Posts Tagged ‘accuracy’

Radical Simplicity: Moving Mountains for Personalized Services

Monday, January 16th, 2012 by Cam Cullen

Steve Jobs once said, “That’s been one of my mantras – focus and simplicity. Simple can be harder than complex: You have to work hard to get your thinking clean to make it simple. But it’s worth it in the end because once you get there, you can move mountains.”

There are three pillars to Intelligent Policy Enforcement: Awareness, Analysis, and Control. Solutions that lack any of these capabilities will limit the ability of an operator to deliver personalized services to their customers. The more they lack, the more limitations that will be imposed on the operator – whether those limits are scalability, performance, granularity, or even operational in nature.

Let’s be upfront about something: Personalized services are HARD.

Think of all the pieces that have to fall into place to deliver a personalized service.

1) A subscriber activates their device, and attempts to access the network. The network authenticates the subscriber and/or device to ensure that they are allowed access to the network.

2) Subscriber “service” must be determined and provisioned to the network. The service “attributes” may include bandwidth allowed, volume of data (either bulk or per service), priorities, security filters, value-added services, or even multi-device correlations. Depending on how the network is configured, this provisioning could be a single transaction to a single system, or many transactions to many systems (policy, charging, access, etc.). The subscriber and their data traffic must be associated with whatever “attributes” that are necessary to deliver their purchased service to their devices.

3) The systems that have been provisioned with the service “attributes” must enforce those attributes without causing instability of the systems or requiring additional CAPEX due to “unexpected” scalability issues.

4) The systems must perform network and business intelligence gathering on all traffic to feed the analytics systems for future service development. The more information that can be collected, the more valuable the data is for the operator. This data has to be collected with as small of an interval as possible, as the longer the interval the more you lose granularity (for example spikes in bandwidth and latency can be smoothed out over long intervals)

5) If volume/event-based charging or CDR/UDR generation is required, every charging event can result in a transaction with the charging infrastructure, and especially at session closing time.

6) In a network where phones go into an “idle” mode or where subscriber movement between locations is tracked, every time this happens can result in more transactions.

And the more services you offer, the harder it gets…

With the announcement of our PL20000 today, we believe that we have delivered the final piece of the puzzle for an ecosystem that will enable the largest network operators in the world to compete on an even keel with smaller, more agile network operators. The limitations faced by large operators are focused on scalability and complexity. Scalability is now a moot point, as the PL20000 can meet the needs of even the most stressful network deployments with the massive performance increase and the port density/support for 100GE interfaces.

The “Simplicity” part is harder (hence the title of the blog). We have been adding capabilities for the last few years to make service creation simpler for operators. Report Studio is a perfect example of this – operators can now ask questions of the network staff and the network team can use Report Studio to answer them based on the massive statistics repository available in the PacketLogic Intelligence Center (which handles an order of magnitude more statistics than other solutions). Our Virtual Services Capability enables operators to define their own services based on application properties (URLs, http referrers, content-type, device-types, file-type, etc.) without intervention by Procera or our signature development team. We have unmatched capability with our PacketLogic Subscriber Management system – in terms of performance (over 100,000 transactions per second), capacity (up to 20,000,000 subscribers in a single system), and ease-of-use (customizable service dictionaries, easy to use interface, interfaces to multiple policy and charging systems simultaneously).

We believe that we are driving solutions towards the nirvana of ‘Radical Simplicity” where carrier-grade IPE solutions do not have to be managed by acolytes of technology that speak arcane languages (ok, maybe program arcane languages). The combination of performance, scalability, and ease-of-use will take Intelligent Policy Enforcement to mass-market deployments for operators of all sizes. Operators that are forced to wait 9 months for a service launch to capitalize on a new trend will find that the trend is no longer hot. Operators that must plan for a performance or scalability reduction for every new service that they launch will very quickly find their solutions very capital intensive and inefficient.

