra fees end customers which type
of network access is used for the IPsec features provided
by AWS vGW. Some tenants can deploy a specific application instance. To
get better visibility, separate virtual routers (e.g. vR1 & vR2) can
be deployed to
connect to internet differentiate traffic and to/from different cloud GWs. It
is important for some enterprises to traffic from the secure channels
from vGW and DirectConnect, e.g. vR1 & vR2. Others may have one
virtual router connecting be able to both types of traffic. observe the specific
behaviors when connected by different connections.

Customer Gateway can be customer owned router or ports physically
connected to AWS Direct Connect GW.
+------------------------+
| ,---. ,---. |
| (TN-1 ) ( TN-2)|
| `-+-' +---+ `-+-' |
| +----|vR1|----+ |
| ++--+ |
| | +-+----+
| | /Internet\ For External
| +-------+ Gateway +----------------------
| \ / to reach via Internet
| +-+----+
| |
| ,---. ,---. |
| (TN-1 ) ( TN-2)|
| `-+-' +---+ `-+-' |
| +----|vR2|----+ |
| ++--+ |
| | +-+----+
| | / virtual\ For IPsec Tunnel
| +-------+ Gateway +----------------------
| | \ / termination
| | +-+----+
| | |
| | +-+----+ +------+
| | / \ For Direct /customer\
| +-------+ Gateway +----------+ gateway |
| \ / Connect \ /
| +-+----+ +------+
| |
+------------------------+

Figure 1: Examples of Multiple Cloud DC connections.

3.2. Interconnect Private and Public Cloud DCs

It is likely that hybrid designs will become the rule for cloud
services, as more enterprises see the benefits of integrating public
and private cloud infrastructures. However, enabling the growth of
hybrid cloud deployments in the enterprise requires fast and safe
interconnection between public and private cloud services.
For an enterprise to connect to applications & workloads hosted in
multiple Cloud DCs, the enterprise can use IPsec tunnels established
over the Internet or a (virtualized) leased line service to connect
its on-premises gateways to each of the Cloud DC's gateways, virtual
routers instantiated in the Cloud DCs, or any other suitable design
(including a combination thereof).

Some enterprises prefer to instantiate their own virtual
CPEs/routers inside the Cloud DC to connect the workloads within the
Cloud DC. Then an overlay path is established between customer
gateways to the virtual CPEs/routers for reaching the workloads
inside the cloud DC.

3.3. Desired Properties for Networks that interconnect Hybrid Clouds

The networks that interconnect hybrid cloud DCs must address the
following requirements:
- High availability to access all workloads in the desired cloud
DCs.
Many enterprises include cloud infrastructures in their
disaster recovery strategy, e.g., by enforcing periodic backup
policies within the cloud, or by running backup applications in
the Cloud, etc. Therefore, the connection to the cloud DCs may
not be permanent, but rather needs to be on-demand.

- Global reachability from different geographical zones, thereby
facilitating the proximity of applications as a function of the
end users' location, to improve latency.
- Elasticity: prompt connection to newly instantiated
applications at Cloud DCs when usages increase and prompt
release of connection after applications at locations being
removed when demands change.
Some enterprises have front-end web portals running in cloud
DCs and database servers in their on-premises DCs. Those Front-
end web portals need to be reachable from the public Internet.
The backend connection to the sensitive data in database
servers hosted in the on-premises DCs might need secure
connections.

- Scalable security management. IPsec is commonly used to
interconnect cloud gateways with CPEs deployed in the
enterprise premises. For enterprises with a large number or
branch offices, managing the IPsec's Security Associations
among many nodes can be very difficult.

4. Multiple Clouds Interconnection

4.1. Multi-Cloud

4.2. Inter-Cloud Interconnection

Enterprises today can instantiate their workloads or applications in
Cloud DCs owned by different Cloud providers, e.g. AWS, Azure,
GoogleCloud, Oracle, etc. Interconnecting those workloads involves
three parties:

The Enterprise, its network service providers, and
the Cloud providers.

All Cloud Operators offer secure ways to connect enterprises' on-
prem sites/DCs with their Cloud DCs.

