In a controversial, incredibly strange, yet kind of catchy video released last year, Megaupload claimed to have 1 billion users, accounting for 4% of Internet traffic.

Our Global Internet Phenomena Report, which has followed the growth of the service in recent years, confirms that Megaupload was the most popular Storage and Back-Up Service in almost every region.

The table below shows the combined usage of Megaupload and MegaVideo on fixed access networks in several regions from our Fall 2011 study:

What will happen now that Megaupload has been taken down? The most likely scenario is that users will switch to other Storage and Back-Up Services and video sites, causing new category leaders to emerge.  Currently the second most popular Storage and Back-up Service varies by region, but some of the top contenders to replace Megaupload are Rapidshare, zShare, Hotfile, and MediaFire.

As always, we will observe and report on the continued evolution of Internet traffic in future Global Internet Phenomena reports.

I’ll refer to congestion occurring closest to the subscriber as the “weak link”.  Not surprisingly, the most effective (in terms of relieving the congestion) and most fair (to subscribers) way to manage this congestion is by implementing corrective policies that are localized at the site of the congestion itself. In fact, this precise approach is what the FCC defines as “narrowly tailored”.  Referring to Figure 1 below, if congestion occurs at the weak link (see A*) location and affects only users 1-5, then correcting that instance of congestion is ideal because Internet user traffic in other network locations such as locations B and C need not be touched.

Sandvine manages link congestion by examining, identifying, and prioritizing time-sensitive gaming, VoIP and streaming applications, over non time-sensitive applications such as emailing and browsing, or, alternatively, prioritising the users who are causing the least congestion vs. those who are causing the most in a short time interval.  You and I don’t notice if we get an email 2 or 3 seconds later, but we will be upset if our voice calls are garbled and we can’t understand each other. 

Managing congestion to the edge of the network, as close to the affected subscribers as possible, is a competitive differentiator for Sandvine because our competitors only help ISPs manage aggregate traffic patterns.  This aggregate approach is based on the flawed assumption that the macroscopic view (Figure 2) of Internet traffic activity is true for points A, B and C of the networks.  At Sandvine, we manage traffic to the edge of the network (Figure 3) and monitor points A, B and C individually.  This yields a more efficient network: higher utilisation, happier consumers, vs. the alternative approach of reducing capacity artificially (throttling).

This approach means that our congestion management techniques are precisely targeted and extremely effective – users in uncongested parts of the network are not impacted, and users in the congested regions are impacted as minimally and fairly as is technically possible.  The result is the preservation of quality of experience for the masses: maximum quality of experience for the maximum number of subscribers for the maximum amount of time

While it seems obvious that congestion management techniques should only apply where and when there is congestion, competitive approaches do not subscribe to this consideration.  “Capacity Control” techniques that are marketed as congestion management solutions simply drop packets.  For instance, a capacity control solution deployed near the top of the pyramid in Figure 3, might simply drop 20% of all packets of a certain type (e.g. limit bulk traffic to 20Mbps on a 100Mbps link).  On average, this approach delivers the benefit of reducing traffic on congested links by 20% at the cost of impacting every subscriber on the network by dropping 20% of the traffic on all the other links, whether or not they were congested in the first place.  This is approach is neither fair (since it impacts everyone) nor reasonable (since it impacts areas that aren’t congested). A link which had 150% of demand will not be uncongested in this model: it will now be @ 130%. Similarly, a link which was previously @ 80% of demand vs. capacity will now be @ 60%: a net loss of efficiency.

Compounding the confusion surrounding congestion management is the argument that another way to avoid congestion is through the use of usage management principles, like hard or soft monthly quotas (caps).  In this scenario, when a subscriber exceeds the monthly quota, the carrier might impose overage charges in the belief that the subscriber will then self-monitor and reduce data consumption.  The theory goes that if a carrier imposes monthly quotas on all subscribers, then the cumulative power of self-management will limit instances of congestion.  A related method is to lower the speed of users after some volume is achieved in a month. These principles are often disguised as ‘fair use policies’ (mimicking Sandvine’s Fairshare traffic management in name only).