Simplicity is hard. We want to help you move mountains, so we are working smarter (as well as harder!).

Device Awareness: The Key to Service Planning

Thursday, July 22nd, 2010 by Cam Cullen

Recently there has been a lot of talk among DPI vendors about location awareness. Location awareness is critical to understanding where the bottlenecks on your network occur, and giving you the ability to actively manage traffic in those locations. Although this is a key capability of what Procera offers (and in much more depth than many other vendors), I want to focus on another key “awareness” component that is not often talked about that can vastly improve the capability of an operator to plan their capacity upgrades and pricing plans.

What is device awareness? Simply put, device awareness is the ability of the DPI system to understand what device is sending/requesting traffic on the network. With so many “internet-connected” devices on networks today, the variety of places and formats to access data has exploded. You can watch Netflix on many TVs, DVD or Blu-Ray players, dedicated boxes (like Roku), consoles (PS3, Wii, and Xbox), smartphones, tablets, or your PC, and Hulu on many of the same.

Why is device awareness important? Understanding what devices your users have connected to the network helps an operator to understand the potential impact of several different scenarios:

1)     If I have x users with a certain device on my networks with a certain location bias and usage profile, what happens if that number becomes 2x?

2)     If a new application comes out (say for example Netflix for PS3, or Hulu Plus for PS3), and I have x users with that device, what is my exposure if 20% of my users start using the new application?

3)     If I want to launch a service with a new device (new iPhone or Android device) and other operators have announced that the normal usage profile for that device is 200GB/month, what might happen to my network with an aggressive take rate?

For Product and Marketing planners, being able to answer these types of questions ensures that the network can handle any changes in user behavior – either application or device related. For users, it ensures that the network can adapt to their usage and not experience dissatisfaction with network performance as usage patterns morph over time. All operators are very aware of the user dissatisfaction issues floating around the world as Smartphone and mobile broadband laptop connections become mission critical for corporations. Proactive management and planning are possible with the right kind of awareness and reporting engines.

How is device awareness implemented?  There are many ways that this can be gathered on the network. In mobile networks, it is common for operators to include this information in the Business or Operational Support Systems for each user account, and often make it part of the user authentication and accounting exchange where it can be snooped by a DPI system. It can also be determined by the applications used by the subscriber (i.e. if the traffic matches Rock Band for the PS3, the user is most likely using a PS3). Another method is by information gleaned from the user agent or application specific information that can be gathered by the DPI engine during Layer 7 analysis. Although each of these methods is not a fail-safe method to gather device information, the combination of these methods can provide a very close approximation for network operators. For example, a query on how many unique subscribers matched a PS3 signature during the month could provide the basis for determining how many subscribers would have access to the new Hulu Plus video streaming application when it was released for the PS3, and another query to see how many of those subscribers had used Hulu (or any streaming video service) before would provide additional planning input to determine the impact of that new service on their network.

Device awareness is a key aspect of the “Awareness” capability of the Procera solution, and without awareness, analysis and control are not possible. I will have more to say on location awareness in a future blog, as this is another key awareness factor, not only for mobile networks, but also for fixed and wi-fi deployments as well. As devices proliferate, being able to ensure that the content is served appropriately for those devices and ensure the correct amount of bandwidth is available will be critical.

Economy of Scaling To 10G Networks

Tuesday, April 6th, 2010 by Jon Linden

“PacketLogic is now available with gigabit interfaces!” This is not today’s announcement, but it’s also not ancient history. We launched the second generation of PacketLogic with (almost) gigabit capacity in April 2004. Four years later, in May of 2008, we released our high-end PL10000 family with 10G interfaces and what is still the highest capacity of any DPI system available in the market today.

What I’m saying is that we love breaking speed records. Now we’re doing it again; but in a very different way. The new PacketLogic PL8720 is the first 10 Gbps DPI system that comes in a slick, real estate and power efficient 2RU appliance with the market’s highest port-density per rack unit. Most importantly, it’s cost efficient.