Some Cloud Operators allow enterprises to connect via private
networks. For example, AWS's DirectConnect allows enterprises to use rd 3 party provided private Layer 2 path from enterprises' GW to AWS
DirectConnect GW. Microsoft's ExpressRoute allows extension of a
private network to any of the Microsoft cloud services, including
Azure and Office365. ExpressRoute is configured using Layer 3
routing. Customers can opt for redundancy by provisioning dual links
from their location to two Microsoft Enterprise edge routers (MSEEs)
located within a third-party ExpressRoute peering location. The BGP
routing protocol is then setup over WAN links to provide redundancy
to the cloud. This redundancy is maintained from the peering data
center into Microsoft's cloud network.

Google's Cloud Dedicated Interconnect offers similar network connectivity options as AWS and Microsoft. One distinct difference,
however, is that Google's service allows customers access to the
entire global cloud network by default. It does this by connecting
your on-premises network with the Google Cloud using BGP and Google Cloud Routers to provide optimal paths to the different regions of DCs described in the global cloud infrastructure.

All those connectivity options previous
section are between for reaching Cloud providers' DCs and
the Enterprises, DCs, but not between cloud
DCs. For example, to connect When applications in AWS Cloud need to communicate with
applications in Azure Cloud, there must
be Azure, today's practice requires a third-party
gateway (physical or virtual) to interconnect the AWS's Layer 2
DirectConnect path with Azure's Layer 3 ExpressRoute.

Enterprises can also instantiate their own virtual routers in
different Cloud DCs and administer IPsec tunnels among them, which
by itself is not a trivial task. Or by leveraging open source VPN
software such as strongSwan, you create an IPSec connection to the
Azure gateway using a shared key. The strong swan StrongSwan instance within AWS
not only can connect to Azure but can also be used to facilitate
traffic to other nodes within the AWS VPC by configuring forwarding
and using appropriate routing rules for the VPC.

Most Cloud operators, such as AWS VPC or Azure VNET, use non-globally non-
globally routable CIDR from private IPv4 address ranges as specified
by RFC1918. To establish IPsec tunnel between two Cloud DCs, it is
necessary to exchange Public routable addresses for applications in
different Cloud DCs. [BGP-SDWAN] describes one method. Other methods
are worth exploring.

In summary, here are some approaches, available now (which might
change in the future), to interconnect workloads among different
Cloud DCs:

a) Utilize Cloud DC provided inter/intra-cloud connectivity
services (e.g., AWS Transit Gateway) to connect workloads
instantiated in multiple VPCs. Such services are provided with
the cloud gateway to connect to external networks (e.g., AWS
DirectConnect Gateway).
b) Hairpin all traffic through the customer gateway, meaning all
workloads are directly connected to the customer gateway, so
that communications among workloads within one Cloud DC must
traverse through the customer gateway.
c) Establish direct tunnels among different VPCs (AWS' Virtual
Private Clouds) and VNET (Azure's Virtual Networks) via
client's own virtual routers instantiated within Cloud DCs.
DMVPN (Dynamic Multipoint Virtual Private Network) or DSVPN
(Dynamic Smart VPN) techniques can be used to establish direct
Multi-point-to-Point or multi-point-to multi-point tunnels
among those client's own virtual routers.

Approach a) usually does not work if Cloud DCs are owned and managed
by different Cloud providers.

Approach b) creates additional transmission delay plus incurring
cost when exiting Cloud DCs.

For the Approach c), DMVPN or DSVPN use NHRP (Next Hop Resolution
Protocol) [RFC2735] so that spoke nodes can register their IP
addresses & WAN ports with the hub node. The IETF ION
(Internetworking over NBMA (non-broadcast multiple access) WG
standardized NHRP for connection-oriented NBMA network (such as ATM)
network address resolution more than two decades ago.

There are many differences between virtual routers in Public Cloud
DCs and the nodes in an NBMA network. NHRP cannot be used for
registering virtual routers in Cloud DCs unless an extension of such
protocols is developed for that purpose, e.g. taking NAT or dynamic
addresses into consideration. Therefore, DMVPN and/or DSVPN cannot
be used directly for connecting workloads in hybrid Cloud DCs.

