Preparing the PoP

Winding Up for Speeds Beyond T1

Engineering PoPs to keep one step ahead of this bandwidth frenzy, adhering to strict budgets and deploying equipment in a shrinking physical space is a huge challenge.

Of course, expanding bandwidth demand is driven by a growing reliance on the Internet, intranets and extranets and the increasingly complex applications they enable. E-commerce, distance learning, virtual private networks (VPNs) and other applications often include multimedia content and advanced programming that easily swamp dial-up and dedicated 56 kilobits per second (kbps) services. In fact, T1 (1.544 megabits per second [mbps]) circuits are today's mainstay for business-quality access to data networks, while economically minded businesses choose digital subscriber line (DSL) for lower speed access where available.

Now, an increasing number of businesses, universities, government facilities and other institutions have outgrown T1. They are looking for affordable, readily available, higher-speed access. The next logical step would appear to be fiber-based DS-3 (45mbps), with up to 28 times the capacity of T1. However, most businesses require only a fraction of a DS-3's bandwidth and are unwilling to pay its substantial installation, provisioning and monthly access fees. The majority of these business facilities also do not have access to fiber. Service providers need to introduce a new service that will bridge the bandwidth, affordability and availability gap between T1 and DS-3. With more than 50 percent of small and medium-sized businesses predicted to fall into this gap by the year 2002, it is not too soon to begin designing and deploying solutions that enable ISP PoPs to close this bandwidth gap.

Closing the Gap

Ideally, a solution for bridging the gap between T1 and DS-3 and providing affordable, readily available, multimegabit Internet access would maximize the use of the existing network and equipment infrastructure and minimize costs and deployment time. This solution would take advantage of existing tariffs and billing systems to enable rapid new service delivery.

If done well, bundling multiple, inexpensive, readily available, secure and symmetric T1 lines into one scalable, multimegabit, multilink path between the customer premises and the PoP could bridge that gap. It would also be highly reliable, since the failure of one T1 circuit would not bring down the entire access bundle.

Getting more bandwidth by bundling multiple copper links is not a new idea. It has been used extensively in the dial-up integrated services digital network (ISDN) world and there are several solutions for T1 as well. Unfortunately, none of these older T1 bundling technologies is acceptable for general deployment of a multimegabit Internet access service that meets the needs of both ISPs and their business subscribers.

Bit-interleaved inverse multiplexing operates at layer 1 of the open systems interconnection (OSI) model. This technology is supported by bit-based inverse multiplexers that receive a bit stream from a router, insert overhead information and transmit the bits onto multiple T1 circuits--increasing the wide area network (WAN) access capacity. At the other end of the WAN, another inverse multiplexer terminates the T1 bundle and reassembles the bits into their original order. The resulting bit stream is sent to a outer, which performs all layer 2 and above processing.

This technology is typically used to make a point-to-point connection between two routers at speeds faster than T1. Inverse multiplexers from the same vendor must be used at both ends of the connection because each supplier's scheme is proprietary.

Of course, expanding bandwidth demand is driven by a growing reliance on the Internet, intranets and extranets and the increasingly complex applications they enable. E-commerce, distance learning, virtual private networks (VPNs) and other applications often include multimedia content and advanced programming that easily swamp dial-up and dedicated 56 kilobits per second (kbps) services. In fact, T1 (1.544 megabits per second [mbps]) circuits are today's mainstay for business-quality access to data networks, while economically minded businesses choose digital subscriber line (DSL) for lower speed access where available.

Now, an increasing number of businesses, universities, government facilities and other institutions have outgrown T1. They are looking for affordable, readily available, higher-speed access. The next logical step would appear to be fiber-based DS-3 (45mbps), with up to 28 times the capacity of T1. However, most businesses require only a fraction of a DS-3's bandwidth and are unwilling to pay its substantial installation, provisioning and monthly access fees. The majority of these business facilities also do not have access to fiber. Service providers need to introduce a new service that will bridge the bandwidth, affordability and availability gap between T1 and DS-3. With more than 50 percent of small and medium-sized businesses predicted to fall into this gap by the year 2002, it is not too soon to begin designing and deploying solutions that enable ISP PoPs to close this bandwidth gap.

However, this technology was not designed for the Internet or other services that connect many subscribers to a central facility and where the use of standards is essential.

At the PoP, every subscriber requires an inverse multiplexer and associated router port. This is expensive, consumes collocation space and is difficult to admini-ster and maintain. It also precludes the use of multihoming, as bit-interleaved inverse multiplexed T1 circuits must terminate at the same facility.

