White Papers

Strategies for Managing the P2P Phenomenon

Popular, Yet Problematic

Since the inception of peer-to-peer networks in the late 1990s, P2P applications have multiplied and evolved to represent a formidable component of Internet traffic. Most service providers continue to believe that P2P traffic will constitute roughly 60% of network traffic over the next two years. Furthermore, there is a strong possibility that increased IP video content will drive this number even higher, especially when Vudu set-top boxes and video applications like Joost TV begin delivering high-definition content to large audiences with P2P technology.

P2P technology enables the sharing of computer resources and services, including information, files, processing cycles, and storage by means of direct exchanges between participants in a P2P network. As a result, P2P networking is a very efficient and resilient method of distributing content over IP networks. However, its mass popularity and bandwidth-hungry nature (composed of file-sharing traffic, much of which is of dubious legality) are threatening to exhaust service providers' network resources. Furthermore, P2P applications pay little attention to geographical locations when they peer with each other and distribute content over an ever-growing number of hops, a process that uses up network resources and drives up service providers' operating costs.

P2P applications are among the most downloaded types of software on the Internet and, in this consumer-driven world, the popularity of P2P clients changes frequently. As of this writing, some of the more popular P2P applications include BitTorrent, Limewire, KaZaA, Gnutella, WinMX, and eDonkey.

The threat of lawsuits by recording and movie industries may have halted the rise of some P2P networks, but users have simply switched to new and better networks, thus continuing to fuel the growth of P2P. The Yankee Group estimates that there are currently about 58.2 million P2P users worldwide, a number estimated to grow to 135.6 million by the year 2010.



It all started with Napster, which was mainly used to download popular music. Now Internet users worldwide are demanding higher-bandwidth applications such as HDTV shows, movies, and software that dominate the P2P traffic mix. According to CacheLogic, video accounts for more than 60% of global P2P traffic, while music files generate just 11%.



Voice over IP (VoIP) is also enjoying rapid growth, aided by the popularity of applications such as Skype, which uses P2P technology to enable users make phone calls over the Internet using regular PCs.

The bottom line is that P2P will continue to be a key driver of the growth in high-speed broadband services. If a service provider does not meet the bandwidth demands of its P2P users, they will switch to another provider that does. However, high-speed broadband access already is a commodity, which implies that the average revenue per user (ARPU) is already low. Meanwhile, service providers want to roll out revenue-generating services such as IPTV, video on demand, and voice over IP over their IP infrastructures. As a result, there is an urgent need to handle traffic congestion issues caused by P2P in an intelligent manner and to ensure that quality of service is optimized for all users and services on the network.

The First Step: Identify Users and Traffic

In order to implement intelligent approaches to P2P traffic management and deliver seamless performance to their entire customer base, service providers must be able to identify P2P users and traffic. This is where P2P becomes the biggest thorn in the side of service providers. It is just not easy to identify P2P traffic.



Conventionally, Internet applications transported via TCP have been assigned a specific TCP port, making the traffic easy to identify. First-generation P2P applications were designed to use well-defined port numbers. These inflexible configurations made it easy for service providers to monitor specific applications and use routers and content switches to perform traffic shaping or to use blocking techniques to ensure QoS and manage bandwidth usage. However, the current generations of P2P applications have the ability to disguise their existence through the use of arbitrary and dynamic port numbers. Often referred to as "port hopping," this capability, combined with the application's ability to encrypt payload content, makes it extremely difficult to categorize P2P traffic. These camouflage methods also render traditional port-based P2P blocking and shaping techniques ineffective.

Another anomaly is the symmetrical nature of P2P applications, which defy service providers' basic assumptions regarding their subscribers' usage profile, especially in terms of upload/download traffic ratio and times of usage through the day. Most of the residential IP network is used for applications like e-mail and web browsing, which generate a larger amount of downstream traffic for a single upstream request. By contrast, P2P applications are constantly uploading and downloading, even when subscribers are asleep. The constant uploads cause a shift in the upstream/downstream ratio, resulting in congestion on the upstream link because of a larger number of subscribers using the upstream link than anticipated.

Application-Aware Devices

In response to the P2P conundrum, network equipment manufactures are developing a new breed of application-aware devices. These devices can inspect deep inside data packets to determine whether network traffic is carrying P2P applications. Commonly known as deep packet inspection (DPI), this technique looks within the application payload of a packet or traffic stream and makes decisions based on its content. DPI will identify and classify traffic, by comparing a packet payload signature to a library of signatures; this permits finer control than what's available with classifications based just on header information.

DPI enables service providers to gain much better visibility into how subscribers are using broadband, thus helping to determine what kind of new services should be offered in the future. Better visibility into the network also translates into improved performance, troubleshooting, and security monitoring.

Proposed Solutions for Managing P2P

There are many options to effectively manage P2P traffic. However, each option needs to be weighed against the costs involved in implementing these solutions and the need to optimize subscribers' quality of experience.

