One of the most daunting tasks to any cable provider is to optimize the use of our scarcest resource: bandwidth. The difficulties in managing data capacity and throughput – bandwidth in the data world – are further compounded by the lack of tools to help identify who is using it, what it is being used for, when it is being used, and where the ongoing use is requiring us to spend our capex and opex resources.
With Internet protocol detail record (IPDR), we now have the ability to answer those questions and to begin using the data collected to plan for future consumption trends. This article explores the reasons to use IPDR, what a typical deployment may look like, the back-office service management enabled by the data, and ways to use this information real-time to make intelligent decisions to better manage our available bandwidth.
From network to service Traditionally, simple network management protocol (SNMP) provided most visibility into the cable edge network. Though SNMP supports a rich network management model, its range of capability is limited in terms of scalability and reliability. As the name would suggest, SNMP was originally designed to support only network management practices and applications.
The evolution of DOCSIS technology has introduced dynamic quality of service (QoS)-based offerings that expanded the edge management model from network-oriented to service-oriented. Furthermore, with new cable provider technologies and services such as wireless, tru2way, and advanced advertising, the relationship between service and network is no longer singular and static. Instead, there is a growing separation between service and network as providers define complex new services delivered over multiple network technologies. (See Figure 1.) With service management’s role within the provider’s delivery network growing, the need exists for a more detailed, efficient and reliable data collection mechanism enabling visibility into service behavior and the overall subscriber experience.
IPDR beings with it a number of new service definitions and protocol features that provide high-resolution network, service and subscriber visibility. Intelligent management of services through IPDR provides enhanced management models that include:
• Metered billing by service and subscriber usage
• Network capacity analysis and planning
• Edge resource management
• Network inventory and asset management
• Service-level objective (SLO) monitoring for business services
• Quality of experience (QOE) monitoring
• Service optimization and product development analytics Data collection Though IPDR does not promise to replace all forms of traditional polling and network management tools based on SNMP, it does provide unique capabilities for gathering detailed per-subscriber, per-service flow information. For this reason, it is particularly well-suited for collection of per cable modem device information stored in the cable modem termination system (CMTS), such as subscriber traffic usage.
With SNMP, an operator desiring per modem state or counter information would be required to generate hundreds of consecutive requests to the CMTS for specific information on a periodic basis. With IPDR, the CMTS automatically streams records to the collector on a configurable periodic basis at a minimum of every 15 minutes.
The IPDR protocol introduces a number of key benefits that enhance the operator’s ability to collect data from the edge network.
• High availability: There are features built directly into the protocol that enable automated fail-over in the event of connection or server failure. This allows operators to build redundant paths for valuable records streams.
• Reliability: IPDR uses transmission control protocol (TCP) to provide connection-oriented transport reliability. In addition, record acknowledgement on the application layer provides further robustness.
• Scalability: The stream-oriented behavior of the IPDR protocol provides a new way for data to be gathered from the network. The exporter "pushes" event-based records to the collection layer, removing the inefficiencies and cost of polling.
• Efficiency: IPDR implements messages using a binary representation. This results in very compact data records that occupy a minimal amount of network capacity while minimizing the expense of encoding and decoding. Three use cases Capacity analysis and service modeling: As service providers consider the introduction of a new product or service tier, or anticipate the growth of a popular offering, they must carefully assess what capacity resources will be required. The timing and location of capacity upgrades to support existing services, coupled with anticipated demand for capacity for new service, creates a complex capacity management challenge.
In the process of capacity analysis, the service provider relies on an assortment of tools and models to support "what if" scenarios in order to forecast demand and associated capex and opex. Typically, this process is manual and labor-intensive, requiring significant time from expert network engineers to integrate usage data from disparate sources.
In addition, inputs fed to the manual capacity models are based on static assumptions and offline network data that often limit the capability of the model.
IPDR enables new tools for the capacity analysis and planning of services by providing detailed modem usage information on a per service flow and service tier level of granularity.
DOCSIS defines subscriber account management interface specification (SAMIS), a service definition designed to provide subscriber and service.
