May 2000

The Sensei's Guide To Multiservice IP

BY TONY RYBCZYNSKI

No doubt most of us can relate to the Karate Kid's frustration. We've learned that tasks of any importance often require us to observe elaborate preparations. One such task, multiservice IP networking, is no exception. Here, the preparations are meant to ensure reliable, high-performance, e-business-grade operations. No mean feat. But the rewards of e-business-grade multiservice IP networking are immense, and well worth the effort devoted to the necessary preparations, even though these preparations may seem daunting at first, or even bewildering.

The main thing is to be persistent. Then, like the Karate Kid, you'll eventually find that all your exercises and preparations make sense. You'll find that you've done more than just condition yourself; you'll find you've somehow acquired the ability to direct your strength at will, to achieve multiservice IP network mastery.

WHY INSTITUTE MULTISERVICE IP NETWORKING?
To a large extent, multiservice IP networking is about uniting data and telephony applications. But why should anyone take the trouble? Well, there are compelling economic and competitive reasons. Consider just one piece of intelligence, from the Gartner Group. According to Gartner, the dramatic growth of CTI- and Internet-enabled customer care applications has already generated $500 billion in business, and will account for over a trillion dollars in business over the next few years.

These customer care applications constitute the main difference between the first and second waves of e-business. The first e-business wave brought www.everything.com into our everyday life. The second e-business wave (e-business 2) is now forming, adding a human touch to e-commerce, enhancing customer service through e-care, and integrating front and back office systems. Basically, e-business, in its latest incarnation, puts forth a new metric. And that metric is... return on relationship.

PREPARING FOR MULTISERVICE IP NETWORKING
E-business 2 requires fast transaction turnaround, consistent application response, 100 percent availability, and tight user security and administration across the entire network. As more applications gain business-critical status, and as more applications and services are rolled out on these networks, IT managers are under increasing pressure to deliver mainframe- and telephony-grade predictability and reliability to their IP networks, while continuing to evolve the reach and capacity of their networking infrastructures.

But how does one satisfy these diverse goals? How may an enterprise enhance predictability and reliability while also extending its network's commitments? How may an enterprise evolve its IP network to a multiservice IP network for e-business?

In this section, we'll show you how. We will break down the enterprise challenge into seven sub-challenges:

1. Institute Switch-Level Equipment Reliability And Redundancy
Traditional IP networks were built on hubs, bridges, and routers with limited or no redundancy. In multiservice IP networks, switches and routers need to have high-level software and hardware reliability. This goes beyond physical level redundancy and reliability. It must also include software that supports rapid automatic recovery at the switch level and that exhibits short restart and upgrade times.

Another key attribute is the commonality of technology platforms across various environments (for example, wiring closet and campus backbone). Such commonality promotes operational effectiveness, minimizes equipment sparing, and simplifies the customer's environment.

2. Adopt An End-To-End System Level View
Traditionally, IP networks achieved reliability through a combination of non-redundant routers running dynamic routing protocols and applications running error recovery protocols (that is, TCP). In contrast, real-time delay-intolerant applications such as IP telephony and interactive collaboration require rapid recovery from equipment and physical link failures.

In the new era of routing, open, modular, and scalable IP networking solutions are required to deliver e-business-grade reliability; the old era of monolithic proprietary systems just can't exhibit the requisite level of nodal software and networking stability. Simplified network architectures (fewer tiers and fewer boxes) using Layer 2 and 3 switching ease networking engineering and design while minimizing fault isolation times.

Some of the key IP networking standards that enhance fault-tolerant networking include high-performance dynamic routing protocols, and protocols for route balancing across paths, for LAN redundancy, and for logical trunking. Layer 1 and 2 recovery mechanisms should be used wherever possible, since Layer 3 mechanisms are often too slow. These considerations apply across the network as well as to servers, given their critical importance to business operation.

Also, comprehensive network management tools are needed to monitor network status and to facilitate rapid fault isolation across multivendor network environments. Ultimately, the result is a more robust network and a simpler environment. One of the benefits is fewer emergencies. And having fewer fires to extinguish goes a long way towards lowering stress among operational staff.

3. Apply IP CoS To Achieve Application-Optimized Networking
Traditionally, IP networks offered best-effort networking, relying on IT's ability to stay ahead of the traffic growth curve. In multiservice IP networks, application-optimized performance via class of service (CoS) mechanisms is required to ensure performance requirements are met under normal, congested, and failure conditions. Also, while portions of the network may be bandwidth rich, CoS is definitely required across some portions of the network (such as the WAN).

The critical CoS networking standard is defined within the Differentiated Service (DiffServ) architecture, supporting a range of end user signaling schemes, including the ReSource reserVation Protocol (RSVP) and Layer 2 mechanisms.

CoS classification functionality should be implemented as close to the application as possible, ideally in the application itself. Very low latency across network switches and a manageable number of user traffic categories across the network can support various classes, such as real-time delay-intolerant (premium), real-time delay tolerant (platinum, gold), non-real-time mission-critical (silver, bronze), and non-real-time non-mission critical (standard).

CoS capabilities may not yield the desired results if they are added to a highly heterogeneous, unarchitected IP network. However, in an appropriately architected IP network, CoS can help enterprises run their mission-critical applications while moving towards convergence.

4. Institute Closed Loop Policy Management
Policies define how network resources are to be provisioned among devices, end users, and applications to meet both CoS and security needs. A complete policy management solution includes a policy entry console or manager for entering policies, a policy decision point or server that retrieves policies and makes decisions on behalf of routers and switches (referred to a policy enforcement points), and policy repositories. (Policy information may be stored in directories compliant with Lightweight Directory Access Protocol, or LDAP.)