Unfortunately, this approach is undone by the reality that instantaneous contributors to congestion are no more likely to be the network’s monthly heavier users (those who are exceeding their caps) than they are to be the other 99% of subscribers, so congestion remains.  Additionally, a monthly quota across the entire network is not narrowly-tailored to the links that are congested at the times they are congested. Why should a user be slowed down on a link which isn’t busy? The quota applies 24x7x365 on all links equally.

Usage management and traffic management each have their time and place. If your objective is to differentiate the value of service plans in order to maximise your revenue (a commercial objective), use usage management and volume limits. If your objective is a technical one, that of alleviating the effect of congestion on some links at some times in your network, use traffic management. Your users will thank you in both cases.

Sandvine’s network policy control solutions focus on protecting and improving the quality of experience on the Internet. Our award-winning network equipment and software helps DSL, FTTx, cable, fixed wireless and mobile operators better understand network traffic, manage network congestion, create new services and revenues, mitigate traffic that is malicious or undesirable to subscribers, deliver QoS-prioritized multimedia services and increase subscriber satisfaction.

Our Global Internet Phenomena Report is an example of the type of data that our network equipment enables ISPs to collect.  Many of our customers and many in the industry, find these insights valuable, as they allow for proactive capacity planning of Internet networks, enabling fair use connectivity and a high quality experience for the end subscriber. Download it and read for yourself. Feel free to contact us directly with any comments or concerns, we would be more than happy to discuss.

We’ve long been fans of transparency and are public industry advocates for government proposals that seek open and equitable Internet access.  It’s all there for you to read.  No need to bring a bucket.

The most credible reports, such as that from The Telegraph explain that the shut down involved the withdrawal of more than 3,500 Border Gateway Protocol (BGP) routes, used by networks to establish routing between one another. When a border router is turned off or removed, its routes are automatically withdrawn, and the networks beyond it become unreachable. Egypt, although a large country, is served by a small number of border routers.

Other reports, such as this Huffington Post story, appear to be misinformed. This article blamed a solution from a U.S.-based company, Narus, whose products incorporate deep packet inspection (DPI) technology. The article then went on to call for Congress to pass rules on the use of DPI. ⁽¹⁾ It also named a variety of companies, including Sandvine, that include DPI in their solution sets.

As Sandvine has described in its submissions to the Privacy Commissioner of Canada, and other global legislative bodies, the proper focus of any rules should not be on any technology, but rather on applications of technologies. For example, the same Single Lens Reflex (SLR) camera technology used to take photos at a family birthday parties is also used in surveillance cameras that may invade and undermine individual privacy rights. Should rules be created for SLR technology? Or should they focus on the protection of privacy?

Sandvine’s network policy control solutions comprise applications to:

• Manage network congestion
• Report on traffic trends for capacity and business planning
• Mitigate malicious traffic
• Measure network performance
• Create innovative new service tiers

All of these solutions are designed to improve the subscriber’s Internet experience. None of these solutions involve censorship.

In fact, DPI is a key technology used in a very wide variety of consumer and network equipment.  The wireless router in one’s home uses DPI to ensure maximum quality of experience for time-sensitive applications like VoIP or gaming, and also allows multiple devices in a home to share a single Internet connection.

Firewalls, many built right into popular PC operating systems, use DPI technology to analyze packets and identify viruses, trojans, and spam. Libraries, schools and government institutions rely on firewalls to protect themselves and their users from attacks.

Load balancers and routers, indispensible hardware that distribute traffic on the Internet and private networks, use DPI to identify where a given packet or URL should be routed and what priority it should be given.

