[Contents] [Prev] [Next] [Index] [Report an Error] [No Frames]


Consolidated SRC Installation on Solaris Platforms

All the previous scenarios provide top reliability because all the network services—that is, the SAE and RADIUS servers—as well as Domain Name System (DNS) and Dynamic Host Configuration Protocol (DHCP) servers and NIC hosts, are at the same site as the router and are connected by a reliable LAN. However, to maintain this reliability, hardware must be dedicated to this function in every POP, no matter how small, and economies of scale cannot be achieved through consolidation in large hosts.

The SRC software also supports a deployment scenario that allows a trade-off between consolidation of components in large hosts and the risk of less reliable MAN/WAN connections between sites. This scenario, shown in Figure 25, consolidates the network services in regional data centers. Here, the regional data center has:


Figure 25: Consolidated Network Services

Redundancy Schemes

The N to 1 and N to M redundancy schemes are even more important in regional data centers because a server could be serving a very large number of subscribers.

RADIUS

Because RADIUS is stateless, it is enough to configure a sufficient number of RADIUS servers for the load and configure both the routers and the SAE to load balance across them.

NIC Hosts

Regional data centers may or may not have one or more NIC hosts. It is up to service providers to add enough NIC hosts to achieve the desired level of availability and performance.

COPS Connection

For the Common Open Policy Service (COPS) connection between the SAE and JUNOSe routers, special care must be taken. During a failover, existing activated services are not affected; but subscribers cannot log in, activate, or deactivate services until failover synchronization is complete. Thus, it may be desirable to configure multiple SAE machines (for example, tens) in the regional data center to limit the number of subscribers served by any one machine. The JUNOSe routers can be configured with primary, secondary, and tertiary COPS servers, so it is possible to configure many failover schemes. It is also possible for SAE servers to redirect existing and new COPS connections to other, more lightly loaded SAE servers. This COPS connection redirection can be triggered manually during a scheduled maintenance window or automatically based on SAE load monitoring.

Adding or Replacing Hardware

Startup is simplified because there is always a pool of SAE hosts to manage any new routers as they are brought online. In the case of a disastrous server failure, the offending hardware can simply be removed and replaced as time and resources allow. Also, in regularly scheduled maintenance windows, incremental software upgrades can be achieved in the same fashion.


[Contents] [Prev] [Next] [Index] [Report an Error] [No Frames]