|Network Management Zone|
The Importance of Scalability and Performance in an NMS Framework
In my last blog, I explained the importance of having a highly flexible NMS framework that allows you to preserve the same management system regardless of the hardware, OS, and database you use, or even the telecom domain that you are in. I’d now like to discuss the performance and scalability aspects of an NMS.
To begin, system scalability is open to interpretation – it really depends on your vantage point. For example, one of our cable customers has millions of cable modems deployed, and yet the scale of that management network is comparable to another customer of ours who manages dozens of cross-continental optical devices in the long haul space.
So what’s the difference? Well, in the cable operator case, you’re managing millions of $25 devices and each requires only 10 management traps. The optical network, on the other hand, is a string of devices spaced every 50-100 miles and each of those devices has thousands of subcomponents with hundreds of traps to be managed. Every telecom domain is different, but our base dominator is the number of Managed Objects. Most people fail to see this.
A scalable NMS framework can support and be modeled to handle both of these domain environments. What’s more, you can build small and ramp up gracefully to a high scale deployment when the time is right. And that’s a big improvement over years past when you needed to build two separate applications.
Now the point of having a highly scalable system is to allow you to achieve high performance when your network elements grow. In the context of an NMS, high performance is usually seen as a function of the fault system and how many events per second can be processed and persisted in the database. At Zoho, we have seen systems pushing 2,000-3,000 traps per second. And that performance is typically achieved by architecting a distributed system with multiple polling engines – and often adding multi-threading to each collector.
It’s no surprise, then, that a high-performance NMS can be rather costly. The deployed cost of our WebNMS framework, for example, can range from single commodity box of $1,000 or go up to $100,000 or more by throwing more hardware and expensive databases at it. That’s quite a spread. The final cost ultimately depends on the type of management application and how it’s sized.
When you add in the cost of hardware, database and labor, a high-end performance NMS can add up to millions of dollars. However, Tier 1 carriers are willing and expect to pay that price because it translates to the high reliability and performance their customers demand.
On the reverse side, not all systems are large and need to be hardened. Using the same management software, you could manage several dozen devices and applications with a Postgres database running a single commodity Windows or Linux box. Cost is almost a non-factor, yet the result is a highly customized management system.
So scalability/high performance is the second set of factors you need to consider when selecting an NMS framework. And that leads us to high availability, the subject of my next blog...
Eric Wegner is a 20-year veteran of the industry and has 10 years of experience with ZOHO Corp . (formerly AdventNet) working on large and complex network management infrastructures for network equipment manufacturers, service providers and military contractors. Eric joined the company as the first sales person and is now business development manager leading the WebNMS division in North America.