Other protocols such as BGP can be used, as described in [BGP-
SDWAN].

4.2. Desired Properties for Multi-Cloud Interconnection

Different Cloud Operators have different APIs to access their Cloud
resources. It is difficult to move applications built by one Cloud
operator's APIs to another. However, it is highly desirable to have
a single and consistent way to manage the networks and respective
security policies for interconnecting applications hosted in
different Cloud DCs.

The desired property would be having a single network fabric to
which different Cloud DCs and enterprise's multiple sites can be
attached or detached, with a common interface for setting desired
policies. SDWAN is positioned to become that network fabric enabling
Cloud DCs to be dynamically attached or detached. But the reality is
that different Cloud Operators have different access methods, and
Cloud DCs might be geographically far apart. More Cloud connectivity
problems are described in the subsequent sections.

5. Problems with MPLS-based VPNs extending to Hybrid Cloud DCs

Traditional MPLS-based VPNs have been widely deployed as an
effective way to support businesses and organizations that require
network performance and reliability. MPLS shifted the burden of
managing a VPN service from enterprises to service providers. The
CPEs attached to MPLS VPNs are also simpler and less expensive,
since
because they do not need to manage routes to remote sites; they
simply pass all outbound traffic to the MPLS VPN PEs to which the
CPEs are attached (albeit multi-homing scenarios require more
processing logic on CPEs). MPLS has addressed the problems of
scale, availability, and fast recovery from network faults, and
incorporated traffic-engineering capabilities.

However, traditional MPLS-based VPN solutions are sub-optimized for
connecting end-users to dynamic workloads/applications in cloud DCs
because:

- The Provider Edge (PE) nodes of the enterprise's VPNs might not
have direct connections to third party cloud DCs that are used
for hosting workloads with the goal of providing an easy access
to enterprises' end-users.

- It usually takes some time to deploy provider edge (PE) routers at new
locations. When enterprise's workloads are changed from one
cloud DC to another (i.e., removed from one DC and re-
instantiated to another location when demand changes), the
enterprise branch offices need to be connected to the new cloud
DC, but the network service provider might not have PEs located
at the new location.

One of the main drivers for moving workloads into the cloud is
the widely available cloud DCs at geographically diverse
locations, where apps can be instantiated so that they can be
as close to their end-users as possible. When the user base
changes, the applications may be migrated to a new cloud DC
location closest to the new user base.

- Most of the cloud DCs do not expose their internal networks. An
enterprise with a hybrid cloud deployment can use an MPLS-VPN
to connect to a Cloud provider at multiple locations. The
connection locations often correspond to gateways of different
Cloud DC locations from the Cloud provider. The different
Cloud DCs are interconnected by the Cloud provider's own
internal network. At each connection location (gateway), the
Cloud provider uses BGP to advertise all of the prefixes in the
enterprise's VPC, regardless of which Cloud DC a given prefix
is actually in. This can result in inefficient routing for the
end-to-end data path.

- Extensive usage of Overlay by Cloud DCs:

Many cloud DCs use an overlay to connect their gateways to the
workloads located inside the DC. There is currently no standard
that specifies the interworking between the Cloud Overlay and
the enterprise' existing underlay networks. One of the
characteristics of overlay networks is that some of the WAN
ports of the edge nodes connect to third party networks. There
is therefore a need to propagate WAN port information to remote
authorized peers in third party network domains in addition to
route propagation. Such an exchange cannot happen before
communication between peers is properly secured.

Another roadblock is the lack of a standard way to express and
enforce consistent security policies for workloads that not only use
virtual addresses, but in which are also very likely hosted in
different locations within the Cloud DC [RFC8192]. The current VPN
path computation and bandwidth allocation schemes may not be
flexible enough to address the need for enterprises to rapidly
connect to dynamically instantiated (or removed) workloads and
applications regardless of their location/nature (i.e., third party
cloud DCs).

