Worldwide access

Many international Internet service providers can deliver high-speed access to end users by using direct satellite channels. Satellite providers integrate advanced traffic management into their services to provide such access, particularly to ISPs in countries with developing communications infrastructure.

Satellite transmission bandwidth can be expensive, so optimizing bandwidth is key when new Internet access services are deployed. Satellite providers manage their traffic with class-based queuing, which runs on platforms built for managing Internet bandwidth.

This lets them develop differentiated, high-value services for ISPs throughout the world. The robust bandwidth management features include the ability to classify users and allocate bandwidth based on defined traffic classes.

NetSat Express evaluated bandwidth management alternatives for satellite applications and selected Xedia's Access Point 100. NetSat Express deployed the solution in a high-speed LAN-to-LAN configuration with 10 to 100 Mb/s Ethernet connections to both an Internet backbone router and broadcast routers connected to individual satellite modems (Figure 1).

In the U.S., Internet traffic is aggregated at the point of presence via the backbone router and then shaped by the access point for satellite transmission. This approach lets remote ISPs provision the level of service and type of link they require. It is important to plan for flexibility and scalability because remote ISPs tend to scale high-speed access when they can reliably provision high-quality bandwidth.

By relying on the public domain class-based queuing software standard, satellite providers can deliver bandwidth class-of-service control. The queuing provides an open and flexible technology for classifying user traffic and meeting explicit bandwidth commitments across a range of Internet protocol applications, users and flows.

Because control is applied at the IP layer, or Layer 3, class-based queuing does not depend on any Layer 2-specific traffic management mechanisms. As a result, it fits into existing networks and provides a consistent model for quality-of-service control, regardless of whether the underlying Layer 2 technology is Ethernet, frame relay, asynchronous transfer mode or point-to-point protocol.

Flexible bandwidth Class-based queuing can let satellite providers offer the explicit bandwidth controls they need to deliver a committed information rate service to customers, often at premium prices. Remote service providers that can create a hierarchy of user traffic classes can select other value-added services. For example, a remote ISP can benefit from volume bandwidth discounts by buying excess bandwidth and then reselling committed rate bandwidth services to secondary ISPs in their regions via terrestrial links.

Bandwidth borrowing enables satellite services to be offered more efficiently because service providers can optimize link bandwidth costs. For example, shared broadcast satellite services can carry traffic destined for several ISPs over a single shared link. While each user receives a minimum bandwidth commitment, periodic bursts in demand can be accommodated by borrowing from idle or unused bandwidth on the link. This is more cost-effective than provisioning extra bandwidth.

Bandwidth borrowing can therefore be offered as a value-added feature, providing downstream ISPs access to idle bandwidth to support bursts in demand. As demand grows, borrowing also becomes an important planning tool.

Satellite providers can monitor borrowing activity-such as the frequency of attempts and the number of failed attempts-to convince downstream ISPs when it is time to upgrade to faster connections.

Software implementations based on class-based queuing can capture downstream use statistics and provide downstream ISPs with the information they need to justify increased bandwidth investment. The software must continuously monitor performance and collect statistics to meet service commitments for each traffic class.

This advanced monitor and control capability lets satellite providers track and bill access bandwidth while anticipating needs for additional bandwidth. It also offers the potential to deliver voice over IP by assigning dedicated bandwidth to voice traffic flows to prevent dropped blocks during transmission.

Applications via satellite Remote ISPs can evaluate several kinds of Internet access satellite services. Two-way satellite services require satellite transmission capabilities in the remote region and the U.S. with class-based queuing traffic management at one or both locations. This application is preferable in regions with little or no terrestrial infrastructure for high-speed Internet connections.

One-way satellite services can be delivered via an outbound satellite link from the U.S. with a local terrestrial link to the Internet. This service reduces both the latency caused by satellite delay and Internet access costs.

The traffic management software can shape traffic on the outbound link, with return traffic coming over lower-cost local terrestrial links from other service providers. This application requires asymmetrical bandwidth management and the ability to shape traffic based on flows in a single direction.