Acquire More Bandwidth

When network resources are constantly overwhelmed, the most obvious approach is to acquire more bandwidth or upgrade the existing infrastructure to handle the increased load. However, this solution has a very limited life span since P2P tends to expand to fill the available bandwidth. This approach also involves costly capital expenditures.

Block P2P traffic

This implies blocking, at the network access point, TCP ports that are commonly used by popular P2P applications. The aim is to reduce bandwidth usage by striking all P2P traffic entering the network.

However, it is not easy to block P2P traffic since it is able to masquerade as non-P2P traffic. Many P2P applications enable users to select a desired port or assign ports dynamically, with the sole intent of circumventing standard P2P blocking practices. Even if technical difficulties can be overcome with the new generation of application-aware devices, there is a bigger concern for service providers: maintaining customer satisfaction for the consumer group that is fueling the growth in high-speed broadband services. Blocking all P2P traffic will most certainly lead to customer dissatisfaction and aggravate subscriber churn.

Implement Bandwidth Caps

Some service providers have tried to impose usage caps to curb P2P subscriber behavior. When a subscriber exceeds a certain level of aggregated bandwidth activity over a period of time, the provider may charge the subscriber extra for the broadband service or give the subscriber's traffic a lower priority.

Capping bandwidth also enables service providers to introduce tiered pricing schemes. By charging different prices for each service tier, determined by usage patterns, service providers are able to recoup some of the costs incurred by the heavy P2P traffic.

Although bandwidth caps discourage most subscribes from using more than the allocated cap, they do not do enough to reduce service providers' bandwidth costs. In addition, upstream traffic, which is very prevalent in P2P, remains untouched and continues to congest the network.

Perhaps the most important objection to bandwidth caps is that subscriber frustration with this approach is likely to aggravate customer dissatisfaction, especially if other service providers do not implement similar caps. This is, in turn, could impact the long-term profitability of a provider that implements caps.

Utilize Network Caching

Caching P2P content enables service providers to maintain a repository of the most frequently downloaded P2P files in a local network. By confining traffic to a local network, this solution minimizes some of the downstream bandwidth and transit cost associated with P2P traffic. It also has helps maintain a high quality of experience and service satisfaction.

Despite its seeming advantages, caching is fraught with legal landmines. Service providers would no longer be providing just the infrastructure but, potentially, free copyright-infringed content as well. Without caching service, providers are not responsible for the nature of traffic passing through their networks. If they use caches, however, they are technically storing content that could push them into violation of copyrights.

Another issue with caching is that it does not address the upstream traffic congestion issue. P2P users from other networks still consume bandwidth when they download files from another service provider's subscribers.

Shape P2P Traffic

Traffic shaping provides a mechanism for controlling the volume of traffic being sent into a network (bandwidth throttling) and the rate at which the traffic is being sent (rate limiting). It simply buffers and prioritizes traffic traversing the network. This permits a service provider to give priority to revenue-generating traffic, leaving whatever bandwidth is left over for P2P. Each individual data packet that arrives at a network access point is examined and classified, based on an identification key found in the packet. Based on the priority of each category of traffic, the packets are then entered into a queue and transmitted. In the context of P2P shaping, P2P packets are sent last over whatever bandwidth is left after all the higher priority traffic has been sent.

The main advantage of traffic shaping is that service providers can gain a degree of control over their networks and can prioritize traffic to suit their subscriber base. In turn, associated P2P costs can be reduced in a way that avoids the drawbacks of completely blocking P2P traffic.

Since shaping relies on how accurately network managers can identify packets as P2P, this technique is susceptible to a range of evasion tactics. The P2P development community has shown itself to be very resistant to shaping techniques in the past, and has developed several tactics for hiding the true identity of packets. Offsets of the packet identifier or signature can be varied, the content can be encrypted to mask its true nature, and traffic signatures across multiple connections or packets can be concatenated (that is, the identity of a particular packet can be determined only by comparing it with several other packets/connections).

Even if traffic shapers could identify P2P packets accurately, the disadvantage is that it adversely affects the experience of all subscribers on the network - not just P2P users. Shaping requires that each data packet be inspected in order to be classified as P2P. This introduces significant network latency and slower processing times for all traffic, since each packet is queued, inspected, and then queued again for transmission. Shaping therefore indirectly penalizes non-P2P users by introducing undue latency into real-time services such as VoIP, IPTV, and VoD.

Implement Stateful, Policy-Based QoS Management

As service providers strive to deliver competitive multiplay services over a single converged infrastructure, stateful policy-based QoS management becomes paramount in optimizing subscribers' user experiences for all services.

Policy-based QoS delivery at the application layer is another critical advantage provided by application-aware forwarding devices. With their deep packet inspection capabilities, they can manage and prioritize hundreds of thousands, if not millions, of individual service traffic flows.

This "stateful inspection" is crucial in order to intelligently classify and route traffic because it provides an overall understanding of the significance of traffic content traversing two end-points. Packets are also deeply inspected and classified by matching payload signatures to a signature library that identifies the nature of packet flow.

