NETCONF T. Zhou Internet-Draft G. Zheng Intended status: Standards Track Huawei Expires: May 15, 2018 E. Voit Cisco Systems A. Clemm Huawei A. Bierman YumaWorks November 11, 2017 Subscription to Multiple Stream Originators draft-zhou-netconf-multi-stream-originators-01 Abstract This document describes the distributed data collection mechanism that allows multiple data streams to be managed using a single subscription. Specifically, multiple data streams are pushed directly to the collector without passing through a broker for internal consolidation. 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 in RFC 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 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 May 15, 2018. Zhou, et al. Expires May 15, 2018 [Page 1] Internet-Draft Multiple Steam Originators November 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 (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. Table of Contents 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 2 2. Use Cases . . . . . . . . . . . . . . . . . . . . . . . . . . 3 2.1. Use Case 1: Data Collection from Devices with Main-board and Line-cards . . . . . . . . . . . . . . . . . . . . . 3 2.2. Use Case 2: IoT Data Collection . . . . . . . . . . . . . 4 3. Solution Overview . . . . . . . . . . . . . . . . . . . . . . 5 4. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 7 5. Security Considerations . . . . . . . . . . . . . . . . . . . 7 6. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 7 7. References . . . . . . . . . . . . . . . . . . . . . . . . . 7 7.1. Normative References . . . . . . . . . . . . . . . . . . 7 7.2. Informative References . . . . . . . . . . . . . . . . . 8 Appendix A. Change Log . . . . . . . . . . . . . . . . . . . . . 8 Appendix B. Subscription Decomposition . . . . . . . . . . . . . 8 Appendix C. Publication Composition . . . . . . . . . . . . . . 10 Appendix D. Subscription Management . . . . . . . . . . . . . . 10 Appendix E. Notifications on Subscription State Changes . . . . 11 Appendix F. Configured Subscription and Call Home . . . . . . . 11 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 11 1. Introduction Streaming telemetry refers to sending a continuous stream of operational data from a device to a remote receiver. This provides an ability to monitor a network from remote and to provide network analytics. Devices generate telemetry data and push that data to a collector for further analysis. By streaming the data, much better performance, finer-grained sampling, monitoring accuracy, and bandwidth utilization can be achieved than with polling-based alternatives. Zhou, et al. Expires May 15, 2018 [Page 2] Internet-Draft Multiple Steam Originators November 2017 YANG-Push [I-D.ietf-netconf-yang-push] defines a transport- independent subscription mechanism for datastore updates, in which a subscriber can subscribe to a stream of datastore updates from a server, or update provider. The current design involves subscription to a single push server. This conceptually centralized model encounters efficiency limitations in cases where the data sources are themselves distributed, such as line cards in a piece of network equipment. In such cases, it will be a lot more efficient to have each data source (e.g., each line card) originate its own stream of updates, rather than requiring updates to be tunneled through a central server where they are combined. What is needed is a distributed mechanism that allows to directly push multiple individual data substreams, without needing to first pass them through an additional processing stage for internal consolidation, but still allowing those substreams to be managed and controlled via a single subscription. This document will describe such distributed data collection mechanism and how it can work by extending existing YANG-Push mechanism. The proposal is general enough to fit many scenarios. 2. Use Cases 2.1. Use Case 1: Data Collection from Devices with Main-board and Line- cards For data collection from devices with main-board and line-cards, existing YANG-Push solutions consider only one push server typically reside in the main board. As shown in the following figure, data are collected from line cards and aggregate to the main board as one consolidated stream. So the main board can easily become the performance bottle-neck. The optimization is to apply the distributed data collection mechanism which can directly push data from line cards to a collector. On one hand, this will reduce the cost of scarce compute and memory resources on the main board for data processing and assembling. On the other hand, distributed data push can off-load the streaming traffic to multiple interfaces. Zhou, et al. Expires May 15, 2018 [Page 3] Internet-Draft Multiple Steam Originators November 2017 +-------------------------------------+ | collector | +------^-----------^-----------^------+ | | | | | | +-------------------------------------+ | | | | | | | +-----+------+ | | | | | main board | | | | | +--^-----^---+ | | | | | | | | | | +---+ +---+ | | | | | | | | | +----+----+---+ +---+----+----+ | | | line card 1 | | line card 2 | | | +-------------+ +-------------+ | | device | +-------------------------------------+ Fig. 1 Data Collection from Devices with Main-board and Line-cards 2.2. Use Case 2: IoT Data Collection In the IoT data collection scenario, as shown in the following figure, collector usually cannot access to IoT nodes directly, but is isolated by the border router. So the collector subscribes data from the border router, and let the border router to disassemble the subscription to corresponding IoT nodes. The border router is typically the traffic convergence point. It's intuitive to treat the border router as a broker assembling the data collected from the IoT nodes and forwarding to the collector[I-D.ietf-core-coap-pubsub]. However, the border router is not so powerful on data assembling as a network device. It's more efficient for the collector, which may be a server or even a cluster, to assemble the subscribed data if possible. In this case, push servers that reside in IoT nodes can stream data to the collector directly while traffic only passes through the border router. Zhou, et al. Expires May 15, 2018 [Page 4] Internet-Draft Multiple Steam Originators November 2017 +-------------------------------+ | collector | +---^-----------^------------^--+ | | | | | | | | | | +-------+--------+ | | | border router | | | +----^------^----+ | | | | | | | | | | +---+ +---+ | | | | | +---+----+---+ +---+----+---+ | IoT node 1 | | IoT node 2 | +------------+ +------------+ Fig. 2 IoT Data Collection 3. Solution Overview All the use cases described in the previous section are very similar on the data subscription and publication mode, hence can be abstracted to the following generic distributed data collection framework, as shown in the following figure. A Collector usually includes two components, o the Subscriber generates the subscription instructions to express what and how the collector want to receive the data; o the Receiver is the target for the data publication. For one subscription, there may be one to many receivers. And the subscriber does not necessarily share the same address with receivers. In this framework, the stream originators have the Master role and the Agent role. Both the Master and the Agent include two components, o the Subscription Server manages capabilities that it can provide to the subscriber. o the Publisher pushes data to the receiver according to the subscription information. Zhou, et al. Expires May 15, 2018 [Page 5] Internet-Draft Multiple Steam Originators November 2017 The Master knows all the capabilities that the attached Agents and itself can provide, and exposes the Global Capability to the Collector. The Collector cannot see the Agents directly, so it will only send the Global Subscription information to the Master. The Master disassembles the Global Subscription to multiple Component Subscriptions, each involving data from a separate telemetry source. The Component Subscriptions are then distributed to the corresponding Agents. When data streaming, the Publisher located in each stream originator collects and encapsulates the packets per the Component Subscription, and pushes the piece of data which it can serve directly to the designated data Collector. The Collector is able to assemble many pieces of data associated with one Global Subscription, and can also deduce the missing pieces of data. +-------------------------------------+ | Collector | | +------------+ +------------+ | | | Subscriber | | Receiver <-------+ | +-^----+-----+ +------^-----+ | | | | | | | | +-------------------------------------+ | Global | |Global | push | Capability | |Subscription | | +-------------------------------------+ | | | | Master | | | | +--+----v------+ +------+------+ | | | | Subscription | | Publisher | | | | | Server | | | | | | +--^----+------+ +-------------+ | | | | | | | +-------------------------------------+ | Component | | Component push | Capability | | Subscription | +-------------------------------------+ | | | | Agent | | | +--+----v------+ +-------------+ | | | | Component | | Publisher | | | | | Subscription | | +------+ | | Server | +-------------+ | | +--------------+ | +-------------------------------------+ Fig. 3 The Generic Distributed Data Collection Framework Master and Agents may interact with each other in several ways: Zhou, et al. Expires May 15, 2018 [Page 6] Internet-Draft Multiple Steam Originators November 2017 o Agents need to have a registration or announcement handshake with the Master, so the Master is aware of them and of life-cycle events (such as Agent appearing and disappearing). o The Master relays the component subscriptions to the Agents. o The Agents indicate status of Component Subscriptions to the Master. The status of the overall subscription is maintained by the Master. The Master is also responsible for notifying the subscriber in case of any problems of Component Subscriptions. Any technical mechanisms or protocols used for the coordination of operational information between Master and Agent is out-of-scope of the solution. We will need to instrument the results of this coordination on the Master Node. Note: Some preliminary considerations on the solution details are now listed in the appendix for reference. The detailed solution need to be discussed and will be added if the WG accepts the problem statement. 4. IANA Considerations This document makes no request of IANA. Note to RFC Editor: this section may be removed on publication as an RFC. 5. Security Considerations It's expected to reuse the existing secure transport layer protocols, such as TLS [RFC5246] and DTLS [RFC6347], to secure the telemetry stream. The Collector cannot access the Agent directly but to negotiate the security parameters with the Master. However the data streams are actually generated by the Agents which are invisible to the Collector. So mechanisms may need to consider when adapting secure transport layer protocols here. the detailed solution is TBD. 6. Acknowledgements 7. References 7.1. Normative References [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, DOI 10.17487/RFC2119, March 1997, . Zhou, et al. Expires May 15, 2018 [Page 7] Internet-Draft Multiple Steam Originators November 2017 [RFC5246] Dierks, T. and E. Rescorla, "The Transport Layer Security (TLS) Protocol Version 1.2", RFC 5246, DOI 10.17487/RFC5246, August 2008, . [RFC6347] Rescorla, E. and N. Modadugu, "Datagram Transport Layer Security Version 1.2", RFC 6347, DOI 10.17487/RFC6347, January 2012, . [RFC7950] Bjorklund, M., Ed., "The YANG 1.1 Data Modeling Language", RFC 7950, DOI 10.17487/RFC7950, August 2016, . 7.2. Informative References [I-D.ietf-core-coap-pubsub] Koster, M., Keranen, A., and J. Jimenez, "Publish- Subscribe Broker for the Constrained Application Protocol (CoAP)", draft-ietf-core-coap-pubsub-02 (work in progress), July 2017. [I-D.ietf-netconf-yang-push] Clemm, A., Voit, E., Prieto, A., Tripathy, A., Nilsen- Nygaard, E., Bierman, A., and B. Lengyel, "YANG Datastore Subscription", draft-ietf-netconf-yang-push-11 (work in progress), October 2017. Appendix A. Change Log (To be removed by RFC editor prior to publication) v01 o Minor revision on Subscription Decomposition o Revised terminologies o Removed most implementation related text o Place holder of two sections: Subscription Management, and Notifications on Subscription State Changes Appendix B. Subscription Decomposition Since Agents are invisible to the Collector, the Collector can only subscribe to the Master. This requires the Master to: Zhou, et al. Expires May 15, 2018 [Page 8] Internet-Draft Multiple Steam Originators November 2017 1. expose the Global Capability that can be served by multiple stream originators; 2. disassemble the Global Subscription to multiple Component Subscriptions, and distribute them to the corresponding telemetry sources; 3. notify on changes between portions of a subscription moving between different Agents over time. To achieve the above requirements, the Master need a Global Capability description which is typically the YANG [RFC7950] data model. This global YANG model is provided as the contract between the Master and the Collector. Each Agent associating with the Master owns a local YANG model to describe the Component Capabilities which it can serve as part of the Global Capability. All the Agents need to know the namespace associated with the Master. The Master also need a data structure, typically a Resource-Location Table, to keep track of the mapping between the resource and the corresponding location of the Subscription Server which commits to serve the data.When a Global Subscription request arrives, the Master will firstly extract the filter information from the request. Consequently, according to the Resource-Location Table, the Global Subscription can be disassembled into multiple Component Subscriptions, and the corresponding location can be associated. The decision whether to decompose a Global Subscription into multiple Component Subscriptions rests with the Resource-Location Table. A Master can decide to not decompose a Global Subscription at all and push a single stream to the receiver, because the location information indicates the Global Subscription can be served locally by the Master. Similarly, it can decide to entirely decompose a Global Subscription into multiple Component Subscriptions that each push their own streams, but not from the Master. It can also decide to decompose the Global Subscription into several Component Subscriptions and retain some aspects of the Global Subscription itself, also pushing its own stream. Component Subscriptions belong to the same Global Subscription MUST NOT overlap. The combination of all Component Subscriptions MUST cover the same range of nodes as the Global Subscription. Also, the same subscription settings apply to each Component Subscription, i.e., the same receivers, the same time periods, the same encodings are applied to each Component Subscription per the settings of the Global Subscription. Zhou, et al. Expires May 15, 2018 [Page 9] Internet-Draft Multiple Steam Originators November 2017 Each Component Subscription in effect constitutes a full-fledged subscription, with the following constraints: o Component subscriptions are system-controlled, i.e. managed by the Master Node, not by the subscriber. o Component subscription settings such as time periods, dampening periods, encodings, receivers adopt the settings of their Global Subscription. o The life-cycle of the Component Subscription is tied to the life- cycle of the Global Subscription. Specifically, terminating/ removing the Global Subscription results in termination/removal of Component Subscriptions. o The Component Subscriptions share the same Subscription ID as the Global Subscription. Appendix C. Publication Composition The Publisher collects data and encapsulates the packets per the component subscription. There are several potential encodings, including XML, JSON, CBOR and GPB. The encoding of the data records follows the YANG schema, so that the composition at the Receiver can benefit from the structured and hierarchical data instance. The Collector may be able to assemble many pieces of data associated with one subscription, and can also deduce the missing pieces of data. The Collector recognizes data records associated with one subscription according the Subscription ID. Data records generated per one subscription are assigned with the same Subscription ID. For the time series data stream, records are produced periodically from each stream originator. The message arrival time varies because of the distributed nature of the publication. The receiver assembles data generated at the same time period based on the recording time consisted in each data record. In this case, time synchronization is required for all the steam originators. Appendix D. Subscription Management A Global Subscription can be rejected for multiple reasons. Some are related to the Subscription Decomposition and Component Subscription. New error codes are defined to indicate why a datastore subscription attempt has failed. The subscription result with the failure reason is returned as part of the RPC response. Zhou, et al. Expires May 15, 2018 [Page 10] Internet-Draft Multiple Steam Originators November 2017 Appendix E. Notifications on Subscription State Changes Each component subscription maintains its own subscription state and is responsible for sending its own OAM notifications (for example, when the component subscription is suspended or when it can resume). TBD. Appendix F. Configured Subscription and Call Home TBD. Only about the message layer which is transport independent. Authors' Addresses Tianran Zhou Huawei 156 Beiqing Rd., Haidian District Beijing China Email: zhoutianran@huawei.com Guangying Zheng Huawei 101 Yu-Hua-Tai Software Road Nanjing, Jiangsu China Email: zhengguangying@huawei.com Eric Voit Cisco Systems United States of America Email: evoit@cisco.com Alexander Clemm Huawei 2330 Central Expressway Santa Clara, California United States of America Email: alexander.clemm@huawei.com Zhou, et al. Expires May 15, 2018 [Page 11] Internet-Draft Multiple Steam Originators November 2017 Andy Bierman YumaWorks United States of America Email: andy@yumaworks.com Zhou, et al. Expires May 15, 2018 [Page 12]