Within each SAMIS IPDR stream, the following modem usage information is included on a semi-real time basis:
• Bi-directional octet and packet counters (both upstream/downstream)
• Service context providing the service class name (SCN) in a string defined by the provider; for example, "VOIP-UPSTREAM" or "DATA-TIER-LIGHT"
• Topological context including the CMTS interfaces to which the modem is associated and customer premises equipment (CPE) devices that are attached
• Policy enforcement statistics for the number of delayed or dropped packets
Identifying resource congestion: One of the challenges to network growth is determining when a resource is in a congested state and making decisions to increase capacity, while at the same time avoiding over-provisioning of resources.
Before we can make decisions on how to avoid congestion, we need to agree on when a resource is congested. DOCSIS now defines QoS attributes ServiceSlaDropPkts and ServiceSlaDelayPkts, from which we can infer the point in time that the network load results in a decrease in the useful work done by the network, thereby causing packet drops and delays.
Alternatively, a percent utilization could be the measured congestion point. In that case, the summation of service flows that are identified with a resource via the CmtsMdIfName or CmtsMdIfIndex attribute can be compared to a specific value, for example 70 percent for downstream or 60 percent for upstream. The amount of time above those thresholds may then be considered congested periods where customers may be impacted.
Global resource management: In today’s world, resources are permanently allocated through careful planning and modeling. A best guess at future utilization is made based on trends, past performance and extrapolated future consumption.
With IPDR, we can now have a standards-based method for representing the usage patterns of service flow utilization. With the per-flow data collected and stored in repositories that can collate this information based on selected criteria, it is now possible to manage those edge resources in a quasi-real-time manner by allocating them to the services that need them most.
An example would be the allocation of downstreams during peak business hours to data edge devices rather than to underutilized residential video channels. At peak residential video times, the same resources are taken back from the data edge device where they now are at low utilization.
With hooks from a global session resource manager (GSRM), which maintains state information on the edge resources and tracks using via IPDR, and a session resource manager (SRM) making requests for the resources, we can now use a standard mechanism to track usage patterns and allocate our most precious resources to the services that can make the most efficient use of them. Just the beginning As IPDR becomes the ubiquitous protocol for service management in the edge network, the opportunity will emerge to expand its role to other service offerings. With the multi-play network evolution bringing together disparate service offerings, such as data, multimedia, voice, video and gaming, there is a greater need for a singular view into the allocation of resources. IPDR fulfills this central role by providing the framework needed by different services to describe their unique view of consumption in universal terms.
IPDR has the potential to unify bandwidth, policy, capacity and resource decisions by providing a core functionality that will allow a common language to be spoken by the multi-play network. This will allow for the universal edge resource manager (ERM) to manage the shared resources across all of a provider’s offerings. (See Figure 2.) Conclusions Resource over-provisioning to provide specific service level agreements is costly to the service provider. As resource consumption is stochastic both in quantity and time, it is only through over-provisioning that we can provide customers with reasonable assurance that their expectations will be met.
By performing semi-real-time analysis on service flow usage data, we will be able to make predictions on where consumption will be occurring and use admission control techniques to provide the QoE the customer has been committed to receive. IPDR service-flow level consumption data will feed into the analysis engine to allow the dynamic resource evaluation and reallocation needed to reduce the large capex and opex spends historically made to provide the highest QoE possible from existing resources.
As we move forward from the world of network management to service management, it’s clear that IPDR will play an increasingly critical role in the service provider’s toolkit.
Jeff Finkelstein is senior DOCSIS engineer at Cox Communications. Reach him at Jeff.Finklestein@cox.com. Jason Schnitzer is founder and principal of Applied Broadband. Reach him at Jason@appliedbroadband.com. Sidebar: Additional Resources • IPDR: www.ipdr.org ("IPDR/SP Protocol Specification Version 2.3", March 2, 2007); see also TM Forum, www.tmforum.org
• SNMP: http://www.ietf.org/html.charters/OLD/snmpv3-charter.html
• DOCSIS: www.cablemodem.com
• Tru2way: www.tru2way.com