| none | S. Homma |
| Internet-Draft | NTT |
| Intended status: Informational | X. de Foy |
| Expires: September 6, 2018 | InterDigital Inc. |
| A. Galis | |
| University College London | |
| March 5, 2018 |
Gateway Function for Network Slicing
draft-homma-coms-slice-gateway-01
This document describes the roles and requirements for a slice gateway that is a data plane function or function group for connecting/disconnecting and compose/decompose network slice subnets and providing network slices from end to end. The interworkings between management and control elements at the management and control planes with the gateway function for controlling and orchestrating end-to-end network slices are also presented in this document.
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). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.
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."
This Internet-Draft will expire on September 6, 2018.
Copyright (c) 2018 IETF Trust and the persons identified as the document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://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.
Network slicing is an approach to create separate virtual networks in support of service depending on several requirements on the same physical resources, and it enables networks to adapt to requirements, which is diverse more, inexpensively and flexibly. It's also expected to enhance usability of infrastructural networks for tenants and create new business opportunities. For example, by using network slices lent from infrastructure operators, other industrial companies can provide communication services including ensurance of network transport without having physical infrastructure.
From a business point of view, a slice includes a combination of all the relevant network resources, functions, and assets required to fulfill a specific business case or service, including OSS, BSS and DevOps processes.
From the network infrastructure point of view, network slice requires the partitioning and assignment of a set of resources that can be used in an isolated, disjunctive or non- disjunctive manner for that slice.
From the tenant point of view, network slice provides different capabilities, specifically in terms of their management and control capabilities, and how much of them the network service provider hands over to the slice tenant. As such there are two kinds of slices: (A) Inner slices, understood as the partitions used for internal services of the provider, retaining full control and management of them. (B) Outer slices, being those partitions hosting customer services, appearing to the customer as dedicated networks.
Network slices are established with combination of various technologies, such as software defined network (SDN), network function virtualization (NFV), or traffic engineering, and managed/operated with automation technologies such as orchestrator.
Assumed use cases of network slices include establishment of virtual networks whose qualities are guaranteed from end to end under the supervision of multi-domain orchstrators. In such cases, a network slice subnet is created on each domain, such as access network and core network, and such network slices are composed of connected subnets.
Network slice subnets are built based on specification of the underlay network, and thus the used technologies might vary. Therefore, a gateway function, which enables to connect subnets while adapting the differentiations and forward data packets to/from the appropriate next subnet, is required. Defining a new data plane technology is not a goal of this draft. This draft aims to specify management-related requirements for an SLG, which may be implemented using existing data plane technologies.
In this document, the gateway function is called slice gateway or SLG, and the role and requirements are described.
SLG main role is the enablement of interworkings between data plane with management and control elements for controlling and orchestrating end-to-end slices.
Use cases of network slices are discussed in several Standard Developing Organizations (SDOs). Some examples are described in use cases document ([I-D.netslices-usecases]).
In some proposed use cases, an NS is structured across multiple network domains. The capability of NS subnets might be different because the components are domain-specific. In particular, the differentiation in capability between different administrative domains is large.
For connecting some different NS subnets and providing a NS that guarantees the prescribed quality from end to end, SLGs are required to connect such NS subnets. SLGs enable to provide E2E-NSs independently of specifications of underlay networks by hiding the differentiations and connecting between NS subnets. An overview of this concept is shown in Figure 1. SLGs glue NS subnets established on each domain and provide an E2E-NS.
E2E-NS
________________________A_________________________
/ \
_____________ ____________ _________________
/ // / / /
end +---+ NS +---+ NS +---+ +---+ NS ,------.
host==>|SLG| Subnet |SLG| Subnet |SLG+-+SLG| Subnet( Server )
+---+ #1 +---+ #2 +---+ +---+ #3 `------'
/____________//___________/ /________________/
/____________//___________/ /________________/
: : :
: : :
.--. .--. .--.
( )-. ( )-. ( )-.
