Programmable Deep Packet Inspection (DPI) for Service Providers
by
Tim Waters,
VP of Marketing
12/17/2007
Service providers around the
globe are in the process of converging legacy and future network services to a
common IP infrastructure. While global IP networks have created great
opportunities for growth and business transformation, they also present a new
set of challenges to the service providers operating these networks.
Key Network Infrastructure
Challenges
First, broadband Internet
services are commodity services with razor thin margins. As legacy cash-cow
services are migrated to IP, it is essential that service providers create new
value-added IP services (above and beyond basic Internet access) which will
allow for revenue growth with high profit margins. Secondly, it is essential for
service providers to optimize the capacity of their network to deliver a wide
variety of network services to different subscribers such that bandwidth
allocation is aligned with maximizing customer satisfaction and service revenue.
Figure 1 -- Driving
Revenue via Service Creation
Finally, threats of DDOS and
other attacks are escalating. Service providers must have the capability to
protect their networks from current and emerging security threats.
These are the key challenges that
service providers face in an IP centric world -- unfortunately, the switches
and routers in the current IP network infrastructure are not well suited to
solving these problems.
A Rapid Migration to The
Next Generation DPI Network Infrastructure
In order to address the
challenges of IP convergence, many service providers are utilizing Deep Packet
Inspection (DPI) technology. DPI products allow service providers to monitor and
control traffic at all layers of the protocol stack (including the application
layer) based on a set of policies.
There are three fundamental
business drivers for implementing DPI network platforms:
Maximize service revenue;
Minimize network capital
expenditures (CapEx) and operating expenditures (OpEx);
Limit the risk of security
threats to the network.
Service
providers have successfully deployed the first generation of DPI products. These
are application-specific point products targeted at solving very specific
problems in the network; for example, controlling P2P traffic, preventing
network intrusions (IDP), or implementing network access control based on
identity and policy.
Many of these products are adept
at solving specific problems; however they are not flexible enough to resolve
the emerging network challenges or implement new services as infrastructure
dynamics and requirements change. This means that new hardware needs to be added
to the network as new requirements emerge. Adding hardware can be extremely
costly, especially in networks that require OSMINE certification and complex OSS
integration. New hardware also introduces additional points of failure in the
data path, reducing overall service availability. Furthermore,
application-specific DPI products can not effectively keep pace with the
velocity of change required to respond in a timely manner to new requirements.
In order to avoid adding new
hardware and redesigning networks as requirements change, a second generation of
programmableDPI products is now being successfully
implemented in networks. Programmable DPI products allow service providers to
run multiple DPI applications at line speed, adding new functionality using
software only updates. These applications can provide bandwidth-on-demand,
traffic management, intrusion detection and prevention, usage based billing,
content insertion, security services, and other complex application layer
services. Using a programmable DPI architecture, service providers have the
flexibility to implement new value added services while optimizing network
traffic and securing the network from security threats.
The Business Drivers
for Programmable DPI Platforms
Driving Top Line Revenue:
One of the key drivers for DPI
network deployments is maximization of service revenues and profitability.
Programmable DPI functionality allows service providers to offer a wide variety
of value-added services on top of basic broadband access. Today most service
providers are focused on providing the basic set of triple play services
consisting of Internet, telephone, and TV service. However, even basic telephone
and TV services are being threatened by Internet voice services (Skype, Yahoo
Messenger, Google Talk) and video content services (YouTube, iTunes, Netflix).
In order to justify the large capital investments that are being made in
broadband infrastructure, it is critical that service providers move away from
the "dumb pipe" model of broadband access to a "smart pipe" model of
premium services. By using "smart pipes" service providers can play a
pivotal role in the emerging broadband service value chain.
Premium services require traffic
management, monitoring, and content modification at the application layer. For
example, a customer might subscribe to a gaming service that provides additional
network bandwidth for gaming web sites at certain times of the day. Another
example is a premium subscription to an Internet video-on-demand service that
provides extra bandwidth for real time video viewed over the Internet. Other
types of services involve application specific traffic monitoring, allowing
service providers to implement usage based billing for certain applications. A
third class of premium services enables services providers to insert and/or
modify the content in application streams -- for example, advertisement
insertion in video streams allowing targeted advertising based on subscriber and
video viewing habits.