Asymmetrical traffic management is important because in many parts of the world it may be more cost-effective for ISPs to download higher-speed bandwidth via satellite while uploading information to the Internet via relatively lower-speed, land-based infrastructure. It is important to select traffic management technology that is not dependent on bidirectional traffic flows so remote ISPs can have maximum flexibility in provisioning Internet bandwidth.

Another application is a broadcast service, in which outgoing data destined for multiple sites is carried on a single, shared broadcast link rather than on dedicated links. This approach optimizes bandwidth use and reduces ISP access charges. It leverages the bursty nature of Internet traffic by efficiently carrying traffic destined for several users over a single connection.

The downstream difference Downstream ISPs can create new revenue opportunities by buying excess bandwidth and serving as primary ISPs that resell bandwidth to secondary ISPs in the region. For example, an ISP can provision a 768 kb/s asymmetrical link and resell 256 kb/s of the bandwidth to secondary ISPs in the region over terrestrial lines.

Each secondary ISP can be classified by the traffic manager as a "child" of the primary ISP, and the satellite provider can allocate bandwidth to ensure that each secondary ISP receives its committed service levels. Secondary ISPs receive committed service levels, and both ISPs can borrow idle bandwidth during peak periods.

As downstream service providers increasingly resell bandwidth, they will require more control of their access connection. As they feed more secondary ISPs, they will eventually require local traffic management at the satellite downlink using a traffic manager to control traffic flows. This application lets regional ISPs worldwide download excess bandwidth and resell it to secondary ISPs while controlling local allocations in real time.

The downlink ISPs can tune bandwidth allocations in real time for secondary ISPs, enabling them to offer variable class-of-service levels with alternate pricing schemes. The primary downstream providers creating the managed bandwidth services will require increased operational control over access bandwidth so they can deliver multiple service classes and dynamically allocate bandwidth according to the secondary ISPs' demands.

Therefore, service providers at the downlink can more effectively manage the Internet bandwidth they are buying in bulk, creating variable pricing and service schemes for secondary ISPs.

Users worldwide require increasingly higher speeds for Internet access, and managed bandwidth services via satellite can help regional ISPs gain dominant positions in their local markets. However, these applications require advanced traffic management control.

The satellite provider must be able to flexibly classify traffic for transmission, and ISPs at the downlinks will require local management control so they can implement adjustable pricing and provisioning services. Satellite transmission with traffic management using class-based queuing can provide the management control, flexibility and scalability required to support ISP bandwidth to enable a truly global Internet.

The National Research Group at Lawrence Berkeley National Laboratory developed class-based queuing as a superset of existing queuing technologies. It enables configuration by three complementary sets of parameters:

Familial relationships. Traffic flows are hierarchical, with flows defined according to their place in the hierarchical tree. Characteristics of new flows are inherited from parent flows, and each flow is placed within the established hierarchy.

Classification. Traffic can be classified by protocol, department, Internet protocol subnet or IP address. Multiple attributes can be applied to a traffic flow for further prioritization. Flow definitions and packet-matching criteria are defined to route inbound packets to a particular class.

Link-sharing parameters. Class-based queuing lets service providers control the amount of sharing within specified bandwidth management limits. Classes are assigned bandwidth rates and priority levels, and parameters are established to define the elasticity and maximum allowable size of the queue. It is more practical for managed bandwidth satellite services than earlier technologies such as priority queuing, fair queuing and weighted fair queuing.

Priority queuing is a straightforward method of ranking traffic flows. Starvation-a lack of deterministic behavior-is a common limitation of simple priority queuing mechanisms. Because class-based queuing enables committed bandwidth for even low-priority traffic flows, it guarantees service to all remote Internet service providers.

Fair queuing assumes that traffic can be effectively separated into well-identified flows so that each flow receives an equal share of transmission bandwidth. Weighted fair queuing improves on fair queuing by enabling service providers to allocate bandwidth on a variable basis. Weighted fair queuing does not let the user specify a way to allocate unused bandwidth, while class-based queuing frees up unused bandwidth for borrowing by other traffic flows.

Unlike class-based queuing, weighted fair queuing is more of a sharing solution and does not integrate explicit bit-per-second rate controls to guarantee or limit bandwidth. Class-based queuing provides increased flexibility over these older approaches and delivers more granular bandwidth control and more explicit bandwidth management that lets providers optimize allocations of satellite capacity.