.' Access ' .' Core ' .' Data '
( Network ) ( Network ) ( Center )
( -' ( -' ( /Cloud -'
'-( ) '-( ) '-( )
'---' '---' '---'
\______ ______/ \_____ _____/ \________ ________/
V V V
Domain#1 Domain#2 Domain#3
\_____________ _____________/ \________ ________/
V V
Domain of Administrator#A Domain of Administrator#B
Figure 1: E2E-NS composed of multiple NS subnets
Moreover, identification of user service traffic and their allocation/disallocation to the appropriate NS subnet are required at the edges of E2E-NSs, as shown in Figure 2, and SLGs might take on these roles.
+-----+ _______________
end | |-->/_______________
host ====>| SLG | NS Subnet#1
|@Edge| _______________
| |-->/______________
| | NS Subnet#2
| | : :
+-----+
Figure 2: NS subnet selection of SLG
Note that, this model has the assumption that transitions of data packets from one NS subnet to another are executed at only SLGs. Also, an SLG is not necessarily implemented as a single device or virtual machine (VM).
The architecture overview of NS management system is shown in Figure 3. Orchestrators manage whole resources including network elements and server resources (i.e., routing, bandwidth, compute or storage). In this figure, the resources including network elements and server resources are managed by resource orchestrators installed in each domain, and the E2E-orchestrator and network service orchestrator handle resource orchestrators.
NSs are requested from NS tenants via the portal system and the order of creations of an NS is given to the E2E orchestrator from the portal system via BSS/OSS. When an NS across multiple administrative domains are requested, the portal system that received the request forwards the order to create NS subnets to the other infrastructure providers' systems via Cross-Segment Slice Manager. The details of COMS architecture are described in the architecture document ([I-D.qiang-coms-architecture]).
SLGs are also controlled via orchestrators. An SLG basically belongs to a network element, and it might also belong to server resource if it runs as a VNF. (An example of position of SLG deployed as a VNF is shown in Appendix A.)
SLGs are located at the edges of each NS subnet. They translate data packets into the appropriate form and send them to the next NS subnet. SLGs located at the end of E2E-NSs additionally provide identification of data packets and select the assigned NS subnet based on the identification result.
The information model used in this architecture is described in information model document ([I-D.qiang-coms-netslicing-information-model]).
NS Tenant
|
. .|. . . . . . . . . . . . . . . . . . . . .
. +-v---------+ .
. |Portal/GUI +--+ .
. +-+---------+ | . . . . . . . . . .
. | +-v-----------------------+ . . +------------+ .
. | |CSS-Mngr./TNS-Orch. |<------->|CSS-M/TNS-O | .
. | +-+-------+---------------+ . . +-+--------+-+ .
. | | | . . | | .
. +-v-----+ | | . . +-v-----+ | .
. |BSS/OSS|<-----+ | . . |BSS/OSS| | .
. +-+-----+ | . . +-+-----+ | .
. | | . . | | .
. +-v--------------------v-----------+ . . +-v--------v-+ .
. | E2E-Orch./Network Service-Orch. | . . |E2E-O/NS-O | .
. +-+--------------------+-----------+ . . +-+--------+-+ .
. | | . . | | .
. +-v----------------+ +-v----------------+ . . : .
. | Resource Orch.#1 | | Resource Orch.#2 |.. . . . . . . . . . .
. +-+---------+------+ +-+---------+------+ . Administrative
. | | | | . Domain#2
. +-v------++-v------+ +-v------++-v------+ .
. |Network ||NFV | |Network ||NFV | .
. |Ctrl. ||Ctrl. | |Ctrl. ||Ctrl. | .
. +-+------++-+------+ +-+------++-+------+ .
. | | | | .
. +-v------++-v------+ +-v------++-v------+ .
. |Network ||Server | |Network ||Server | .
. |Elements||Resource| |Elements||Resource|.. .
. |in ||in | |in ||in | .
. |Domain#1||Domain#1| |Domain#2||Domain#2| .
. +--------++--------+ +--------++--------+ .
. . . . . . . . . . . . . . . . . . . . . . .
Administrative Domain#1
CSS-Mngr./CSS-M:Cross-Segment Slice Manager
TNS-Orch./TNS-O:Transport Network Slice Orchestrator
Figure 3: Overview of NS Management Architecture
An SLG is basically a component in the data plane and has the roles of data packet processing. Moreover, it is required to have functions for control/management processes such as connecting to underlay networks or managing NSs.