It is clear from these examples
that in order to participate in the premium Internet service value chain,
service providers must be able to monitor and control network traffic and
content at the application layer. It is also critical that DPI products have the
flexibility to support new services as market opportunities are presented. DPI
network processors must be programmable in order to achieve this necessary
flexibility, allowing service providers to adapt their service offerings to
dynamic market requirements.
Maximizing Customer
Satisfaction While Optimizing CapEx and OpEx:
It is equally important that
service providers minimize network capital and operating expenses as new
services are rolled out. This means that network utilization must be optimized
to align service delivery with customer satisfaction. This is a difficult
problem in many networks today because most network elements are not able to
implement traffic management above layer 4. Since most web applications run over
http on TCP port 80, it is impossible for standard routers and switches to
classify and manage web application traffic -- it is all treated as http
Internet traffic. First generation DPI products have attacked the most pressing
problem of rate limiting P2P traffic, however, many of these products are not
flexible enough to address the more general problem of providing fair network
use based on customer, application, service pricing, and service dynamics.
Current trends in Internet content services indicate that the requirements for
traffic management are unpredictable. Therefore, service providers must
implement a programmable DPI solution so that software can be upgraded to
support new requirements for traffic management and control as necessary. By
effectively managing network traffic based on applications, service providers
can optimize utilization of their networks which has the effect of reducing both
capital and operating expenses.
Maximum Network Security
DPI is also required for to
insure best-in-class network security. Threats -- including DDOS attacks,
worm propagation, VoIP service hijacking, toll fraud, and credit card fraud --
continue to escalate in both quantity and severity. In order to protect networks
from these threats it is necessary to implement application layer firewalls,
intrusion detection and protection (IDP), and network monitoring and control
based on identity, services, and most importantly, policy management.
First generation security devices
have implemented basic firewall and IDP capability but, as threats have become
more sophisticated, it is necessary to monitor and control network traffic based
on a more complex set of policies. For example, traffic control policies might
be based on a combination of application behavior, protocol anomaly detection,
vulnerability signatures, and user identity. As threats become more
sophisticated it is important to be able to add new security functions into the
network as necessary, driving the need for a programmable DPI
product that is software upgradeable. This allows new security applications to
be implemented in the network, thus dealing with new threats without replacing
existing network hardware or implementing a network redesign.
Figure
2 brings these three market
drivers together and presents an overview of how programmable DPI can transform
a "dumb pipe" broadband network into a "smart pipe" thus allowing
service providers to 1) increase service revenues and margins, 2) optimize
network utilization, minimizing operating expenses and 3) protect the network
from security threats.
Figure 2 -- Business
Drivers for Programmable DPI
Composition of the
Next Generation DPI Device -- 7 Key Requirements
Now that the key business drivers
have been established for DPI, we will examine some of the specific requirements
for next generation DPI network devices.
1. Programmable
By definition, DPI network
infrastructure devices must be programmable, allowing for significant software
flexibility, scalability, and in service software upgrades. These capabilities
enable an emerging class of DPI applications supporting new premium services,
traffic management, and network security policies to be implemented without fork
lift upgrades of network hardware.
2. Active and Passive
Monitoring
Some service providers use DPI
products to monitor how subscribers, services, and applications are using the
network, utilizing this data in network planning and engineering. Others want
the DPI products to play an active part in controlling network traffic to
deliver new services and optimize bandwidth allocations. The DPI platforms must
be capable of both these functions. Furthermore, this functionality must be able
to be modified to support new requirements by upgrading the software in the DPI
device.
3. Full Packet
Scanning
DPI devices must be able to scan
every packet at line speed. Scanning must include all network protocol layers as
well as application content. Additionally scanning must be able to correlate
session layer content in sequential packet streams, providing maximum
flexibility to implement services or security threat mitigation at the session
layer.