In reality, there is very little difference between the network infrastructure in countries where citizens enjoy the free and open exchange of information and those that unreasonably restrict such rights.  The difference lies not in technology, but in the principles and policies that are implemented.

Footnote:

1.  On a related note, legislation has been proposed in the US (Bill S.3480 – Protecting Cyberspace as a National Asset Act of 2010, a.k.a. the Internet “Kill Switch” bill),  that provides the President with the ability to control systems that constitute critical cyber-infrastructure.

Notably, for Sandvine, the Order appears to preserve an environment in which deployment of Sandvine’s Network Analytics, Fairshare, Usage Management and Quota Management policy applications can continue to thrive.  Other significant aspects of the Order are highlighted in the sections below.  Sandvine continues to provide solutions to service providers globally, in support of their compliance with their respective regulatory and legislative environments.

The Rules in Brief

The basic rules require:

1. Transparency of traffic management practices, performance characteristics and the commercial terms of service. Mobile broadband providers must also disclose third-party device and application limits and any relevant criteria for use of such third-party offerings.
2. No blocking of lawful content, applications, services, or non-harmful devices. A less onerous no blocking rule applies to mobile broadband providers, which may not block access to lawful websites and may not block applications that compete with their own video or voice telephony services.
3. No unreasonable discrimination such as to favour one’s own application offerings over competitive offerings. The FCC acknowledged that some discrimination is beneficial. They also noted that paid prioritization is likely to be unreasonable, but will be subject to case-specific review. This rule does not apply to mobile providers.
4. All of the rules are generally subject to Reasonable Network Management.

Relevance to Sandvine’s Solution Portfolios

Business Intelligence

While the Order does not address Business Intelligence directly, gathering usage data about a broadband Internet access network would seem to be a necessary first step in effecting Network Policy Control and assessing any reasonable network management practice. Without this data there would be no way to determine the reasonableness of any network management technique. It certainly appears that the gathering of network data is not precluded in any way by the Order, and, to the contrary, may be a necessary pre-requisite to assessing compliance with the Order.

Traffic Optimization

Not surprisingly given the history of the debate, the Order deems an application-agnostic approach, such as the one that Sandvine introduced to the market with its Fairshare solution, as ‘likely reasonable’. Paragraph 73 of the Order states:

“Differential treatment of traffic that does not discriminate among specific uses of the network or classes of uses is likely reasonable. For example, during periods of congestion a broadband provider could provide more bandwidth to subscribers that have used the network less over some preceding period of time than to heavier users. “

The Order does not take a position on whether application-specific Traffic Optimization techniques are inherently reasonable or not, presumably leaving the issue open to case-by-case analysis. The Order seems to anticipate the deployment of such techniques, as in the Transparency discussion (paragraph 56) the Order requires service providers to disclose:

“…whether and why the provider blocks or rate-controls specific protocols or protocol ports, modifies protocol fields in ways not prescribed by the protocol standard, or otherwise inhibits or favors certain applications or classes of applications.”

The Order is clear that content, applications, services and non-harmful devices cannot be impaired or degraded to render them effectively unusable, unless doing so can be deemed necessary for reasonable network management.

Service Creation

The Order also seems to allow the introduction of tiered or differentiated services offerings, such as those supported by Sandvine’s Quota Management and Usage Management solutions. For example, the Order states (paragraph 72) that service tiers based on subscriber usage levels are reasonable:

“Some commenters suggest that open Internet protections would prohibit broadband providers from offering their subscribers different tiers of service or from charging their subscribers based on bandwidth consumed. We are, of course, always concerned about anti-consumer or anticompetitive practices, and we remain so here. However, prohibiting tiered or usage-based pricing and requiring all subscribers to pay the same amount for broadband service, regardless of the performance or usage of the service, would force lighter end users of the network to subsidize heavier end users. It would also foreclose practices that may appropriately align incentives to encourage efficient use of networks.  The framework we adopt today does not prevent broadband providers from asking subscribers who use the network less to pay less, and subscribers who use the network more to pay more.”

The Order further states that broadband offerings which give users choice of their own traffic priorities (paragraph 71) are reasonable:

“Thus, enabling end users to choose among different broadband offerings based on such factors as assured data rates and reliability, or to select quality-of-service enhancements on their own connections for traffic of their choosing, would be unlikely to violate the no unreasonable discrimination rule, provided the broadband provider’s offerings were fully disclosed and were not harmful to competition or end users. We recognize that there is not a binary distinction between end-user controlled and broadband-provider controlled practices, but rather a spectrum of practices ranging from more end-user controlled to more broadband provider-controlled. And we do not suggest that practices controlled entirely by broadband providers are by definition unreasonable.”

Operations Management

The Order explicitly recognizes the ability of service providers to take action with respect to harmful network traffic and to utilize solutions such as the Network Protection solution offered by Sandvine (paragraph 88):

“Broadband providers may implement reasonable practices to ensure network security and integrity, including by addressing traffic that is harmful to the network.”

In a footnote that describes traffic that is harmful to the network, the Order identifies spam, botnets, and distributed denial of service attacks.

Additionally, the Order’s Transparency requirements (paragraph 56) state that certain performance characteristics of service offerings be disclosed, such as the “expected and actual access speed and latency, and the suitability of the service for real-time applications.” Certain applications in Sandvine’s Operations Management suite could help service providers satisfy this requirement.

Reasonable Network Management and Transparency

Reasonable Network Management

The Order defines reasonable network management (paragraph 82):

“A network management practice is reasonable if it is appropriate and tailored to achieving a legitimate network management purpose, taking into account the particular network architecture and technology of the broadband Internet access service.”

So, different techniques may be deployed in different networks, thereby recognizing the architectural differences between fixed and mobile networks, for example. In the same paragraph, the Order further states:

Legitimate network management purposes include: ensuring network security and integrity, including by addressing traffic that is harmful to the network; addressing traffic that is unwanted by end users (including by premise operators), such as by providing services or capabilities consistent with an end user’s choices regarding parental controls or security capabilities; and reducing or mitigating the effects of congestion on the network. The term “particular network architecture and technology” refers to the differences across access platforms such as cable, DSL, satellite, and fixed wireless.

In paragraph 92, the Order recognizes that other practices not included in the definition above may also be considered a “reasonable network management practice” and that broadband service providers “should have the flexibility to experiment, innovate and reasonably manage their networks.”

The Order further recognizes that “appropriate and tailored” in the definition of reasonable network management does not rise to the same “narrowly and carefully tailored” standard as previously used by the FCC’s Comcast Network Management Practices Order in 2008. The FCC expressly acknowledged that this previous standard was unnecessarily restrictive (paragraph 85). In the same paragraph the Order states that service providers “need not necessarily employ the most narrowly tailored practice theoretically available to them.”

Transparency

Service Providers must be aware of the detailed Transparency requirements of the Order, most notably for disclosure of Network Policy Control practices in respect of (Paragraph 56):

• Congestion Management: If applicable, descriptions of congestion management practices; types of traffic subject to practices; purposes served by practices; practices’ effects on end users’ experience; criteria used in practices, such as indicators of congestion that trigger a practice, and the typical frequency of congestion; usage limits and the consequences of exceeding them; and references to engineering standards, where appropriate.
• Application-Specific Behavior: (as discussed under Traffic Optimization above)
• Device Attachment Rules: If applicable, any restrictions on the types of devices and any approval procedures for devices to connect to the network.
• Security: If applicable, practices used to ensure end-user security or security of the network, including types of triggering conditions that cause a mechanism to be invoked (but excluding information that could reasonably be used to circumvent network security).

In paragraph 59, the Order notes that “We also expressly exclude from the rule competitively sensitive information, information that would compromise network security, and information that would undermine the efficacy of reasonable network management practices.” In paragraph 56, with respect to disclosure of the commercial terms of a service, the Order states that service providers must disclose in their Privacy Policy “whether network management practices entail inspection of network traffic, and whether traffic information is stored, provided to third parties, or used by the carrier for non-network management purposes.”

While public comments to the NPRM were streaming in, a federal appeals court finally issued a ruling in the Comcast vs. FCC case that launched the Network Neutrality debate in full force in 2008. The court found that the FCC didn’t have authority over Comcast’s network management practices.  So who owns the Network Neutrality debate? The FCC has started another process to try to get the jurisdiction that the courts said that they did not have, which caused countless US Congressmen to caution the FCC that they were stepping on the toes of Congress. All the while, industry was working together to try to come up with their own proposal. 

Last week, Verizon and Google issued a very public stance on the network neutrality topic.  In essence, their joint policy proposal enshrines the notions of reasonable network management (like the draft rules in the NPRM and the FCC’s broadband Policy Statement of 2005 before it) and the importance of transparency of such practices.
At Sandvine, we’ve long advocated that an “unmanaged network is not a neutral network”.  We took the opportunity to educate policy makers on the nature of today’s Internet; outlining how certain applications and users, left unmanaged, take more than a fair share of network resources, to the detriment of the quality of experience of others.  Access to the Internet needs to be equitable.  In consultations with the FCC, and other industry leaders such as the National Cable and Telecommunications Association, I have recommended the following criteria for ‘reasonable network management’:

1. Narrowly-tailored
Management is implemented only where congestion exists and when congestion is causing quality of experience issues for a large number of subscribers.

2. Proportional and reasonable effect
Policy has an effect on subscribers or applications that is proportional to the effect the user or application is having on the network.  Policy applies the smallest reasonable intervention to alleviate congestion and improve quality of experience for the majority of subscribers.

3. Legitimate and demonstrable technical need
Congestion management and quality of experience issues can be demonstrated to exist in the network and management’s technical remedies are effective in achieving its targeted goals.

4. Transparent disclosure
Operator can disclose its traffic management policies in a simple, useful and predictable manner.

5. Auditable
Service provider can demonstrate that the above requirements are met through its auditing and reporting capabilities.

It seems like the principles of reasonable network management and transparency are gaining broad acceptance across the industry.  While Sandvine supports the spirit behind the Verizon Google (V-G) proposal, I believe that one substantial improvement could be made with respect to the appropriate use of prioritization, where the V-G proposal is unclear. In one instance the proposal says: “Prioritization of Internet traffic would be presumed inconsistent with the non-discrimination standard, but the presumption could be rebutted.” And in another:
“Reasonable network management includes any technically sound practice: … to prioritize general classes or types of Internet traffic, based on latency”. 

I believe that the V-G proposal is trying to avoid anti-competitive prioritization (such as between service providers of a given application class, such as Google versus Yahoo!), while supporting prioritization for the purposes of managing congestion and preserving the quality of experience for users and applications (such as prioritizing time-sensitive VoIP traffic over e-mail during times of congestion), which I support. I am confident that any rules that may result from this proposal would clarify this point.

The V-G proposal wisely carves out “managed services”, which allows service providers to offer their own differentiated value-added IP-based services, such as an IP video or voice service, and prioritize that traffic to give subscribers the quality of service they expect. Also, the V-G proposal suggests that wireless networks be excluded from the rules, which I believe is a prudent approach for now. Given the dynamic evolution of wireless networks, a wait-and-see-how-it- works-in-fixed approach seems acceptable to me.  And last but not least, the V-G proposal calls for limited FCC jurisdiction, preventing a blank-slate of authority by the FCC over Internet broadband access, and essentially keeping other aspects of the Internet value chain unregulated. This is the outcome that Internet content providers are looking for, and the outcome that will continue to spur growth in consumer content choice.