Furthermore, an SLG might be required to support handling services provided on NSs in addition to controlling of NS because an SLG is an edge node on an E2E-NS.
In this section, we describe the requirements for an SLG in terms of the following aspects and their interworkings.
SLGs at the edge of E2E-NSs MUST have the capability to identify and classify data packets, and assign them to the appropriate E2E-NS. This requirement varies depending on the location.
SLGs MUST provide functions for transport data packets depending on the specifications of the underlay networks.
____________________________
/ . . . . . /
+-----+ . . +-----+
| |. . . .| |
| SLG | | SLG |
| |* * * *| |
+-----+ * * +-----+
/ * * * * * /
/___________________________/
NS Subnet
*** : Main-route
... : Sub-route
Figure 4: An example of traffic distribution by SLG
In NSaaS, isolation control is required for avoiding an NS being affect by other NSs. Traffic engineering or QoS control is ones of the most fundamental approaches to prevent disturbances between NSs.
If an SLG is composed of a combination of several components, a service chaining mechanism is required to make them work together and achieve SLG functionality.
Moreover, some NSs may traverse NFVs such as firewalls or cache servers for providing value-added services to their users. In such cases, SLG might be required to support service chaining mechanisms, such as handling of network service header (NSH) defined in [RFC8300]. If an NS includes the service chaining architecture defined in [RFC7665], some SLG would be required to support following functions; classifier(CF), service function forwarder (SFF), and inter boundary node(IBN). (Details of CF, SFF and IBN are described in SFC documents; [RFC7665], [I-D.ietf-sfc-hierarchical].)
SLG MUST have interface to its controller or operation systems for set parameters related to the data plane functions described in Section 5.1.1. In addition, an SLG at the edges of E2E-NSs MUST have interfaces to authentication servers.
An SLG MUST support address resolution or routing mechanisms to connect to underlay network elements including routers or L2 switches.
For preventing entry of irregular traffic to NSs, an SLG at the edge of E2E-NS MUST support AAA mechanism for incoming traffic. Also, when an SLG connects to another SLG in other administrative domain, SLGs should have a mechanism to confirm that the connection is established with the regular processes. For example, an SLG is required to support authentication of the opponent SLG with key information indicated from higher-level operation systems.
In management of NSs, OAM or monitoring mechanisms for both underlay and overlay networks is required for SLGs. For an underlay network, an SLG MUST have OAM functions to confirm connectivity to interconnect equipment. For an overlay network, an SLG MUST have OAM functions to confirm connectivity to the some node on the same NS, and measure the traffic amount of flowing packets on each NS.
In NSaaS, some NS tenants may need delivery of an individual service to each user, device, or application on the same NS. For such service deliveries, an SLG might be required to identify and classify user traffic based on some information such as subscriber ID or payload of data packets. Also, an SLG should be controllable from the NS tenant.
An NS accommodates several communication devices and SLGs might be required to have fair queueing mechanisms for maintaining service quality of each user. Also, different types of service traffic that have different priorities might coexist on an NS. For example, some NS providers might provide telephone and internet access services to their users with an NS. In such cases, SLG might be required to provide QoS control mechanisms for enforcing priority control based on service priorities.
These QoS controls are executed depending on the information of inner packets and are independent of isolation mechanisms as infrastructure. An SLG might be required to have a hierarchical QoS control mechanism in case that both QoS controls for services over NSs and isolation between NSs are required.
SLG might be required to support steering or service chaining function for conveying data packets to the appropriate network functions deployed on an NS based on the classification result and user's contract information.
An SLG might have interfaces to controllers for managing user policies on each NS. Some controllers might be deployed on the same NS. If some controllers are located at external networks, they might require SLGs to have APIs.
In an NSaaS, collection of telemetry information of each NS might be required for understanding traffic usage. Thus, an SLG might be required to support to collect and report telemetry information of connected NSs.
This section describes considerations related with deployment of SLGs.
For providing E2E-NSs on existing network infrastructures, some components located at boundaries of domains are required to have the same set of functionality as an SLG. Examples of such components in each domain type are described below.
There are mainly three types of SLG for creating E2E-NS across multiple administrative domains. The requirements of each SLG type are listed in Appendix B.
