Saturday, 8 January 2011

I've actually done a fibre cost-benefit analysis

You may not know this, but most state governments are sitting on big fibre networks. Quite a few government utilities have rolled out fibre on their own over the last few decades to support their own operations.

For instance, the electricity transmission companies have strung it along their pylons. The RTA has run it along motorways. I'm sure the water mob have laid fibre alongside their pipes. Same with gas. The railways have run it alongside every bit of track in the metro area (not worth it out in the country - surprise, surprise). The railway signalling network used to run over a copper network; most of it is now run over fibre. Not that signalling needs a lot of bandwidth - I remember being told once that it needs 9600kb at most.

Back in my public service days, I worked with a government agency on a cost benefit analysis for hooking up to one of these fibre networks. The fibre was already in place - it ran past most of the larger offices for this agency, which had premises all over the metro area. It was just a matter of tapping into the fibre and putting in the right sort of network kit.

Doing the costing was easy - we needed so many switches at such and such a cost. eg - 50 sites, two switches per site, $10k per switch. Pretty simple maths.

We brought in a cabling company to estimate the hauling of the last mile fibre tails - we needed to bore under roads and pipes and so forth, and that costed up to $80,000 (depending on the difficulty and length of the bore). We also needed to cost in the professional services required to install and configure all that stuff, along with additional power outlets in some places and all sorts of trivial bits and pieces.

Many of the benefits were also easy to quantify. Every office had a small server room. That server room had its own air conditioning and an uninterrupted power supply (and both needed maintenance and had finite lives, and the air con units had to run 24x7). Depending on the location, it would have anywhere from 2 to 10 servers, and all were life expired (ie, old and crap). We added up the cost of centralising hundreds of servers into a single data centre, where a trivial number of servers would be required, versus replacing the whole damned lot (which had to be done anyway as they were cactus). The cost of the existing backup system was incredible - each site required a set of daily, weekly and monthly backup tapes, and backup tapes and tape drives aren't cheap. I think the saving in backup tapes alone was well over $100,000.

Then there was the saving in operating system and application licences - they went from dozens of email servers to just a few. The number of backup software licences was also cut dramatically - they were over $1,000 a pop.

The costing started getting squishy when we got to the technical support aspects - how much would we save on having engineers driving all over the countryside supporting all this kit? That was hard to pin a number on, but we did our best.

In the end, I think we calculated that the project would pay for itself in about 9 months - it was a complete no-brainer.

In typical public service fashion though, the manager of the server team vociferously opposed the project as it would cut into his empire. He preferred to spend millions of dollars of our money on replacing all that kit out in the field because it would make him look more important.

Thankfully, it went ahead. I came back later to see what benefits had been realised, and how the costs had gone. Funnily enough, for a government project, the costs actually came in almost smack on the money. The benefits were realised, along with a couple of unexpected ones that we hadn't thought about - they managed to get rid of two vehicles and some satellite support offices because they didn't need so many people out in the field anymore. They could work from a desk and support stuff remotely, which meant instead of doing 3 hours of driving and 4 hours of work, they could actually do 7 hours of work per day.

We tried to quantify the productivity improvements that the users would get from a faster network - what impact would going from a 2Mb WAN link to 1Gb have? Yes, it was now possible to move large files around quickly (which was actually vital for a couple of projects that generated massive files, including lots of video), but putting numbers on it was next to impossible. You can fudge some stuff up - we did - but you end up with very questionable numbers that fall apart when anyone starts poking them with a sharp intellect.

So it is possible to do a cost-benefit analysis on this sort of thing. It actually didn't take that long - a week at most, with input from half a dozen people. If the costs and benefits are staring you in the face, it's just a matter of crunching the numbers.

That's the problem with the NBN - the benefits aren't staring us in the face. With the Snowy Mountains scheme, I think most people could grasp the benefits of generating cheap hydro power and supplying irrigation water to farmers. With the Harbour Bridge, people could grasp the improvement over catching a ferry to North Sydney and the opening up of land to the north for housing. I'm still trying to grapple with what the NBN is going to deliver.

And before people start going on about health and education, after we finished that fibre rollout for that government agency, we had a look at doing all the hospitals and schools that were within easy reach of that fibre network. It was a no-brainer as well - we were raring to do it. However, it all fell over thanks to inter-agency bickering. We could have had most public hospitals in the metro area and lots of public schools on a 1Gb network 6 or 7 years ago, but the political will wasn't there to ram it through.

No comments: