- 1 - INTERNET-DRAFT October 1, 1996 IP Over Cable Data Network Service draft-ietf-ipcdn-ipcabledata-spec-00.txt October 1, 1996 Masuma Ahmed mxa@terayon.com Terayon Corporation Guenter Roeck groeck@cisco.com Cisco 1. Status of this Memo This document is an Internet-Draft. 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." To learn the current status of any Internet-Draft, please check the "lid-abstracts.txt" listing contained in the Internet-Drafts Shadow Directories on ds.internic.net (US East Coast), nic.nordu.net (Europe), ftp.isi.edu (US West Coast), or munnari.oz.au (Pacific Rim). 2. Abstract This document describes the application of IP over a cable data network service environment configured as a logical IP subnetwork (LIS). Specifically, this document describes the cable data network interfaces to support IP, IP service features, IP address assignment using Dynamic Host Configuration Protocol (DHCP), Address Resolution Protocol (ARP), and other service-specific issues relating to supporting IP over cable data network service. This document considers only directly connected IP end-stations and the router operating in the conventional LAN based paradigm over a cable data network. As background information, this document also provides an overview of the cable data network - 2 - service, the architecture and the related network interfaces. This document does not specify an Internet Standard of any kind. It is presented for discussion purposes only. 3. Conventions The following language conventions are used in the items of specifications in this document: * MUST, SHALL, or MANDATORY - this item is an absolute requirement of the specification. * SHOULD or RECOMMEND - this item should generally be followed for all but exceptional circumstances. * MAY or OPTIONAL - this item is truly optional and may be followed or ignored according to the needs of the implementor. 4. Introduction The goal of this specification is to allow compatible and interoperable implementations for transmitting IP packets over cable data network service. This memo defines only the operation of IP over cable data network service and is not meant to describe the operation of cable data network service. Note that the cable data network service described in this document is referred to as high speed cable data service (HSCDS) in the Request For Proposals [1] (RFP) issued by CableLabs. The cable data network service is a public carrier service. Therefore, supporting IP over a public carrier service has issues such as security, scalability, fairness, charging based on service tiers, traffic management and should be dealt with appropriately. This document tries to address some of these issues. In this document, the cable data network is defined as an end-to-end network consisting of three overlaid networks; IP routed network, a data link layer subnetwork and the physical HFC access networks. A functional diagram of the end-to-end cable data network is shown in Figure 1. In this configuration, IP packet data service is supported using the - 3 - packet data bearer service capabilities of the data link layer cable sub-network which in turn is supported using the physical transmission medium and the Medium Access Control (MAC) protocol of the physical layer in the Hybrid Fiber Coax (HFC) access networks. Note that the data link layer subnetwork is a routable network unlike a bridged network and may support functions such as Address Resolution Protocol (ARP) filtering. ________ IP Routed Network |Router|--------------------------------------------|PC| |______| Data Link Layer Subnetwork |headend|----------------------------|modem| |equip. | | | HFC Access Network ---------------------------- Note: The Data link layer subnetwork shown here is a routable network and is not a bridged network. Figure 1: End-to-end Cable Data Network The rest of the document details the support of IP, Address Resolution Protocol (ARP), and IP address assignment over cable data network service. As background information, a brief overview of the cable data network service along with the cable data network architecture is provided in Appendices A and B. 5. Cable Data Network Architecture As mentioned earlier, a cable data network consists of three overlaid networks; IP overlaid network, data link layer subnetwork and HFC access network. This section describes the requirements associated with the IP network over the cable data network service only. Specifications of the data link layer and the physical layer networks are beyond the scope of this document. As background information, a brief description of the physical and data link layer networks is provided in Appendix B. - 4 - 5.1 IP Routed Network The end-to-end cable data network MUST provide the internetworking capabilities by using IP as the network layer protocol technology. A router MUST be used to provide the layer 3 connectivity between different customer equipment and the wide area network. Therefore, the interface provided by the router to the customer equipment MUST be a network layer interface and the data transferred MUST be a routable protocol which may be routed to the backbone network belonging to the same carrier network or to the Wide Area Network (WAN). The router MUST provide all internetworking between customer equipment (e.g., PCs) attached to the cable data modems and between cable modem users and the WAN. 6. Cable Data Network Interfaces The network components of the cable data network and the related interfaces are shown in Figure 2. We followed the conventions as much as possible with a few exceptions used in the HSCDS RFP issued by CableLabs to name the network elements and the associated interfaces of the cable data network. The network components of the cable data network include: * cable data modem (CDM) and PC/WorkStation at the subscriber premise * cable data modem termination system (CDMTS) at the distribution hub or headend * router, Dynamic Host Configuration Protocol (DHCP) server and local web servers at the headend In the sections below, the router interfaces to support IP over cable data network are described. Other relevant interfaces of the end-to-end cable data network are also briefly described. - 5 - Headend or Distribution Hub |------------------------| |________ --------- | ||Local | |Manage-| | ||WWW | |ment | | ||Server| |System | | |---^---- -^----^-- | | | | | | | | |--|----|---| | | | ---I/M1 ---I/M2| | | | | |___v___v_ |-v---|| I/F3 |-------| I/F2 |---|I/F1 ____ I/F7 ||Router |<---|-->|CDMTS|<---|-->|HFC |<--|-->|CDM|<-|->|PC| To <--|-->|_^_____| I/F4 |-----|| |Access | |---| |__| WAN | | | | |Network| | | | |<----------------|----|----------------------------->| | ---I/F5 | I/F6 | | | ||-v----| | ||DHCP | | ||Server| | ||______| | -------------------------- I/F: Network Interface I/M: Management Interface Figure 2: Cable Data Network Interfaces 6.1 IF/1 Interface The I/F1 is the interface between the CDM and the PC at the subscriber premise. The I/F1 interface supports native Ethernet and IEEE 802.3 Medium Access Control (MAC) protocols over 10Base-T physical interface. The I/F1 interface carries transparently the higher layer protocols (e.g., IP) above the data link layer protocol to the PC (or workstation). The specification of this interface is beyond the scope of this document. - 6 - 6.2 I/F2 Interface The I/F2 is the interface between the CDM and the HFC access network. The I/F2 supports an RF digital transmission interface between the CDM and the HFC access network and performs upstream RF channel signal modulation and downstream RF channel signal demodulation functions. In addition, I/F2 supports a data link layer interface to the HFC network providing network access control and data delivery functions. The specification of this interface is beyond the scope of this document. 6.3 I/F3 Interface The I/F3 is the interface between CDMTS and the HFC access network. The I/F3 interface performs almost the same protocol functions as the I/F2 interface with a few exceptions. The I/F3 interface at the CDMTS is used to control and manage a number of CDMs in the HFC access networks. Therefore, one of the primary functions of the I/F3 interface is to manage and control the usage of upstream and downstream RF channel resources by the subscriber modems. Also, at the physical level, the following differences exist between the I/F2 and I/F3 interfaces: - upstream and downstream channel frequencies (e.g., I/F3 upstream and downstream frequencies are opposite to those at the I/F2) - receive and transmit power levels In addition, it is possible that the I/F3 may aggregate more than one fiber nodes and as such the I/F3 interface may have different Bit Error Rate (BER) and Signal to Noise Ratio (SNR) than the I/F2 interface. The specification of this interface is beyond the scope of this document. 6.4 I/F4 Interface The I/F4 is the interface between the CDMTS and the router located at the headend or the distribution hub. Separation of the router and the CDMTS may be an implementation issue and as such the I/F4 interface is vendor implementation specific. Therefore, the specification of I/F4 interface is - 7 - beyond the scope of this document. 6.5 /F5 Interface The I/F5 is the interface between the router and the IP address server which in this case is the Dynamic Host Configuration Protocol (DHCP) server. The I/F5 interface is a traditional IP routed network from the headend router to the DHCP server(s). As the data transmitted across this network is native IP, the choice of LAN and WAN media is extremely flexible. It is possible that the router or the CDMTS itself may contain the DHCP server functions and thus the I/F5 interface may support a proprietary interface depending on a specific vendor's implementation. Therefore, the specification of I/F5 interface is beyond the scope of this document. 6.6 I/F6 Interface The I/F6 is the IP interface between the router located at the headend or distribution hub and the PC located at the subscriber premise. The I/F6 interface MUST support the IP network layer interface between the router located at the distribution hub/headend and the PC (or workstation) located at the subscriber premise. This interface MUST support dynamic assignment of network layer address, i.e., the IP address to the PC on PC power up using DHCP [4]. This interface is described in detail in Section 8 below. 6.7 I/F7 Interface The I/F7 is the Wide Area Network (WAN) interface between the router and the public backbone network. This interface supports all of the required standard WAN interfaces supported in a public carrier network. Specification of the I/F7 interface is beyond the scope of this document. 7. IP Service Features The types of IP service features that may be supported over cable data network service include: - 8 - * Guaranteed and best effort IP service delivery (e.g., by using RSVP and Integrated services protocol) * Packet/protocol filtering (e.g., packet access, filtering, forwarding, and control) * Subscription based service provisioning (e.g., access to the IP service via a service order process) * Dynamic and static configuration of IP addresses to subscriber's end systems (using DHCP) * Different tiers of IP service (e.g., using IP access list) * IP multicast service 8. Logical IP Subnetwork Configuration In the Logical IP Subnetwork (LIS) configuration, each separate administrative entity configures its hosts and routers within a closed logical IP subnetwork. Each cable data network can be considered to be under one administrative entity, i.e., under the jurisdiction of one cable data network service provider. The cable data network can be configured as a single or multiple IP subnetworks depending on the geographic span and physical architecture of the cable data network configuration and the number of hosts supported in the network. In general, the router in the cable data network MUST support at least one subnetwork configuration (referred to as `router LIS configuration'). The hosts within the same subscriber premise MUST have direct access to the other hosts belonging to the same host subnet configuration but MUST not have direct access to the other cable data network service hosts supported in the same router LIS. All hosts within the same host LIS MUST have the same IP network/subnet number and address mask, i.e., all of the IP devices on each of the Ethernet interfaces of the subscriber CDMs MUST be on the same IP router subnet. Depending on the cable data network service requirements, it is RECOMMENDED that the router providing LIS functionality over the cable data network service be able to support more than one LIS. Therefore, the router SHOULD be configured as a member of one or more LISs. All members within a router LIS MUST have the same IP network/subnet number and address mask. - 9 - As mentioned in Appendix A, RF channels are used as the physical transmission medium in the HFC access networks to support cable data network service. In addition, separate RF channels at different RF frequency spectrum are used for upstream and downstream transmission. Also, depending on the CATV network lay-out, two-way CATV data transmission may be supported using a single downstream RF channel and multiple upstream RF channels. For the purpose of this document, the downstream RF channel and the associated upstream RF channels used for two-way data transmission are considered as a single two-way RF transmission entity. Depending on the span of the cable data network and the number of hosts supported per RF transmission entity, a router LIS MUST be configured to support all hosts connected to a single or multiple RF transmission entities. The router providing interconnection of differing LISs MUST be able to support multiple sets of parameters (one set for each connected LIS) and be able to associate each set of parameters to specific IP network/subnet number. The router MUST be able to provide multiple LISs support with a single physical I/F4 interface between itself and the CDMTS. Similarly, a router MUST be able to support a single LIS that spans over multiple CDMTSs. Also, the router MUST be able to provide a single LIS support to more than one RF transmission entities with a single physical I/F4 interface between itself and the CDMTS. Note that, as mentioned earlier, the router and the CDMTS functions may be combined into a single entity. In such a case, the I/F4 related requirements described here do not apply. Hosts that are not within the same subscriber premise but within the same IP router subnet as well as of different IP router subnets MUST communicate via the IP router. Therefore, the hosts within the same router LIS MUST not have direct access to each other. The router MUST support sending IP packets to any and all hosts within the same router LIS as well as of differing router LISs but the hosts within the router LIS MUST send packets to the router only. Since it is expected that only a small amount of the cable data network service traffic will be from one host to another, this will not cause excessive relay traffic, but does have significant impact on the IP subnet model. 8.1 Address Resolution Protocol The hosts and router had the same subnet mask for the large router subnet and the hosts that happened to talk to many other hosts on the same router subnet may be required to support very large (e.g., 10,000 entries) Address Resolution Protocol (ARP) tables. Therefore, the router MUST view a - 10 - single or multiple RF transmission entities in the cable data network as one subnet (e.g., 1,000 to 10,000 hosts). Normally, ARP [5] is used between hosts and the router, and between hosts. ARP used in the cable data network for each of these cases is described below. * Router to Host To avoid scaling and security problems with use of ARP over a large IP router subnet (e.g., 1,000 to 10,000 hosts), the router MUST not ARP for the MAC address of the host. Instead, the router MUST assume that DHCP is used by the IP hosts. In the process of relaying the DHCP requests between the hosts to the DHCP server, the router MUST capture the MAC address of the host and the host's IP address assigned by the server. The router MUST bind this information together into its ARP table. The entry in the ARP table MUST be flagged to prevent it from aging out normally. Unicast ARP MAY be used to validate the entry and refresh it. * Host to Router The DHCP MUST communicate the default IP gateway address to the host. Through configuration in the DHCP server, the IP address of the router MUST be supplied to the host. The host MUST issue a normal ARP for the IP address of the router. The subscriber CDM MUST encapsulate this packet to send it upstream. The router MUST answer this ARP normally. * Host to Host Hosts ARPing other hosts attached to the same I/F1 interface MUST not leave the I/F1 interface. However, for hosts ARPing other hosts within the router LIS, the router MUST use the proxy ARP capability to answer these ARP requests. 8.1.1 ICMP Data from one host to another on the same router subnet MUST be sent via the router. When two hosts are on the same subnet, the router would normally send an ICMP Redirect to inform the first host that a better (in this case, direct) path exists. However, since the cable media does not support direct host to host communications within the same router subnet, the router MUST do the forwarding and MUST suppress the ICMP messages. - 11 - 8.2 IP Address Assignment A host attached to the CDM at the subscriber premise MUST use DHCP to obtain its configuration and IP address. The router MUST participate in all DHCP exchanges between the host and the DHCP server. For example, upon power-up, the host may broadcast a DHCP message on its local Ethernet segment. The host may optionally include any host configuration parameters that it may need. The subscriber modem transmits this packet upstream to the router. Upon receiving the packet, the router adds its IP address to the gateway IP address field in the DHCP packet and may forward the packet to one or more DHCP servers. The DHCP servers send DHCP packets to the router with each packet containing offered IP addresses available for use which the router forwards to the host. The host selects an offered IP address and sends back a DHCP request message for a lease on that address to the router which forwards the packet to the DHCP server. The DHCP server sends an acknowledgement indicating a successful lease of the address. The router adds an ARP entry, binding the IP address to the Ethernet MAC address of the host and forwards the DHCP acknowledgement to the host. 8.2.1 IP Broadcast Address It is RECOMMENDED that the router and the hosts within the IP subnet of the cable data network be able to receive and transmit IP packets with any of the four standard IP broadcast addresses as specified in RFC1122 [6]. Members upon receiving an IP broadcast or IP subnet broadcast packets for their LIS, MAY process the packet as if addressed to that station. However, depending on the cable data network service requirements, the router SHOULD have the capability to suppress packets received with broadcast IP address. 