6. Problem with using IPsec tunnels to Cloud DCs
As described in the previous section, many Cloud operators expose
their gateways for external entities (which can be enterprises
themselves) to directly establish IPsec tunnels. Enterprises can
also instantiate virtual routers within Cloud DCs to connect to
their on-premises devices via IPsec tunnels.

6.1. Scaling Issues with IPsec Tunnels

If there is only one enterprise location that needs to reach the
Cloud DC, an IPsec tunnel is a very convenient solution.

However, many medium-to-large enterprises usually have multiple sites and
multiple data centers. For workloads and apps hosted in
cloud DCs, multiple sites need to communicate securely with those
cloud
workloads and apps. This section documents some of the issues
associated with using IPsec tunnels to connect enterprise premises
with cloud gateways.

6.1. Complexity of multi-point any-to-any interconnection

The dynamic workload instantiated apps hosted in cloud DCs, Cloud DC needs gateways have to communicate
with multiple branch offices and on-premises data centers. Most
enterprises need multi-point interconnection among multiple
locations, which can be provided by means of MPLS L2/L3 VPNs.

Using
maintain many IPsec overlay paths tunnels to connect all branches & on-premises data
centers to cloud DCs those locations. In addition,
each of those IPsec Tunnels requires CPEs to manage routing among Cloud DCs
gateways and the CPEs located at other branch pair-wise periodic key
refreshment. For a company with hundreds or thousands of locations, which can
dramatically increase the complexity
there could be hundreds (or even thousands) of the design, possibly IPsec tunnels
terminating at the
cost of jeopardizing the CPE performance.

The complexity of requiring CPEs to maintain routing among other
CPEs cloud DC gateway, which is very processing
intensive. That is one of the reasons why enterprises migrated from Frame Relay
based services to MPLS-based VPN services.

MPLS-based VPNs have their PEs directly connected to the CPEs.
Therefore, CPEs only need to forward all traffic to the directly
attached PEs, which are therefore responsible for enforcing the
routing policy within the corresponding VPNs. Even for multi-homed
CPEs, the CPEs many cloud operators only need allow a limited
number of (IPsec) tunnels & bandwidth to forward traffic among the directly
connected PEs. However, when using each customer.

Alternatively, you could use a solution like group encryption where
a single IPsec tunnels between CPEs and
Cloud DCs, SA is necessary at the CPEs need to compute, select, establish GW but the drawback is key
distribution and maintain
routes for traffic to be forwarded to Cloud DCs, to remote CPEs via
VPN, or directly. maintenance of a key server, etc.

6.2. Poor performance over long distance

When enterprise CPEs or gateways are far away from cloud DC gateways
or across country/continent boundaries, performance of IPsec tunnels
over the public Internet can be problematic and unpredictable. Even
though there are many monitoring tools available to measure delay
and various performance characteristics of the network, the
measurement for paths over the Internet is passive and past
measurements may not represent future performance.

Many cloud providers can replicate workloads in different available
zones. An App instantiated in a cloud DC closest to clients may have
to cooperate with another App (or its mirror image) in another
region or database server(s) in the on-premises DC. This kind of
coordination requires predicable networking behavior/performance
among those locations.

6.3. Scaling Issues with IPsec Tunnels

IPsec can achieve secure overlay connections between two locations
over any underlay network, e.g., between CPEs and Cloud DC Gateways.

If there is only one enterprise location connected to the cloud
gateway, a small number of IPsec tunnels can be configured on-demand
between the on-premises DC and the Cloud DC, which is an easy and
flexible solution.

However, for multiple enterprise locations to reach workloads hosted
in cloud DCs, the cloud DC gateway needs to maintain multiple IPsec
tunnels to all those locations (e.g., as a hub & spoke topology).
For a company with hundreds or thousands of locations, there could
be hundreds (or even thousands) of IPsec tunnels terminating at the
cloud DC gateway, which is not only very expensive (because Cloud
Operators usually charge their customers based on connections), but
can be very processing intensive for the gateway. Many cloud
operators only allow a limited number of (IPsec) tunnels & bandwidth
to each customer. Alternatively, you could use a solution like
group encryption where a single IPsec SA is necessary at the GW but
the drawback here is key distribution and maintenance of a key
server, etc.