This theme was also echoed at the Light/Heavy Reading panel in which our CTO, Don Bowman, participated in titled “Policy, DPI & Net Neutrality: Heaven or Hell”.  The general consensus by the panelists was that the definition of reasonable and transparent network management, although ambiguous, did provide service providers the latitude necessary to manage their networks for optimal subscriber experience and to move ahead with their plans. 

More specifically, closer to home, there is a lot of buzz about how to optimize mobile traffic to improve the subscriber quality of experience.  This is in the form of advanced techniques for delivering video since video is the “killer app” in mobile, just as it is on the fixed side.  Service providers want video on their network but need to ensure good quality – Sandvine solutions can easily interoperate with such specialized applications for traffic optimization.  Implementing network efficiencies, either at the edge with 4G technologies or in the core, is one of the 4 prongs that AT&T Mobility CEO Ralph de la Vega advocated in Tuesday’s CTIA keynote to spur mobile growth, the other 3 being releasing more spectrum, leveraging compatible technologies (Wifi, Femtocells), and optimizing applications for delivery over the air.  It’s exciting to be part of this mobile Internet explosion!

For example, as a result of being situated next to one of the world’s largest fresh water lakes, the city of Chicago offers an un-metered water service and flat-rate billing based upon the size of your house.

One would initially think that this isn’t necessarily a bad thing, but what happens, is that it leads to a tremendous waste of water. After living in the city for a while, Chicagoans tend to get lax on how they use Lake Michigan’s resources. Simple tasks like watering the lawn and not paying attention (or at times even caring) can lead people to water the sidewalk or city streets. Trust me, I lived there and did it from time to time (of course not on purpose) and I saw many of my neighbors do it.

This isn’t all that much different from the Internet — downloading large files and never using them or sending large attachments to everyone in your address book effectively wastes bandwidth that could be used by others.

I now live in the Sonoran Desert where water is a precious commodity and I pay for every drop I consume. Now that my monthly bill is tied directly to how much I use, I quickly take all the recommendations on how to consume less water; such as ensuring that I have no leaky faucets, minimizing or eliminating the grass (or sidewalks) I water and not washing the family car in the driveway.

Now of course this isn’t a perfect analogy as water has intrinsic value per liter and a packet does not. And water companies are less concerned about peak usage as they have the ability to store reservoirs of water to tap into during peak hours. But I think the analogy is close enough to make the point that people should be made aware of their consumption of network bandwidth so as not to waste it like many of my former neighbors did with their water in Chicago.

At the end of the day, bandwidth is a limited network resource. If there is any way we can make consumers aware of how they are using their bandwidth to ensure they are not wasting it, it would greatly help everyone’s quality of service when using their Internet service. Usage-based billing is one of the most effective ways service providers can manage the costs of building out their networks while at the same time making users aware of their usage.

However, over the years discrimination has become synonymous with prejudice, and it is in this light that the network neutrality debate has focused. However, differentiating one thing does not have to come at the expense of another. Differentiation can be a means of reducing waste and increasing efficiency. Consider the case of a nation in south-east Asia. Assume a popular networked video game console is launched, and that all of the gaming servers are located in Redmond, Washington, USA. Assume that the nation has 2 primary means of reaching the west coast of the US: one that goes east to west, and one west to east. Based on distance alone, the east to west underwater cable will have the lowest latency. But, the overland west to east cable will also have many more routing exchanges, which add even more latency (as much as 1-2ms per exchange).

In this environment, if the ISP were to route gaming traffic destined to Redmond via the short cable, and email traffic via the long cable, there would be an advantage given to the gaming traffic, and no disadvantage to the email traffic. Thus differentiation does not need to come at the expense of anything. This network is more efficient, and all users would achieve a better experience using it.

Is it the intent of the new fifth principle to disallow such activity? The explicit wording would appear to do so. But that goes against the spirit of innovation and relentless search for efficiency that has been the core of the Internet culture.