4. Traffic
Classification and Management
Classification of network traffic
must be flexible and allow service providers to classify traffic by multiple
parameters which include: subscriber, service, application, origin/destination,
IP address, and other parameters related to services and protocols. Traffic
classification could also be based on usage and traffic patterns of subscriber
sessions. Based on the network traffic classification and policies, DPI
platforms must allow service providers to control traffic flows. This includes
blocking, rate shaping, and redirection of packets to different destinations.
5. Packet
Modification and Content Insertion
In some cases it is necessary to
modify packet headers or application layer content. There are many other
potential services and applications that could require packet modification.
Therefore, it is important that a DPI platform deployed in a service
provider's network has this capability.
6. Packet Generation
Some services or security defense
systems require DPI processors to generate packets. For example, proxy servers
might require packet generation to set up connections, defense mechanisms might
need to send TCP packets to throttle to stop malicious TCP attacks, or some
applications (such as SIP) could require in-band signaling. DPI platforms must
have the capability to flexibly generate packets based on the application and
service objectives.
7. Billing Policy and
Control
A critical component of service
management is billing. Today, most Internet services are based on fixed rate
billing -- a legacy of the "dumb pipe" broadband access model. As premium
services are implemented it will be important to be able to bill customers based
on usage, bandwidth, service offering, time of day, response time, and other
parameters that define a service level agreement (SLA). The DPI processor must
have the capability to monitor many different aspects of user session, collect
data, and generate billing records.
Such programmable DPI devices are
now emerging into the marketplace. Bivio's DPI platform is a good
example of a programmable DPI processor that allows service providers and their
systems integrator partners to run DPI applications in a standard Linux
environment, yet deliver the flexibility and scalability required to optimize
carriers' evolving network infrastructure. Custom, commercial and open-source
applications, tailored to maximize carrier revenue opportunities and/or minimize
further CapEx and OpEx investments are easily and seamlessly ported to the
platform. This model allows service providers to add and enhance new services
and manage network application traffic as requirements change without adding new
hardware to the network. Furthermore, carriers can operate their networks at
line speed, thus minimizing bottlenecks and maximizing customer satisfaction.
In summary, DPI devices are
essential components of next generation IP networks and one should ensure that a
programmable DPI solution is implemented which is sufficiently flexible to solve
current as well as future challenges.
About
the Author
Tim Waters has more than 18
years experience in marketing data and telecommunications products and
services. He is responsible for overall marketing and product
management of Bivio networking appliances.
Prior to joining Bivio,
Waters was vice president of marketing and business development at
NetDevices, Inc., a leading supplier in the emerging enterprise service
gateway marketplace. He has held similar positions at SkyStream
Networks, a manufacturer of IP video headends, and Onetta, Inc., where
he led the creation of the intelligent optical engine market. Waters
previously was vice president of marketing and business development for
Promatory Communications, a leading supplier of next-generation DSL
access multiplexers, subsequently acquired by Nortel Networks.
Earlier in his career Waters served as vice president of data product
management for Ameritech in Chicago, Illinois, and held several
management positions at Paradyne Corporation and AT&T Computer
Systems.
Waters holds an MBA from
the Harvard Graduate School of Business Administration and a BA in
economics from College of the Holy Cross.
About
Bivio Networks
Bivio
Networks has developed a family of industry leading, deep packet
inspection and processing platforms that combine unparalleled scaling of
network performance, processing power, and application agility. Bivio's
network appliance platforms feature a groundbreaking architecture
specifically optimized for wire-speed execution of emerging network
services that increasingly demand deep packet processing combined with
high network throughput. Based on open industry standards, Bivio
Networks fuses unmatched flexibility with uncompromising performance to
enable its customers to overcome existing bottlenecks and deliver the
foundation of the next-generation network infrastructure. Headquartered
in Pleasanton, California, Bivio Networks is backed by Goldman Sachs,
InterWest Partners, Storm Ventures, Venrock Associates, and Silver Creek
Ventures.