What you can't see going on in your networks can cost you. A lot. And we've learned the hard way from VoIP that networks matter to the bottom line -- even though the industry still struggles to define how to measure that cost.
The biggest problem has always been visibility -- it is almost impossible to determine exactly what the network is doing. Even the most experienced network engineers rely on crude tools like ping and traceroute to "see" down the network path. But those simple tools don't show very much (see, What Ping doesn't tell you). Today, network managers piece together fragments of data from device counters and packet sniffers to generate fuzzy mental images of network performance that never really reflect the immediate truth.
And how much does not seeing hurt? With network management spending approaching $9 billion worldwide, Gartner analyst Jeff Snyder determined (2004) that 85% of networks were not ready for VoIP. And with the call for pre-deployment assessment now strident and widespread, the industry has focused on achieving network health prior to introducing the application. And yet, Network World recently estimated that 82% of network incidents are reported by end users, and Netuitive stated that approximately 50% of network alarms were false positives.
And all this due to lack of visibility. What you can't see….
Fortunately, emerging from the darkness of our ignorance comes a clear picture -- it shows a three-dimensional space that is increasingly well-defined and potentially quite manageable. Two of the dimensions correspond directly to two recent developments in the networking industry:
- End-to-end visibility into networks
- Coupling Quality of Experience (QoE) to networks through application modeling
End-to-end visibility is the "horizontal" dimension in the 3-D network picture, describing the end-to-end behavior of the network. And app-to-network coupling is the "vertical" dimension. The third dimension across time is provided by the now-familiar approach of continuous monitoring. But let's look deeper at these more recent developments that have enabled the 3-D network view.
Network visibility allows you to determine what the application actually has access to with regard to the end-to-end path. Composed of hundreds of elements such as switches, routers, cables and wireless media, and finally NICs and drivers, the network path literally defines the underlying experience of the application. Without visibility, the green connectivity light and ping are the only immediate indications of what the network offers.
Visibility is an essential capability required for a range of activities from troubleshooting to performance assessment to remediation and provisioning confirmation. It doesn't tell the whole story, however.
First, consider a simplified view of the standard OSI model -- the distinct domains are clearly defined. At the bottom is the end-to-end network and all its behaviors. Next is the application layer, almost entirely localized to the end-hosts, where everything from the operating system to the user interface resides. And finally, at the top, is the fabled "Layer 8" or "wet ware" -- although often thought of as the human component, particularly in the case of VoIP, it might just as easily be a business process or another application or service.
The relationship between these layers is relatively clear as well. At the interface between the network and the application are simple parameters composed in terms of measures like loss, latency and jitter -- packet behaviors in the case of the typical IP network. And, although the application doesn't reside in the midst of the end-to-end network, it can often be modeled and subsequently embedded there. Appropriate network models for voice, video, data and transactional applications have been developed that account for their peculiarities and requirements.
Coupling the layers
Such models substitute for an actual Application layer and make it possible to translate the immediate network behavior into a measure of the QoE at the top layer. In the VoIP world, that is most often represented by MOS (Mean Opinion Score). Although a rather limited metric (see MOS: A love-hate relationship), it offers visibility vertically from the network up to the user. In this way, coupling between the network layer and the user layer can be established and maintained.
More than just a way to measure, this coupling can support the definition of the network, based on target performance requirements. In essence, network SLAs and QoS mechanisms can be directly informed by the requirements of a particular QoE instead of being composed of best-effort estimates from black-art engineering wizards. In this way, a closed cycle of network management can be implemented, from the establishment of required QoE (based on business requirements), to the automated definition of SLA and QoS for network design, and the subsequent validation of the results of the implementation (and subsequent revision of requirements and/or remediation and provisioning of the network)."
With the addition of continuous monitoring, this newly available two-dimensional image of application performance becomes 3-D and offers the complete picture of network performance. From this view, based firmly in network behaviors and delineating the critical QoE over time, the costs of ignorance can truly be seen. Poorly performing applications hit directly on the business bottom line -- if you can see it happening, you can do something about it. The alternative is unthinkable.
About the author: Chief Scientist for Apparent Networks, Loki Jorgenson, PhD, has been active in computation, physics and mathematics, scientific visualization, and simulation for over 18 years. Trained in computational physics at Queen's and McGill universities, he has published in areas as diverse as philosophy, graphics, educational technologies, statistical mechanics, logic and number theory. Also, he acts as Adjunct Professor of Mathematics at Simon Fraser University where he co-founded the Center for Experimental and Constructive Mathematics (CECM). He has headed research in numerous academic projects from high-performance computing to digital publishing, working closely with private sector partners and government. At Apparent Networks Inc., Jorgenson leads network research in high performance, wireless, VoIP and other application performance, typically through practical collaboration with academic organizations and other thought leaders such as BCnet, Texas A&M, CANARIE, and Internet2. www.apparentnetworks.com