Stateful QoS policy management does not depend on limiting subscriber usage, but rather on intelligently rerouting P2P traffic via cost-effective paths, thereby delivering substantial cost savings. Network congestion at the network access point is also minimized, freeing up bandwidth for other revenue-generating subscriber services.



Stateful QoS policing can be extended to deliver quality based on subscriber or service awareness or a combination of the two. In the first approach, subscriber awareness, forwarding decisions are made by classifying traffic flows from a particular subscriber. This ensures that an end user receives the bandwidth and quality of service (QoS) promised by the provider. In the second approach, service awareness, forwarding decisions are based on traffic flows from a particular service type. This ensures that the appropriate QoS is applied to each service. The third approach - hierarchical QoS - is a hybrid of the other two. Hierarchical QoS applies policing to flows based on service type at one stage and at a second stage applies shaping to a subscriber's bundle of services.

By intelligently inspecting traffic flows, service providers gain an understanding of what kinds of traffic are traversing their networks. Once P2P traffic has been identified, providers can apply application- or subscriber-specific polices, thereby managing the cost of bandwidth consumption, transit costs, and most importantly, subscribers' performance expectations. Stateful QoS policing has the potential to reduce costs and, at the same time, optimize subscribers' quality of experience.

Predeployment Testing and Validation Ensure Intelligent Traffic Management

It is clear that using application-aware devices with deep packet inspection capability is a long-term, viable solution that will help service providers manage and mitigate the P2P phenomenon.

Matching every packet byte against thousands of pattern characters at wire speed is a computationally intensive task. Combined with sophisticated bandwidth management and QoS schemes, the packet-forwarding mechanism of application-aware devices is much more complex. Therefore, comprehensive predeployment testing and validation becomes crucial to ensure that these solutions will help providers intelligently manage P2P traffic.

A key element that service providers need to consider as they evaluate and compare possible solutions is the application-aware device's ability to prioritize and deliver optimum quality of service for profit-generating, real-time services such as IPTV and VoIP in the presence of bandwidth-hogging P2P applications.

Test Methodology

Generate a Realistic Mix of Traffic to Exercise Per-flow QoS Policy Management, Validate QoE

To fully exercise and validate the application-aware device's ability to classify and prioritize traffic flows, service providers need to be able to emulate realistic stateful traffic that characterizes real traffic. Realistic emulated traffic must include a mix of services, including VoIP, IPTV, VoD, HTTP, and P2P application traffic with the same characteristics as real or expected customer traffic. A realistic traffic mix is essential to ensure that a QoS management system is forwarding real-time VoIP and video traffic with low latency and, when network bandwidth limits are reached, that it is rate-limiting P2P traffic.

Ensure Scalability of QoS Policy Management

Service providers must measure application performance and verify the capacity and stability of application-aware devices using real service traffic that emulates expected and future numbers of users. Scalability testing is the single most effective way to reduce costs in the process of planning, estimating capacity, and designing networks. To accurately measure user capacity, the characteristics of emulated traffic such as P2P file size distributions, VoIP call duration profiles, IPTV channel-changing profiles, and the proportion of traffic from each application service must closely match those of real or expected customer traffic.

Manage the Proprietary Nature of P2P Protocols

Unlike traditional application-layer protocols such HTTP and SIP, which are specified in IETF standards, many P2P applications are proprietary and therefore not covered by any open standard. With no open standard and the constant evolution of some P2P application protocols, P2P traffic is very difficult to simulate.

This is a challenge for both manufacturers and service providers, especially during design and evaluation testing. A testbed containing thousands of real P2P clients is expensive and difficult to create and maintain. As a result, manufacturers and service providers have to depend on testing tools that combine advanced stateful P2P application-replay capabilities and an extended library of P2P signatures that mimic P2P protocol behavior at scale.

P2P Requires Intelligent Approaches

Service providers are eager to deploy intelligent approaches to manage the ever-growing popularity of P2P. The impact of P2P applications on network traffic patterns, capacity planning, and infrastructure upgrades is significant. Service providers must find ways to manage high-bandwidth P2P traffic while maintaining customer satisfaction as well as the overall subscriber user experience for all services - especially revenue-generating IPTV and voice services.

Application-aware devices with deep packet inspection capabilities, coupled with advanced stateful QoS policy management systems, are positioned to deliver viable solutions to deal with the P2P phenomenon. Successful deployment of these application-aware devices and the introduction of new, tiered subscriber policies require comprehensive predeployment validation. It is essential for service providers to validate the functional and scalable performance of these devices with real-world application traffic that simultaneously emulates subscriber services such as IPTV, VoIP, web browsing, email, and P2P up/downloads. Providers need to characterize service or application performance, subscribers' quality of experience for each service, the capacity of their networks to provide those services, and device scalability. All these parameters will have a direct influence on subscriber satisfaction and network CAPEX costs.

[ back | top of page | back to white papers ]