IBM to Ratchet-Up Network Quality-of-Service

In homogeneous Token-Ring networks, Quality-of-Service (QoS) has never been much of a problem, but in heterogeneous client/server environments, network QoS is more often than not a key consideration. And because now more than ever homogeneous systems must learn to make nice with their more diversified brethren, network QoS should be a primary concern of mainframe and AS/400 administrators as well.

With the announcement of its Application Driven Networking initiative, IBM hopes to provide a simplified means for network management--and a step in the direction of guaranteed QoS.

IBM positions Application Driven Networking as not simply an initiative, but as a network policy management architecture, which it maintains is designed to substantially reduce the complexity of establishing networking policies. With Application Driven Networking, Big Blue reasons, network administrators can consolidate policy decisions at the system management level, regardless of the complexity or heterogeneity of their enterprise networks, providing a single interface for management within the enterprise.

Application Driven Networking enforces QoS guarantees by allowing network administrators to control the performance and security of their network by assigning policies enforced at the application and network levels. In this model, network administrators can choose to give important applications top priority, while correspondingly downgrading the QoS of less critical applications.

Like many management-oriented initiatives--such as the proposed Active Directory from Microsoft--Application Driven Networking is enabled by means of a directory, in this case an "e-network" directory front-end to IBM's DB2 database. According to Bruce Dillon, a development manager of policy-based network management with IBM's Networking Hardware Division, the e-network uses the common information model, which is a method of relating data objects hardware and users throughout an organization. Information can be retrieved from the e-network directory by means of the Lightweight Directory Access Protocol (LDAP).

But the strength of Application Driven Networking, Dillon maintains, is that it takes advantage of existing information. "Application Driven Networking takes advantage of application configuration information that's already in the directory," he says. "With our presence in the end systems like AS/400, the configuration information for the other systems already exists in the directory, and application-driven networking is all about leveraging that information."

According to Sam Alunni, president of research firm Sterling Research (Sterling, Mass.), IBM's Application Driven Networking initiative is but the first step in an overall plan to provide total QoS management of all hardware devices--including server or mainframe CPUs--throughout the enterprise.

"We're going to find that there's going to be some integration at the directory level of the setting of policies for business applications," Alunni says. "For example, there's a whole set of queue structures inside a CPU that can be used to implement policy as well. So when you set a policy for a bunch of users that need access to a financial application, it'll not only be the network that gets set up for the application, but the CPU resources [utilization and processor time, for example,] as well."

Alunni points to moves on the part of Cisco Systems (San Jose, Calif.) and Hewlett-Packard (Palo Alto, Calif.) as examples of other industry players who are already aligning themselves in a similar fashion. "Cisco and HP are demonstrating this right now," he indicates. "They've taken the HP internal CPU resource scheduling technology and they've integrated it at the directory level with the Cisco technology."