8.2.2 IP Multicast Address The IP multicasting method specified in RFC1112 [7] requires a Network Service Interface which provides a multicast-like ability to provide dynamic access to the local network service interface operations: - JoinLocalGroup (group-address) - LeaveLocalGroup (group-address) Security, subscription and subscriber billing related implications associated with dynamic subscription and removal from group address lists of any host in a router IP subnetwork require further study. Also, methods to support - 12 - IP multicasting over data link layer protocol of the cable data network service require further study and will be addressed in the future. 8.3 IP Service Tiers Cable data network service providers may support different tiers of IP service using different charging schemes. Depending on the service tier subscribed to, a host can have access to different servers and application services such as premium web pages, guaranteed bit rate packet, multicast, etc. Different tiers of IP service MAY be supported using the IP access list. By arranging the IP address assigned to fall into one of several ranges, the number of access lists required may be reduced to a very small number. The router MAY support such capability by modifying the DHCP Address Assignment packet to include the subscriber's cable modem ID in the DHCP `client identifier' field. Note that the subscriber's hosts MUST not know the cable modem ID. This will be done transparently to them. 8.4 Security The IP security issues such as supporting authenticated end-to-end IP transmission, e.g., using data encryption are beyond the scope of this document. 9. Issues Issues associated with cable data network service configurations to support capabilities such as IP multicasting, IP tunneling and Virtual Private Network (VPN) configuration include: - procedures for performing routing updates between the headend router and the modem router (in this case, the modem at the subscriber premise supports routing functions) - ability to create virtual private IP routed network - filtering of IP packets from outgoing routing protocol updates - 13 - 10. Acknowledgements Special thanks to Jim Forster and Dennis Picker for their valuable suggestions and critical review of the document. In addition, the author would like to thank Amir Furhman and Steve Lin for helpful discussions on the topic. 11. Appendix A: CATV Data Network Service Examples of CATV data network service capabilities include: *packet data delivery to subscriber cable data modem (CDM) with minimum peak bit rate of 500 kbps in the downstream direction. The maximum peak bit rate can be up to 40 Mbps. *packet data delivery to subscriber cable data modem (CDM) with maximum peak bit rate of 10 Mbps in the upstream direction. The minimum peak bit rate can be as low as 28 kbps. Various implementations of cable data network service supporting a number of data link layer protocols are available today. Most of these implementations support data link layer protocol for the cable data network service using slot and frame approach in both upstream and downstream directions. In the HFC access network, the downstream direction is described as the transmission of data flow from the network to the subscriber and the upstream direction is described as the transmission of data flow from the subscriber to the network. In the downstream direction, usually broadcast mode is used to distribute traffic to the subscribers from the cable headend equipment. In the upstream direction, the network resources are shared and subscribers have to contend for it. As an upstream resource arbiter, the cable headend equipment allocates and manages upstream bandwidth to the subscribers using data link layer bandwidth management algorithm. Radio Frequency (RF) channels in the upstream and downstream directions over HFC access networks are used as the physical medium to transport the cable data network service. Various combinations of the modulation techniques are used for digital transmission of the cable data network service over - 14 - analog transmission medium of the HFC access networks. Examples of different modulation techniques include: * Spread spectrum modulation technique such as Direct Sequence Spread Spectrum * Quaternary Phase Shift Keying (QPSK) technique * Quadrature Amplitude Modulation Technique (QAM) with modulation order of 16, 64, and/or 256 *Orthogonal Frequency Division Multiplexing (OFDM) technique The RF channels are configured to run between the cable modem at the subscriber premise and the channel controller at the headend. Upstream channel is shared among all the subscribers in the HFC networks and various physical layer access algorithms in addition to data link layer bandwidth management algorithms are used to access the