Policy-based configuration management operates on the basis of ports, users (including mobile users), and applications, using LDAP to extract policy information from directories, and the COPS (Common Open Policy Service) protocol and CLI to communicate with network switches. The ability to perform dynamic and time of day policy configuration updates needs to be provided to satisfy business cycle requirements.

"Closed loop" policy management includes configuration of edge devices, enforcement of policies in the network, and verification of performance as seen by the end user application through service level management. Enforcement of policies in the network also includes admission controls of applications vying for access to network and application resources.

5. Maintain Flexibility
Traditionally, leased lines have been used to connect sites. We now see attractive alternatives, including virtual private lines using frame relay, ATM, and (increasingly) IP VPNs. In these carrier environments, a high degree of flexibility is required if multiservice IP networks are to be extended reliably, negotiating various interface, signaling, price/performance, and CoS attributes.

Another dimension of flexibility is the ability to support various redundancy options, particularly when working over the public network services (from one or multiple carriers). Options include Layer 1, 2, and 3 mechanisms such as SONET and DWDM redundancy, ATM multilink, IP dual homing, and ISDN backup.

Rapid detection of access and service failures on IP, frame relay, and ATM are also a necessity. It is also necessary to support flexible mapping of application traffic with different CoS needs onto carrier frame relay and ATM services and IP VPNs. Doing so, however, makes sense only if you are confident your carriers are fulfilling their service level agreements (SLAs). Needless to say, the ability to monitor carrier compliance is essential.

6. Encourage Internal Enterprise Initiatives
IP networks have evolved from local PC LANs to wide-scale routed networks. Operational and engineering approaches evolved as well; however, these approaches hardly suffice for business-grade networking. (For example, resetting network equipment to recover from failures just won't do.)

Multiservice IP networking requires that enterprises rethink these procedures and engineering practices. Enterprises need to determine the appropriate level of switch and network redundancy commensurate with their business priorities. They also need to establish more rigorous operational procedures that recognize the transition from best-effort networking to always-on multiservice IP networking.

Enterprises need to define organizational policies to decide which users or applications should be assigned to different service classes. Enterprises should develop strategies to bring about the eventual integration all directory information in a common LDAP-based directory structure. Finally, enterprises should perform risk analyses, which should inform security policies by identifying what needs to be protected, including network resources, computer resources, and company data.

7. Partner With The Right Suppliers
Strategic partnerships with proven, global suppliers represent an important asset for the enterprise IT manager. Ideally, vendor partners should be committed to proactive involvement in (and product compliance with) open IP industry standards. They should have a broad experience base in applications and their networking needs, through their own enterprise network as well as by working with customers across industries. They should also be proven suppliers of highly scalable, reliable multiservice networks supporting voice and data over packet-based technologies, and have products that are certified to work with carrier services on a global basis.

Strategic partners must follow strict software and hardware design methodologies for delivering products of the highest quality. Finally, strategic partners should deliver 24x7 support and a full range of professional services on a global basis.

CONCLUSION
E-business 2 is heralding a new era in enterprise networking, encouraging an application perspective in the delivery of highly reliable, consistent, and secure networking for a full range of real-time and non-real-time e-business applications. If you are to succeed in this new era, you'll need to make the necessary preparations.

These preparations can be broken down into seven general exercises, as indicated in this column. They do involve a good deal of work, perhaps so much work that you may become absorbed in the individual exercises, and lose sight of the overall goal. (The same way the Karate Kid became exasperated with yard work during his martial arts training.) But have faith. The seven exercises towards multiservice IP networking will show you how to build e-business-grade enterprise networks.

Tony Rybczynski is director of strategic marketing and technologies for Nortel Networks' Enterprise Solutions unit. For more information, visit the company's Web site at nortelnetworks.com. E-mail comments to [email protected].

When you run e-business-critical applications over a multiservice IP network, delivering the necessary performance and network availability presents some special challenges. You have to do more than provide 99.999 percent switch availability.

Examples of e-business-critical applications include enterprise resource planning, e-commerce, transaction services, collaborative applications, and financial and customer service applications. If these applications are to receive consistent and reliable connectivity, latency, and throughput, enterprise IT managers must take an application perspective of network performance.

The application perspective admits a broader view of performance and reliability. For example, such a view encompasses a variety of failure conditions, some of which may be less than obvious. Consider the following list of failure conditions. It covers a range of failure types, any of which could compromise network reliability:

• Hardware failures that degrade connectivity, processing, or bandwidth capacity. (In certain industries, such failures could cause revenue losses in the tens of thousands of dollars per minute.)
• Non-hardware failures that cause logical disconnects. (For example, a remote site could be physically connected, but logically unreachable if, say, RIP hop count limits have been exceeded. Also, such disconnects could prevent an application from seeing its network resources, including printers.)
• Network congestion resulting in increased packet loss, which could impact any applications that aren't based on TCP/IP. (For example, poor quality of streamed audio or video. Poor voice quality could lead telephony users to hang up.)
• Back-end processing delays. (Such a delay may prompt a user to abandon a Web transaction. Of course, the user may then visit a competitive site.)
• Exceeding expected timeframes for critical file transfers. (Such delays could conflict with financial book closing intervals or customer contractual commitments.)
• Denial of service attacks that prevent users from accessing your web site.

Most enterprises will evolve their networks from a best-effort IP world. Some will do so on the way towards meeting the needs of selected applications requiring special treatment (for example, customer care, transaction services, ERP). Other enterprises will run the network as an internal service or utility offering a range of service level agreements (SLAs). Major challenges are preparing the network for real-time applications and developing a management and networking infrastructure that can deliver SLAs.