Net/Systems Management /

Performance and availability management blend in new product generation

By John Morency
Network World on Network/Systems Management, 01/18/99

In previous columns, I've discussed the two product segments that collectively support network bandwidth management. These include network traffic shapers (such as the PacketShaper from Packeteer and the AC product line from Allot Communications) and Web traffic load balancers (from vendors such as Arrowpoint Communications, F5, IPivot and Network Engines). As the segment name implies the primary benefit of both sets of products is the delivery of improved network application performance that is more consistent and predictable. This benefit is delivered through active traffic shaping and rate control (in the case of traffic shapers) or through traffic load sharing across a set of discrete back end servers (in the case of Web traffic load balancers).

As we enter 1999, performance is just one of the benefits being promised by a new generation of Web traffic management products. Note the use of the term "Web traffic management" vs. "Web traffic load balancer." The reason for this shift is that at least two segment players, HydraWEB Technologies and Alteon Networks, have recently announced extensions to their product lines that support high availability and disaster recovery features as well as traffic management.

In the case of Alteon, new functionality includes the ability to load share a back-end Web server cluster across both the local and wide areas. In its new switch software release, ACElerate 5, support exists to direct traffic to back-end servers that may reside either on a local segment or on a remote IP subnet across a routed backbone leveraging both IP and MAC address information. While WAN redirection may not necessarily be the most desirable approach in steady state operation for performance reasons, its support becomes extremely valuable in the event of single server failure or complete disaster recovery for an entire server cluster. A new inter-switch protocol called the Distributed Site State Protocol (DSSP) is used to enable an ACE-switch network to maintain a reasonable degree of inter-switch state synchronization. The potential benefit is more globally consistent performance and an increased set of options to exercise when back-end server failures inevitably occur.

HydraWEB has also recently announced local and wide area server cluster support in its Hydra Hydra 100 (HH100) product. In addition, HydraWEB takes a server based agent approach in which HydraWEB-supplied agents execute on each individual server within a logical cluster in order to provide both performance scheduling and availability status information to central switches. The result is a dispatching algorithm that takes into account both network and application information in both load balancing traffic and managing back end server availability. The user tradeoff is use of proprietary vendor features for improved functionality. In addition, HydraWEB also allows the system or network manager to implement a local performance or availability management policy to complement the policy management services that are already supplied as part of the product.

Extensions such as these for improved performance as well as availability services will likely characterize an increasing set of products as we move through the coming year. I welcome the input of any vendors who are currently in the process of developing similar products or users with experience in using these types of products as the content basis for a future column update.