upstream resources. One or a combination of the following physical layer access algorithms is used to support cable data network service in the upstream direction. * Synchronous Code Division Multiple Access (S-CDMA) method * Time Division Multiple Access (TDMA) method *Frequency Division Multiple Access (FDMA) method 12. Appendix B: Cable Data Network Architecture and Interfaces The physical and data link layer portion of the cable data network architecture is described below. B1. HFC Access Network The physical HFC access network is a a shared-media, tree and branch architecture with analog transmission over fiber used for trunks and coaxial cable used for accessing the end systems. The majority of the existing HFC access networks support sub-split systems where the upstream frequency spectrum is supported from 5 to 30 MHz (and 42 MHz in the upgraded systems) and the downstream frequency spectrum is - 15 - from 50 to 550 MHz (and 750 MHz in the upgraded systems). There are also systems that support mid-split (5 to 108 MHz in the upstream direction, and 162 MHz and above in the downstream direction) and high-split (5 to 174 MHz in the upstream direction, and 243 MHz and above in the downstream direction) systems, however, these systems are primarily used in institutional networks. A physical lay-out of the HFC access network is illustrated in Figure 3. As shown, a typical HFC access network consists of fiber nodes and cascaded amplifiers with remote distribution hubs centrally controlled from a central cable headend system. Depending on network configurations, a single headend in the cable data network can support from 40 to 200 or larger number of fiber nodes and each fiber node can support from 500 to 2000 or even larger number of households. - 16 - To other<--//---| DH or --//--->||<----SONET Ring ________ |<------> HE || (digital) | | |Co-axial || |------|Fiber |----|Distribut(500/ || (analog) | | | |-ion 2000 _____||______ Fiber Optics| |--->|Node | |<------> homes | |<-----//----| | |______| passed) |Distribution|------//------| |Hub (DH) | |<----------> |or | ________ | |Head End | Fiber Optics | |---|<-Co-axial (500/ |(HE) |<-------//-------|Fiber | Distribution 2000 |____________|------//-------->|Node |---|<---------->homes || |______| |<---------->passed) || 20,000/100,000 To other <--//--|| homes passed 40 to 200 DH or --//--->| Fiber Nodes HE Figure 3: An Example HFC Access Network RF channels usually 6 MHz wide are used to transport analog services such as NTSC video, and digital services such as cable data network service, in the HFC access networks. An RF channel is the physical layer parameter of the HFC access network that extends from the physical layer interface of the cable data modem (CDM) located at the subscriber premise to the cable data modem termination system (CDMTS) located at the headend or distribution hub. Separate RF channels in different frequency spectrum are used for upstream and downstream transmission. Distribution hubs are remotely located from the headend and are configured to support one or more fiber nodes. These remote hubs are interconnected back to a centralized headend via digital transmission medium such as SONET ring. 13. Terminology In this document, the following terminology is used consistent with the Cablelabs HSCDS RFP. * CDM is the cable data modem at the subscriber premise. * CDMTS is the cable data modem termination system at the headend or distribution hub. - 17 - * Customer equipment is the equipment at the subscriber premise such as a PC or workstation. * HE is the cable head end. * DHE is the Distribution Hub Equipment. * Carrier equipment is the equipment such as CDM, CDMTS, HE that belongs to the public carrier network. * I/F refers to the network interface in the CATV data network. * I/M refers to the management interface in the CATV data network. 14. Authors' Addresses Masuma Ahmed Terayon Corporation 2952 Bunker Hill Lane Santa Clara, CA 95054 Phone: (408) 486-5207 Fax: (408) 727-6205 Email: mxa@terayon.com Guenter Roeck Cisco 174 Tasman Drive Santa Clara, CA 95054 Phone: (408) 527-3143 Fax: (408) 727-6205 Email: groeck@cisco.com - 18 - References 1. "High Speed Cable Data Service Request for Proposals", Cable Television Laboratories, April 1995. 4. Droms, R., "Dynamic Host Configuration Protocol", RFC1531, Bucknell University, October 1993. 5. Plummer, D., "An Ethernet Address Resolution Protocol - or - Converting Network Addresses to 48 bit Ethernet Address for Transmission on Ethernet Hardware", STD 37, RFC826, MIT, November 1982. 6. Deering, S., "Requirements for Internet Hosts - Communication Layers", STD 3, RFC1122, USC/Information Sciences Institute, October 1992. 7. Deering, S., "Host Extensions for IP Multicasting", STD 5, RFC1112, Stanford University, August 1989.