Peer-to-peer traffic – friend or foe?

Most everyone is aware of the havoc peer-to-peer traffic can cause on a service provider’s network.  But many may not be aware that some solutions now exist to control this traffic. 

The first question that needs to be reviewed is, do you want to control this traffic?  There are two divergent schools of thought on this. Many service providers see P2P traffic as a way to charge for additional value-added services. Others still believe they want to block P2P traffic from their network and turn off these accounts.  The latter group might find some new opportunities for service offers embedded in this article.

The problem

Let’s start with a little bit of background to make sure everyone is on the same page.  As P2P traffic continues to grow, service provider’s network usage patterns are changing and the current school of thought related to provisioning is no longer sufficient.  This growth in traffic causes network congestion, performance deterioration and ultimately customer dissatisfaction.  These items combined may also cause customer churn. 

Traditionally, information is exchanged via the Internet through a client/server model.  With P2P exchange of data, the creation of decentralized groups allows for information to flow over the public Internet in an anonymous logical fashion. The individual users of these applications are shielded via this anonymity.  In addition to the issues created due to copyright infringement via this form of information exchange, the service providers no longer have the ability to forecast network capacity based on historical subscriber usage patterns. There are four key areas where service providers are feeling the pinch:

1) Upstream/downstream traffic is flipped where the upstream traffic is much larger then the downstream traffic.  This results in network congestion on the upstream link that was never planned for with initial broadband deployments.

2) Time of day usage statistics no longer apply.  Previously, service providers could assume peak usage at certain times of the day and lower usage at other times.  With P2P applications, the computers are often left to transfer data throughout the day in an unattended fashion.

3) Previously, peering traffic always traversed the Internet to another location.  In today’s world, two home users can form a direction connection.

4) Over-subscription assumptions no longer apply.  A handful of power users can “hog” all of the bandwidth deployed for a much larger usage base.

Recent studies have identified that 60% of all Internet traffic is driven by P2P applications.  In order for service providers to continue to provide appropriate performance to their entire customer base at a profitable price point, some changes will need to occur.  First they must be able to identify the P2P users and restrain them.  Second, they must be able to identify the users driving this traffic in order to charge them appropriately.  Based on this knowledge, new subscription plans can be put in place to ensure all users have the appropriate bandwidth available to them.

The solution

One solution to the service provider’s headaches listed above are products from a company called P-Cube.  P-Cube’s Service Control Platform is a dedicated network device that provides the tools for service providers to detect and control P2P application usage and resulting bandwidth consumption.  This platform maintains the state of every network conversation while executing detailed inspections of every data packet up to the application layer (Layer 7).  This is done by detecting P2P application signatures that are usually witnessed during the initial message exchange between two network hosts.  It then classifies all traffic for that conversation as P2P.  Some examples of how P-Cube can help a service provider control this problem are as follows:

1) Aggregated Rate Limiting:  Limits all P2P traffic to a certain percentage of the available bandwidth.

2) Upstream Control:  Limits upstream P2P traffic while allowing downstream traffic to continue uninterrupted.

3) Destination Based Classification:  Limits traffic on expensive or congested links.

4) Time of Day Policies:  Provides different limits on P2P usage during different times of the day/week.

5) Subscriber Application Quota:  Enforces a byte cap (quota) per day and then throttles or completely blocks traffic.

6) Subscriber Dynamic Policies:  Dynamic policies to allow users to control their own accounts.

Real world use

SingTel has implemented the P-Cube solution in order to resolve their network issues driven by P2P traffic.  They have been offering Internet access for more then 10 years, and broadband Internet access for the last three.  This service allows for up to 1.5M bit/sec.  They currently have about 100,000 subscribers and are adding roughly 7,000 per month.   By monitoring their outgoing traffic, they found usage to be at levels they would have projected for 200,000 users.

During the last year they have had many subscriber complaints as to unsatisfactory bandwidth and performance, and many customers have churned.  They had to add many costly leased lines to their infrastructure, and have seen international transit traffic soar.  Through the use of P-Cube’s product they found that 60% of their traffic was P2P.  Additionally, they found that only 5% of the users accounted for 70% of the P2P traffic.

SingTel also implemented new services to help control this problem.  They limited P2P traffic from international transit links.  By using regional bandwidth links, internal costs were reduced.  They allowed for unlimited P2P during off-peak hours, but limited during peak.  This shifted the bandwidth burden to times-of-day when casual users wouldn’t be affected.  Lastly, they implemented a quota to enforce the usage.  Once a certain amount of bandwidth was consumed, a quota was implemented. In total, the first year savings are projected to be between $500,000 and $1,000,000 – certainly not something to sneeze at.

For those of you who subscribe to the every cloud has a silver lining theory, P2P traffic may allow for new subscriber services, which can increase your revenue base and allow you to reduce your operational costs… something worth thinking about.