Cisco Cisco Prime Virtual Network Analysis Module (vNAM) 6.3 White Paper

Page 1

Network Visibility in the Data Center:

Best Practices for Staying Ahead

Executive Summary

As data center architectures make the transition to highly virtualized, highly dynamic, internal cloud

forms, traditional methods for monitoring and troubleshooting infrastructure health and performance

have struggled to make the transition. Monitoring activity from the network perspective is an

essential source of operational intelligence, but monitoring tools must provide sufficient capacity and

flexibility to handle rapidly growing traffic volumes and a dynamic, ever-changing resource landscape.

This ENTERPRISE MANAGEMENT ASSOCIATES® (EMA™) whitepaper examines the specific

challenges of network-based monitoring in modern data centers and reviews the most recent evolution

of the Cisco Systems Network Analysis Module solution in light of its ability to meet the new and

emerging requirements.

Evolving Needs of Monitoring in the Data Center

If there is a constant within the realm of today’s data center infrastructures, that constant is change.

Change is becoming more frequent all the time, along many dimensions, all at the same time. Compute

is increasingly virtualized, allowing servers to be turned up, moved, and decommissioned on a moment’s

notice. Storage virtualization is evolving towards programmability, allowing automated allocation and

distribution of capacity. And networks are fundamentally changing, moving away from dedicated

physical infrastructures and towards a mix of physical, virtual, and programmable services. All of this

is happening in order to keep up with ever-accelerating rates of application development, change, and

deployment.
The arrival of cloud services adds another interesting wrinkle. The use of external cloud services

aggravates the change equation, as the normal checks and balances associated with infrastructure/

equipment procurements are cast to the wind. Further, cloud services are often engaged by line of

business as a means to avoid perceived lack of responsiveness and support from traditional IT. Political

and control issues aside, the bottom line is that external cloud, whether private or public in nature, adds

additional change variables together with even less visibility for monitoring and operations assurance

than is possible with private, wholly–owned data center infrastructure.
The response to these pressures is in some ways revolutionary, and in other ways evolutionary. Some will

consider software-defined, programmable approaches to infrastructure to be a true revolution, especially

when contrasted with fixed legacy predecessors. But this path really began with the many incremental

steps taken in virtualizing networks, storage, and finally compute, and thus is in many ways more of

an evolution, in reality. Of equal or perhaps greater impact are the changes in best practices, workflow,

and organizations that are accompanying the next generation data center technology changes. EMA

has been following the steadily growing percentage of operations teams that are organizing around

application-aware monitoring and troubleshooting practices based on cross-domain, service-centric

management paradigms. Again, for some organizations, this is a radical departure from prior, silo-

centric approaches to operations. But other organizations, particularly those who had embraced some

form of integrated service management, such as ITIL, COBIT, or ITSM, will see this shift as an

inevitable evolutionary step.
So what can networking engineers, managers, and operators do to prepare for this service-centric,

application aware approach? The key lies in finding and leveraging as much data as possible about exactly

how the data center infrastructure is performing, in context with the applications in use and the functions/

purposes behind the activity. While EMA recommends that a mix of monitoring approaches be used,

one of the most consistently reliable and valuable viewpoints is that of application-aware performance