Meanwhile, load balancing, which operates at layer 3 and above in the OSI model, intelligently directs packets down parallel paths based upon the packet destination and other criteria. This technology is available in most routers, and has been used extensively to provide higher-capacity WAN access for up to two T1s. The load balancing task becomes increasingly difficult, however, as more WAN circuits are used, often resulting in inefficient use of multiple T1 circuits and high consumption of router processor resources. In addition, the load must be balanced at installation; as load characteristics change over time, performance can degrade. Consequently, businesses do not get their subscribed bandwidth, and the ISP pays a high price for maintaining a cumbersome solution.

Another option that transports asynchronous transfer mode (ATM) cells through parallel T1 circuits and recombines them at the PoP is inverse multiplexing for ATM (IMA). ATM tends to be used in access networks where quality of service (QoS) is more important than bandwidth efficiency. Transmission control protocol/Internet protocol (TCP/IP)-based networks, such as the Internet, tend not to use ATM as an access technology because of the stiff cell tax that can reduce the payload capacity of a circuit by 20 percent to 30 percent due to cell overhead and inefficient transport of 64 byte packets. The longevity of ATM as an access technology is cloudy because of these efficiency problems and the rapid advances being made with IP QoS.

A New Way

A new solution is now available that overcomes the deficiencies of previous multilink T1 technologies and was designed to support the deployment of affordable and effective multilink, multimegabit copper services. Products that implement this solution create a virtual multimegabit access path from the business subscriber to the PoP. This path emulates the throughput and latency characteristics of a single circuit operating at the combined speed of all the bundled circuits. The solution employs programmable packet fragmentation and reassembly techniques that reduce latency and ensure high efficiency utilization of each T1 link in the bundle.

The virtual multimegabit access path is created by bundling multiple T1 circuits together using standard layer 2 multilink protocols, such as multilink point-to-point protocol (PPP) or multilink frame relay. Use of standards means products employing this solution are compatible, a key attribute of any public data service offering. Use of standards also enables complete plug-and-play operation; no load balancing or future maintenance is required for efficient use of each T1 circuit.

Proper use of these protocols, along with layer 1 error information, enables continuous operation of the bundle (at reduced bandwidth) even if one or more of the T1s fail. It also enables the recovery of T1 circuits back into the bundle and a subsequent increase in bundle speed. In addition, the T1s in a bundle can be connected to the PoP and customer premises equipment (CPE) in any order, eliminating the risk of intra-bundle wiring problems throughout the network. These protocols also enable the equalization of differential delay across different T1 circuits in the bundle to ensure proper packet ordering. Customer premises products that employ this new solution employ new technology and design techniques to overcome the performance problems previously associated with multilink PPP solutions.

This multiline solution also requires a new device at the PoP, since most existing routers and switches do not effectively support multilink PPP or multilink frame relay. This new equipment should terminate the virtual multimegabit access paths of many customers, then aggregate packets onto a single, high-capacity uplink such as a 10/100 Base-T Ethernet or high-speed serial interface (HSSI)/DS-3 to connect to the existing router or switch respectively.

Because equipment employing this new solution is transparent to the current network routing topology, service providers can deploy multimegabit copper services while leveraging their existing PoP equipment and minimizing network changes and new maintenance activities. Connecting customer T1 interfaces to the PoP device through a channelized T3 interface enables the service provider to perform adds, moves and changes remotely. However, it is also important to provide optional copper T1 interfaces for those PoPs that don't have access to channelized T3.

Aggregating multiple links at the PoP should provide superior multilink performance and efficiency in the face of small packets and many customers. Support for individual interface buffer management schemes incorporating random early discard (RED) and other policy mechanisms ensures adequate burst tolerance and prevents a single subscriber from interfering with the bandwidth guarantees of other subscribers. Support for strong remote and local management capabilities, such as simple network management protocol (SNMP) and standard command line interface (CLI) functionality, is also key for this type of networking equipment. It is also critical this new multilink PoP aggregation equipment fits a small footprint, is low power and meets NEBS requirements (NEBS is the guideline for environmental conditions--such as fire, earthquake, etc.--that central office [CO] equipment must be able to withstand, as defined by the former research and development arm of the Bell monopoly). With this new breed of multilink PoP aggregation equipment, service providers can effectively support copper-based, multimegabit Internet access services.

This technology goes beyond business Internet access applications and can be used to connect any facility that lacks fiber access to data networks at higher speeds. Service provider PoPs, multitenant facilities and wireless base stations, among others, can take full advantage of the benefits of using multilink multimegabit T1 copper circuits. As the bandwidth explosion continues, service providers can prepare their PoPs for speeds beyond T1 by effectively deploying equipment and services that use T1 to enable multimegabit network access.

Dan Palmer is president and CEO of Tiara Networks Inc., San Jose, Calif. He can be reached at liz@tiaranetworks.com.

Share this article: Email, Slashdot, Digg, Del.icio.us, Yahoo!MyWeb, Windows Live Favorites, Furl
Add this article feed to: RSS, My Yahoo, Newsgator, Bloglines