A quick Google search for ”network upgrade” shows 115 million results, while ”network optimization” showed only 21.8 million. It’s hardly rocket science, but it does reflect one sad truth about the IT industry – if your network is not delivering the service you need, then there will be no shortage of people out there keen to sell you an upgrade, but you’ll be harder put to find someone who will squeeze the needed performance from your existing investment.

Of course, there are times when an upgrade or more equipment is needed but, in my many years experience, I’ve never seen a situation where network tuning for better performance hasn’t delivered a far better return on investment than a forklift upgrade – even if it has only been used to reduce the final capital outlay.

”How much traffic can your network support?” It’s funny how few people can provide an accurate answer to that question. The usual recommendation is to build enough surplus capacity to make sure that overload is never an issue. By the time you do discover the limit, it’s usually too late.

That’s what happened to the Big Brother TV website in its first season – public demand was so high that the website crashed. So, for the second season they piled on the power to be on the safe side and not risk alienating their growing fan base – so a lot of money was spent on a system running at only 30% capacity. In this example, third time really was lucky, because for the third season they called in my company to ask the really basic question: ”what is the least cost system that will have the flexibility to handle realistic peak loads?” The answer to that question not only maintained and strengthened their brand, it also saved them a lot of money.

So what is the secret for finding what a network can handle in extremes, and for tuning it to deliver better business performance and ultimately greater ROI? Let’s reduce it to answering a few basic questions..

Tuning, rather than testing

Legend has it that Sir Frederick Handley-Page would turn up unexpectedly when a new prototype aircraft was on the runway for its test flight and climb inside. Every now and then he’d point to a piece of the structure and say that it was too heavy, or over-designed, and sometimes he even pulled a hacksaw from his pocket and pruned the offending component on the spot. He was invariably right: after all the detail work of the stress analysts, he could look at the overall structure and instinctively tune it for better performance, because he had so many years of experience.

Another aviation legend, Howard Hughes, is said to have asked for all the rivets to be polished smooth on his aircraft to reduce aerodynamic drag – a more obvious remedy, but again it needed lateral thinking to first see it.

The trouble with the phrase ”network testing” is that it suggests a negative process. You can hardly blame the customer who protests: ”I’ve just paid you a fortune for this network upgrade. Are you now telling me I need to bring in another company to see if it works?” So it is hardly surprising that the vendors have so little to say about the advantages of network testing. It is far better for business to sell an over-prescribed system and make sure there is ample capacity in hand.

But, as in the aviation examples, think of the process as fine-tuning the network and you are now in a ”less is more” situation. So the first question is this:

  • Before you spend more money on a system upgrade, how can you be sure it doesn’t already have all or most of the capacity you need?

The answer is that you can’t be sure, unless the network has been analysed under realistic volume conditions for potential performance bottle necks. The good news is that network capacity testing is easily executed and costs relatively little.

So, what is your peak load?

Another question is the one already mentioned above:

  • How much traffic can your network support?

The examination board was lucky in this respect, because they were planning for a particular eventuality: around 100,000 candidates wanting to know their results soon after they were posted online. Each candidate was required to log on and key in their individual PIN number to ensure privacy – so you might think the maximum was clearly defined. However, the examining board anticipated that a lot of candidates would pass on their log-on details among friends and extended families anxious to know what their future held in store. So even here, it took some experience and guess work to estimate the worst case scenario. What was abundantly clear, however, was the analysis of how many new users per second could be accepted, and how many simultaneous users could be supported by the system.

Often one is relying much more on guess work and old statistics. A major disaster or public event can produce a flood of enquiries that can knock-out any network, and it would be asking a lot of the system to be able to handle any such eventuality. There is however, a big difference between simply adding a lot of capacity and waiting to see if it works, and running volume tests to find out just how much traffic the network can take before it breaks down. Knowing in advance does at least allow one to monitor exceptional loads and take some pre-emptive action before the limit is reached.

A major broadcast channel like the BBC knows from experience that if it runs any popular documentary that ends with a website being posted for an organisation – let’s say a the website of an animal protection agency after a programme on animal welfare – then that site is very likely to crash from a rush of enquiries after the showing. That’s why the URLs they post are always through their own BBC website – where they will cache the contents of the organisation’s website in the arms of their own massive system where it has a better chance of being able to take the load.

Pity that nobody ever got sacked for buying you-know-who

If you are feeling a little insecure in the face of unpredictable eventualities, then it is always tempting just to ”pay for the best”. If things do go pear shaped, at least you can say you did not scrimp on the solution. An awful lot of IT budget goes this way – thousands spent so that the IT Director can sleep at night. Sleeping pills don’t cost as much!

So a third key question is this:

  • How do you know that a cheaper option might not work just as well?

The answer is, that you don’t know if you don’t try it! Spending wisely need not mean spending more – customers often find they get better results by reconfiguring existing equipment, or fine tuning less costly systems, than by splashing out on top of the range equipment. And once again, the vendor supplying your upgrade is not the person most likely to suggest putting cheaper options to the test – especially not those from rival producers.

In the case of one major UK high street bank, testing was able to identify potential system bottlenecks and allow the designers to optimise the blend of applications running in the context of the hardware being installed. An 18-month server consolidation programme providing a stringent six week programme of tests integrated into the commissioning process. Each set of tests provided data that allowed the system to be optimised at every step in a realistic production environment. Two factors were critical in these tests: the experience and skills of the test team, and the unit used to generate realistic traffic conditions with easily adjustable parameters and precise reporting of results both in real time and as finished reports.

The most important thing was to be sure that the bank’s mission critical appplications would run reliably across the new network from day one, and not be vulnerable to extreme or unexpected traffic or operating conditions. Another critical issue is the difference between ideal equipment performance figures and what is actually available under real world conditions. Major buying decisions must be informed by thorough pre-testing to make sure money is being wisely spent.

Conclusion

Three simple questions could save your organisation a lot of money, and make some IT vendors a little unhappy:

  • How can you be sure your system doesn’t already have the capacity you need?
  • How much traffic can your network support?
  • How do you know that a cheaper option might not work just as well?

The questions are simple, but finding the answers requires the right equipment and a bit of experience – especially if you are going beyond the immediate answer to fine-tune your system.