This is located at an edge of an E2E-NS, and supports identification, classification and authentication of user traffic in addition to fundamental SLG functions, such as transport and isolation. Also, it might be required to have capabilities for services delivered on an NS.
This is located between NS subnets within a single administrative domain and has only fundamental functions. It is not necessarily required if a common transport mechanism in all domains is used.
This is located between NS subnets established on different domains. It supports authentication for connecting to the opponent SLG in addition to fundamental functions.
The connection form of an SLG varies depending on which type it is. Examples of horizontal connection forms of each SLG type are described below.
*Virtual Layer*
+-----+
host#1 ====>| | _______________
| |-->/_______________
host#2 ====>|E-SLG| NS Subnet#1
| | _______________
host#3 ====>| |-->/_______________
| | NS Subnet#2
: : | | : :
+-----+
////////////////////////////////////////
*Physical Layer*
,--------------------
[UE#1] -----\ /
[UE#2] -----[Edge] Domain#1
[UE#3] -----/ \
: : `-------------------
Edge: Edge Node
Figure 5: Overview of horizontal connection of E-SLG
*Virtual Layer*
+------+
_________ | | ___________
_________/-->|IS-SLG|--> /__________
NS Subnet#1 | | NS Subnet#2
+------+
///////////////////////////////////////
*Physical Layer*
--------------. ,--------------
\ /
Domain#1 [ GW ] Domain#2
/ \
--------------' `--------------
GW: Gateway Node
Figure 6: Overview of horizontal connection of IS-SLG
*Virtual Layer*
+------+ +------+
_________ | | ______ | | ___________
_________/-->|ID-SLG|O______)|ID-SLG|-->/__________
NS Subnet#1 | | Tunnel | | NS Subnet#2
+------+ +------+
///////////////////////////////////////////////////////
*Physical Layer*
--------------------. ,-------------------
Administrative \ / Administrative
Domain#1 [ GW ]---[ GW ] Domain#2
/ \
--------------------' `-------------------
GW: Gateway Node
Figure 7: Overview of horizontal connection of ID-SLG
There are two patterns of vertical connection of SLGs in the middle of E2E-NSs. The first pattern is that the SLGs accommodate only a set of NS subnets, which are composition of the same E2E-NS. In this pattern, such SLGs are not required to support NS subnet selection, however, establishment of a new SLG is required when a new E2E-NS is created. This might causes extra overheads because of deploying many SLGs.
The other pattern is that such SLGs are acceptable to accommodate multiple NS subnets from each domain. In this pattern, SLGs are support NS subnet selection. On the other hand, this pattern can restrain the number of SLGs. Also, it is easy to provide transit of data packets from an NS subnet to other subnet on the same domain.
The overviews of these patterns are shown in Figure 8 and Figure 9.
+-----+
_________ | | ___________
_________/-->|SLG#1|-->/__________
NS Subnet#1 | | NS Subnet#2
+-----+
+-----+
_________ | | ___________
_________/-->|SLG#2|-->/__________
NS Subnet#3 | | NS Subnet#4
+-----+
: : :
Figure 8: Overview of vertical connection of SLG: Separated Pattern
+-----+
_________ | | ___________
_________/-->| |-->/__________
NS Subnet#1 |SLG#1| NS Subnet#2
_________ | | ___________
_________/-->| |-->/__________
NS Subnet#3 | | NS Subnet#4
| |
: | | :
+-----+
Figure 9: Overview of vertical connection of SLG: Shared Pattern
An SLG can be created as either a software or hardware function. NSs are virtual networks created depending on requests from external NS tenants, and thus software would be more compatible with usage for NSs in terms of flexibility or manageability. Moreover, it enables to increase or decrease for each function if SLG is composed of combination of several components. However, it is difficult to provide high performance or sufficient throughput for carrier-grade networks with software function. In addition, it would be difficult to implement sufficient QoS control mechanisms with general servers, because they requires special hardware structures.
On the other hand, hardware appliances are able to provide high throughput compared with software. However, they are inflexible in terms of provisioning.
From the above considerations, operators should prepare SLG in appropriate ways depending on their usages or locations.
SLG provides interconnectivity between NS subnets. The concept and fundamental framework including the related NS information model are described in subnets interconnection document ([I-D.defoy-coms-subnet-interconnection]).