7. Problems of Using SD-WAN to connect to Cloud DCs
SD-WAN can establish parallel paths over multiple underlay networks
between two locations on-demand, for example, to support the
connections established between two CPEs interconnected by a
traditional MPLS VPN ([RFC4364] or [RFC4664]) or by IPsec [RFC6071]
tunnels.

SD-WAN lets enterprises augment their current VPN network with cost-
effective, readily available Broadband Internet connectivity,
enabling some traffic offloading to paths over the Internet
according to differentiated, possibly application-based traffic
forwarding policies, or when the MPLS VPN connection between the two
locations is congested, or otherwise undesirable or unavailable.

7.1. SD-WAN among branch offices vs. interconnect More Complexity to Cloud DCs

SD-WAN interconnection of branch offices Edge Nodes

Augmenting transport path is not as simple as it appears. For an
enterprise with multiple sites, using SD-WAN CPE managed overlay paths among
sites requires each CPE to manage all the addresses that local hosts
have the potential to reach, i.e., map internal VPN addresses to
appropriate SD-WAN Overlay paths. This is similar to the complexity of
Frame Relay based VPNs, where each CPE needed to maintain mesh
routing for all destinations if they were to avoid an extra hop
through a hub router. Even though SD-WAN CPEs can get with the assistance from a central
controller (instead of running a routing protocol) to resolve the
mapping between destinations and SD-WAN paths, SD-WAN CPEs are still
responsible for routing table maintenance as remote destinations
change their attachments, e.g., the dynamic workload in other DCs
are de-commissioned or added.

Even though originally envisioned for interconnecting branch
offices, SD-WAN offers a very attractive way for enterprises to
connect to Cloud DCs.

The SD-WAN

In addition, overlay path for interconnecting branch offices and the SD-WAN for
interconnecting are
different from connecting to Cloud DCs have some differences: DCs:

- SD-WAN for Overlay path interconnecting branch offices usually have two end-
points (e.g.,
end-points (e.g. CPEs) controlled by one entity (e.g., a
controller (e.g.
controllers or management system systems operated by the enterprise).
- SD-WAN for Connecting to Cloud DC interconnects may consider consists of CPEs owned or managed by
the enterprise, while and the remote end-points are being managed or
controlled by Cloud DCs (For the ease of
description, let's call such CPEs asymmetrically-managed CPEs).

- Cloud DCs may DCs.

7.2. Edge WAN Port Management

An SDWAN edge node can have WAN ports connected to different entry points (or devices) with one
entry point that terminates a private direct connection (based
upon
networks or public internet managed by different operators.
There is therefore a leased line for example) and other entry points being
devices terminating the IPsec tunnels, as shown need to propagate WAN port property to
remote authorized peers in Figure 2.

Therefore, the SD-WAN design becomes asymmetric.
+------------------------+
| ,---. ,---. |
| (TN-1 ) ( TN-2)| TN: Tenant applications/workloads
| `-+-' +---+ `-+-' |
| +----|vR1|----+ |
| ++--+ |
| | +-+----+
| | /Internet\ One path via
| +-------+ Gateway +---------------------+
| \ / Internet \
| +-+----+ \
+------------------------+ \
\
+------------------------+ native traffic \
| ,---. ,---. | without encryption|
| (TN-3 ) ( TN-4)| |
| `-+-' +--+ `-+-' | | +------+
| +----|vR|-----+ | +----+ CPE |
| ++-+ | | +------+
| | +-+----+ |
| | / virtual\ One path via IPsec Tunnel |
| +-------+ Gateway +-------------------------- +
| \ / Encrypted traffic over|
| +-+----+ public third party network |
+------------------------+ |
|
+------------------------+ |
| ,---. ,---. | Native traffic |
| (TN-5 ) ( TN-6)| without encryption |
| `-+-' +--+ `-+-' | over secure network|
| +----|vR|-----+ | |
| ++-+ | |
| | +-+----+ +------+ |
| | / \ Via Direct /customer\ |
| +-------+ Gateway +----------+ gateway |-----+
| \ / Connect \ /
| +-+----+ +------+
+------------------------+Customer GW has physical connection domains in
addition to AWS GW

Figure 2: Different Underlays route propagation. Such an exchange cannot happen
before communication between peers is properly secured.