Numbers typically speak louder than words. I happened to read the following earlier today: “Market tracker Dell’Oro Group says 10G, which makes up 25% of the market in terms of revenue, will drive growth in Ethernet switching this year. […] Dell’Oro expects 10G to reach $3.6 billion in revenue for 2010, up from $2.8 billion in 2009.”

This means that not only big carriers deploy 10G these days. Tier 2 and 3 operators, even higher education institutions, deploy 10G. And they will have to in order to cope with a growing traffic volume and to ensure that the Internet remains the innovation engine it is. More speed means more traffic, which means an even bigger need to understand what’s going through your network, and automate detection of potential threats to both users and the performance of the network.

Doing deep packet inspection at a 10 Gbps rate is computing intensive. This is why it historically has been done in chassis-based solutions – products that are and still will satisfy the large operators’ requirements for resilience and upgrade capabilities. The extensive computing intensity is the reason why DPI has been such a big investment that many small to mid size network operators have postponed this decision. Not deploying DPI puts them in an unfavorable competitive situation. We saw this gaping market demand and set out to develop a cost efficient product that’ll upgrade all the way to 10 Gbps – without compromising on functionality.

Are you getting the dignity of this? Do you see why cost efficiency is so important? I’d say this is flat out a game changer! You’ll see. PL8720 will have first-mover advantage in a rapidly evolving market where we put DPI in the hands of a new set of service providers and higher education institutions. They, and their users, will experience the benefits of a full-blown DPI product. Many of these have hesitated to make this investment, though they’re fully aware of the need, because they know they’ll still make further network upgrades and will transition from GE links to 10GE. Now this is not an obstacle anymore.

I just wanted to let you know what we’re doing to change the world of networking. Pretty interesting, huh?!?

Is Accuracy Really That Important?

Wednesday, December 2nd, 2009 by Jon Linden

Trust me, it is. There are a lot of good reasons why we promote “accuracy and control, redefined” in our logo. Ask any operator with DPI experience, and you’ll hear that accuracy is top of the list. And we can, in all honesty, say that DPI didn’t really deliver on this promise originally. The first generation of DPI identified port-hoping filesharing applications good enough to cap them to avoid disaster.

But times have changed. Today we have very sophisticated tools in our bag and our traffic identification engine looks at several criteria when determining what application each individual connection is. We also leverage characteristics, like interactive, streaming, download and bulky to categorize traffic in an application-agnostic fashion.

Online applications have evolved extensively over the years I’ve been working with DPI. Back in the days when IP and TCP were invented, all traffic was client-server-based. The applications were neither time nor quality sensitive, but everyone was happy with a global and resilient network.

Fast forward to today. P2P technology is used to leverage bandwidth and CPU capacity at the edge of the network for faster connectivity and to decrease the traffic being sent over the core network. P2P technology is used by the streaming music services we run all day at the office, as well as the online HD video on demand service we use at home at night. Both Salesforce.com and our office phones run over IP enabling us to work from home as if we were at the office. But we would also be totally paralized if the Internet connection (as well as the redundant link ;-) ) was down. This is how crucial and integrated the Internet is in our lives today, and this is why traffic volumes grow at a pace that outdoes Moore’s law and that saturates pipes.

Of course, these applications are totally different in nature and have different requirements for how to be treated to function properly. Of course, different users have different expectations in different situations at different times of day. Of course, it would be an issue if you treat HD video as filesharing or World of Warcraft as SIP. This will impact the performance of the network, but also your ability to manage expectations and create viable business cases for how to satisfy different user profiles. Step one in any process is analysis, and unless you trust the intelligence you use for your analysis you won’t dare to make decisions based on these facts.

So, this is one of these cases where good enough isn’t good enough. Trust me when I say that you will want to trust the information you have at hand when you make critical decisions.