This section is focused on interconnection between NS subnets established on different administrative domains, and describes considerations related to this condition.
For interconnection between different administrative NS subnets, pre-arrangement of the transport protocol, which is used to connect between SLGs is required. Orchestration systems indicate the protocol and configuration to each SLG.
In addition to establishing connection, quality control of communication is important. SLGs of egress side should execute traffic shaping to prevent some NSs from excessively occupying the link between SLGs. Moreover, some SLGs are connected to several other SLGs that are deployed on the different locations. Therefore SLGs of the ingress side should execute traffic policing to avoid excessive inflow of traffic into some NSs. The parameters for these controls are pre-configured by orchestration systems.
The above approaches are ones of the simplest ways to provide quality assurance of inter-administrative subnets. If there is stricter isolation request, more considerations would be required.
For connecting networks of different administrators, secure interconnection schemes are required. Especially, in an NSaaS, networks might be connected to several networks, and schemes for ensuring secure connectivity would be more important.
SLGs confirm whether the opponent SLG is regular when it requests to connect, and reject the request if the SLG is not regular. In some cases, SLGs might be confirm whether the inner packets received from the other SLGs are sent from regular users.
Requirements and considerations for SLG related to security are described in Section 5 and Section 7.
This memo includes no request to IANA.
The authors would like to thank Li Qiang for her kind review and valuable feedback.
The mapping of SLG as a VM into ETSI NFV MANO architecture is described in Figure 10.
+---------------------------+
| BSS/OSS, Orchestrator |
+-+---------------+---------+
| |
+-v------+ +-v---------+
|SLG-Ctrl| | NFV Orch. |
+-+------+ +-+---------+
| |
,-v------. |
| SLG | |
:========: +-v---------+
| VM |<-----+ VNF Mngr. |
`--------' +-+---------+
+--------+ |
|HostOS/ | +-v---------+
|Server |<-----| VIM |
+--------+ +-----------+
Figure 10: Position of SLG as a VM on ETSI NFV MANO
The requirements for each SLG type are listed in Figure 11.
+---------------++-------+-------+-------+----------------+
| || E-SLG |IS-SLG |ID-SLG | Reference |
+=========================================================+
|*Data-Plane of NS as Infrastructure |
+=========================================================+
|Identification/|| M | O | O |Section 5.1.1.1.|
|Classification || | | | |
+---------------++-------+-------+-------+----------------+
|Transport/ || M | O | M |Section 5.1.1.2.|
|Forwarding || | | | |
+---------------++-------+-------+-------+----------------+
|Isolation || M | M | M |Section 5.1.1.3.|
+---------------++-------+-------+-------+----------------+
|Service Chain || O | O | O |Section 5.1.1.4.|
+=========================================================+
|*Control/Management-Plane of NS as Infrastructure |
+=========================================================+
|IF to Ctrl/OpS || M | M | M |Section 5.1.2.1.|
+---------------++-------+-------+-------+----------------+
|Addr Resolution|| M | M | M |Section 5.1.2.2.|
|/Routing || | | | |
+---------------++-------+-------+-------+----------------+
|AAA || M | - | M |Section 5.1.2.3.|
+---------------++-------+-------+-------+----------------+
|OAM || M | M | M |Section 5.1.2.4.|
+=========================================================+
|*Data-Plane for Service on NS |
+=========================================================+
|Identification/|| O | - | O |Section 5.2.1.1.|
|Classification || | | | |
+---------------++-------+-------+-------+----------------+
|QoS Control || O | O | O |Section 5.2.1.2.|
+---------------++-------+-------+-------+----------------+
|Steering/ || O | - | O |Section 5.2.1.3.|
|Service Chain || | | | |
+=========================================================+
|*Control/Management-Plane for Service on NS |
+=========================================================+
|IF to Service || O | O | O |Section 5.2.2.1.|
|Manager || | | | |
+---------------++-------+-------+-------+----------------+
|Telemetory || O | O | O |Section 5.2.2.2.|
+---------------++-------+-------+-------+----------------+
M: Mandatry, O: Optional, - : Not Required
Figure 11: List of Requirements for each SLG