7.3. Forwarding based on Application
Forwarding based on application IDs instead of based on
destination IP addresses is often referred to Reach Cloud DC as Application based
Segmentation. If the Applications have unique IP addresses, then
the Application Based Segmentation can be achieved by propagating
different BGP UPDATE messages to different nodes, as described in
[BGP-SDWAN-USAGE]. If the Application cannot be uniquely
identified by the IP addresses, more work is needed.

8. End-to-End Security Concerns for Data Flows

When IPsec tunnels established from enterprise on-premises CPEs
are terminated at the Cloud DC gateway where the workloads or
applications are hosted, some enterprises have concerns regarding
traffic to/from their workload being exposed to others behind the
data center gateway (e.g., exposed to other organizations that
have workloads in the same data center).
To ensure that traffic to/from workloads is not exposed to
unwanted entities, IPsec tunnels may go all the way to the
workload (servers, or VMs) within the DC.

9. Requirements for Dynamic Cloud Data Center VPNs

In order to address the aforementioned issues, any solution for
enterprise VPNs that includes connectivity to dynamic workloads or
applications in cloud data centers should satisfy a set of
requirements:

- The solution should allow enterprises to take advantage of the
current state-of-the-art in VPN technology, in both traditional
MPLS-based VPNs and IPsec-based VPNs (or any combination
thereof) that run over the public Internet.
- The solution should not require an enterprise to upgrade all
their existing CPEs.
- The solution should support scalable IPsec key management among
all nodes involved in DC interconnect schemes.
- The solution needs to support easy and fast, on-the-fly, VPN
connections to dynamic workloads and applications in third
party data centers, and easily allow these workloads to migrate
both within a data center and between data centers.
- Allow VPNs to provide bandwidth and other performance
guarantees.
- Be a cost-effective solution for enterprises to incorporate
dynamic cloud-based applications and workloads into their
existing VPN environment.

10. Security Considerations

The draft discusses security requirements as a part of the problem
space, particularly in sections 4, 5, and 8.

Solution drafts resulting from this work will address security
concerns inherent to the solution(s), including both protocol
aspects and the importance (for example) of securing workloads in
cloud DCs and the use of secure interconnection mechanisms.

11. IANA Considerations

This document requires no IANA actions. RFC Editor: Please remove
this section before publication.

12. References

12.1. Normative References
12.2. Informative References

[RFC2735] B. Fox, et al "NHRP Support for Virtual Private
networks". Dec. 1999.

[RFC8192] S. Hares, et al "Interface to Network Security Functions
(I2NSF) Problem Statement and Use Cases", July 2017

[ITU-T-X1036] ITU-T Recommendation X.1036, "Framework for creation,
storage, distribution and enforcement of policies for
network security", Nov 2007.

[RFC6071] S. Frankel and S. Krishnan, "IP Security (IPsec) and
Internet Key Exchange (IKE) Document Roadmap", Feb 2011.

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

[RFC4664] L. Andersson and E. Rosen, "Framework for Layer 2 Virtual
Private Networks (L2VPNs)", Sept 2006.

[BGP-SDWAN] L. Dunbar, et al. "BGP Extension for SDWAN Overlay
Networks", draft-dunbar-idr-bgp-sdwan-overlay-ext-03,
work-in-progress, Nov 2018.

13. Acknowledgments

Many thanks to Alia Atlas, Chris Bowers, Ignas Bagdonas, Michael
Huang, Liu Yuan Jiao, Katherine Zhao, and Jim Guichard for the
discussion and contributions.

Authors' Addresses

Linda Dunbar
Futurewei
Email: Linda.Dunbar@futurewei.com

Andrew G. Malis
Independent
Email: agmalis@gmail.com

Christian Jacquenet
Orange
Rennes, 35000
France
Email: Christian.jacquenet@orange.com

Mehmet Toy
Verizon
One Verizon Way
Basking Ridge, NJ 07920
Email: mehmet.toy@verizon.com