Network Working Group Z. Li
Internet-Draft Huawei Technologies
Intended status: Informational Y. Zheng
Expires: September 14, 2017 China Unicom
J. Zhang
S. Xu
Huawei Technologies
March 13, 2017
An Architecture of Network Artificial Intelligence(NAI)
draft-li-opsawg-network-ai-arch-00
Abstract
Artificial intelligence is an important technical trend in the
industry. With the development of network, it is necessary to
introduce artificial intelligence technology to achieve self-
adjustment, self- optimization, self-recovery of the network through
collection of huge data of network state and machine learning. This
draft defines the architecture of Network Artificial Intelligence
(NAI), including the key components and the key protocol extension
requirements.
Requirements Language
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 inRFC 2119 [RFC2119]
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). Note that other groups may also distribute
working documents as Internet-Drafts. The list of current Internet-
Drafts is at http://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 14, 2017.
Li, et al. Expires September 14, 2017 [Page 1]
Internet-Draft An Architecture of NAI March 2017
Copyright Notice
Copyright (c) 2017 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
(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 . . . . . . . . . . . . . . . . . . . . . . . . 2
2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 3
3. Architecture . . . . . . . . . . . . . . . . . . . . . . . . 3
4. Process . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
5. Classification . . . . . . . . . . . . . . . . . . . . . . . 5
6. Requirement of Protocol Extensions . . . . . . . . . . . . . 5
6.1. Requirement of Southbound Protocols . . . . . . . . . . . 5
6.2. Requirement of Data Collection . . . . . . . . . . . . . 6
6.3. Requirement of Devices . . . . . . . . . . . . . . . . . 6
6.4. Requirement of Northbound Interface . . . . . . . . . . . 6
7. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 6
8. Security Considerations . . . . . . . . . . . . . . . . . . . 7
9. Normative References . . . . . . . . . . . . . . . . . . . . 7
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 7
1. Introduction
Artificial Intelligence is an important technical trend in the
industry. The two key aspects of Artificial Intelligence are
perception and cognition. Artificial Intelligence has evolved from
an early non-learning expert system to a learning-capable machine
learning era. In recent years, the rapid development of the deep
learning branch based on the neural network and the maturity of the
big data technology and software distributed architecture make the
Artificial Intelligence in many fields (such as transportation,
medical treatment, education, etc.) have been applied. With the
development of network, it is necessary to introduce artificial
intelligence technology to achieve self-adjustment, self-
optimization, self-recovery of the network through collection of huge
data of network state and machine learning. The areas of machine
learning which are easier to be used in the network field may
Li, et al. Expires September 14, 2017 [Page 2]
Internet-Draft An Architecture of NAI March 2017
include: root cause analysis of network failures, network traffic
prediction, traffic adjustment and optimization, security defense,
security auditing, etc., to implement network perception and
cognition.
This draft defines the architecture of Network Artificial
Intelligence (NAI), including the key components and the key protocol
extension requirements.
2. Terminology
AI: Artificial Intelligence
NAI: Network Artificial Intelligence
3. Architecture
^ ^
(4)| |(4)
+---------------|--------------+ +---------------|--------------+
| Domain 1 | | | | Domain 2 |
| +------------+ | | +------------+ |
| | Central | | | | Central | |
| (1)| Controller |----------------------| Controller |(1) |
| | with | | | | with | |
| | NTA | | | | NTA | |
| +------------+ | | +------------+ |
| / \ | | / \ |
| (3)/ \ | | / \(3) |
| / \ | | / \ |
| +--------+ +--------+ | | +--------+ +--------+ |
| | | | | | | | | | | |
| |Network | ...... |Network | | | |Network | ...... |Network | |
| | Device | (2) | Device | | | | Device | (2) | Device | |
| | 1 | | N | | | | 1 | | N | |
| +--------+ +--------+ | | +--------+ +--------+ |
| | | |
+------------------------------+ +------------------------------+
Figure 1: An Architecture of Network Artificial Intelligence(NAI)
The architecture of Network artificial intelligence includes
following key components:
(1) Central Controller: Centralized controller is the core part of
Network Artificial Intelligence which can be called as 'Network
Li, et al. Expires September 14, 2017 [Page 3]
Internet-Draft An Architecture of NAI March 2017
Brain'. The Network Telemetry and Analytics (NTA) engines can be
introduced acompanying with the central controller. The Network
Telemetry and Analytics (NTA) engine inclues data collector,
analytics framework, data persistence, and NAI applications.
(2) Network Device: IP network operation and maintenance are always a
big challenge since the network can only provide limited state
information. The network states includes but are not limited to
topology, traffic engineering, operation and maintenance information,
network failure information and related information to locate the
network failure. In order to provide these information, the network
must be able to support more OAM mechanisms to acquire more state
information and report to the controller. Then the controller can
get the complete state information of the network which is the base
of Network Artificial Intelligence(NAI).
(3) Southbound Protocol and Models of Controller: As network devices
provide huge network state information, it proposes a number of new
requirements for protocols and models between controllers and network
devices. The traditional southbound protocol such as Netconf and
SNMP can not meet the performance requirements. It is necessary to
introduce some new high-performance protocols to collect network
state data. At the same time, the models of network data should be
completed. Moreover with the introduction of new OAM mechanisms of
network devices, new models of network data should be introduced.
(4) Northbound Model of Controller: The goal of the Network
Artificial Intelligence is to reduce the technical requirements on
the network administrators and release them from the heavy network
management, control, maintenance work. The abstract northbound model
of the controller for different network services should be simple and
easy to be understood.
4. Process
NAI consists of following processes:
-- Data Collection
From the time aspect, data collection can be divided into real-time
data collection and non-real-time collection.
From the content aspect, data collection can be divided into network
information collection (including topology, tunnels, routing,
equipment configuration, etc.) and traffic collection (the collection
network traffic, network load, device KPI, etc.).
-- Data Storage
Li, et al. Expires September 14, 2017 [Page 4]
Internet-Draft An Architecture of NAI March 2017
Store data collected from network. Many existing big data storage
technologies can be used here.
-- Data Processing
This is preliminary data processing too select effective data and
simply analyse data relationship.
-- Analyse
Analyse engine will provide the data analysis results using machine
learning algorithm.
-- Closed Loop Control
According to the results of intelligent analysis and policy set by
user, the centrol controller will implement closed-loop control of
the network.
5. Classification
NAI can be divided into off-line process and on-line process in
accordance to the time aspect of the data collection and analysis.
Off-line process refers to process of the existing data, or non-real-
time collection data. Although the analysis process will also focus
on the relationship between data and time, but it does not require
real-time analysis. Off-line process is mainly used for two
purposes: (1) training or verification of real-time process design;
(2) trouble shooting or reason analysis for events that have already
occurred.
On-line process is efficient real-time collection, processing and
analysis of the data, to operate network monitoring and event
forecasting. The main purpose of the on-line process are: (1)
network capacity monitoring and precise optimizing; (2) network event
prediction and fast trouble shooting; (3) real-time network
optimization according to the policy.
6. Requirement of Protocol Extensions
6.1. Requirement of Southbound Protocols
REQ 01: The southbound protocol of the controller should be
introduced to meet the performance requirements of collecting huge
data of network states.
Li, et al. Expires September 14, 2017 [Page 5]
Internet-Draft An Architecture of NAI March 2017
The soundbound protocol can be based on the extensions of the
existing traditional protocols such link state colloction protocols,
PCEP[RFC5440], BMP[RFC7854], etc. Or the new protocol like
Telemetry[I-D.kumar-rtgwg-grpc-protocol] can be introduced as the
soundbound protocols. The protocol choice will be based on the
application scenarios of NAI.
6.2. Requirement of Data Collection
REQ 02: The data collected from the network devices includes but not
limites to following information:
-- network topology information
-- routing protocol status
-- IP routes and MAC routes
-- LSP information
-- network traffic inforamtion
-- network configuration
-- network device KPIs
-- log of network elements
-- trap of network elements
-- OAM information
6.3. Requirement of Devices
REQ 03: New OAM mechanisms should be introduced for the network
devices in order to acquire more types of network state data.
6.4. Requirement of Northbound Interface
REQ 04: The abstract network-based service models should be provided
by the controller as the northbound models to satisfy the
requirements of different services.
7. IANA Considerations
This document makes no request of IANA.
Li, et al. Expires September 14, 2017 [Page 6]
Internet-Draft An Architecture of NAI March 2017
8. Security Considerations
TBD.
9. Normative References
[I-D.kumar-rtgwg-grpc-protocol]
Kumar, A., Kolhe, J., Ghemawat, S., and L. Ryan, "gRPC
Protocol", draft-kumar-rtgwg-grpc-protocol-00 (work in
progress), July 2016.
[RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
Requirement Levels", BCP 14, RFC 2119,
DOI 10.17487/RFC2119, March 1997,
<http://www.rfc-editor.org/info/rfc2119>.
[RFC5440] Vasseur, JP., Ed. and JL. Le Roux, Ed., "Path Computation
Element (PCE) Communication Protocol (PCEP)", RFC 5440,
DOI 10.17487/RFC5440, March 2009,
<http://www.rfc-editor.org/info/rfc5440>.
[RFC7854] Scudder, J., Ed., Fernando, R., and S. Stuart, "BGP
Monitoring Protocol (BMP)", RFC 7854,
DOI 10.17487/RFC7854, June 2016,
<http://www.rfc-editor.org/info/rfc7854>.
Authors' Addresses
Zhenbin Li
Huawei Technologies
Huawei Bld., No.156 Beiqing Rd.
Beijing 100095
China
Email: lizhenbin@huawei.com
Yi Zheng
China Unicom
No.9, Shouti Nanlu, Haidian District
Beijing 100048
China
Email: zhengyi39@chinaunicom.cn
Li, et al. Expires September 14, 2017 [Page 7]
Internet-Draft An Architecture of NAI March 2017
Jinhui Zhang
Huawei Technologies
Huawei Bld., No.156 Beiqing Rd.
Beijing 100095
China
Email: jason.zhangjinhui@huawei.com
Xu Shiping
Huawei Technologies
Huawei Bld., No.156 Beiqing Rd.
Beijing 100095
P.R. China
Email: xushiping7@huawei.com
Li, et al. Expires September 14, 2017 [Page 8]