I can remember a time five years ago when, as Webreality’s infrastructure manager, the idea of not physically having servers would have sounded like part of a bad joke.
At that time, Webreality had moved into virtualisation on our own physical infrastructure, benefiting from being able to host multiple virtual servers on a small number of larger physical machines.
However, we still had a file server in each office, our own Exchange server and a lot of small servers consuming vast amounts of power that we could never fully keep track of.
Back then it still seemed that the future was more CPU power in smaller boxes taking up fewer racks, and perhaps managed servers at large server hosting specialists. But every time you deployed a new machine, out came a box of DVDs.
The problem with owning your own hardware is that at the rate software changes, it’s out of date before you’ve even finished the gruelling configuration process. Yet there we were, spending hundreds of thousands of pounds on equipment and space, the electricity to run it all and then on the air conditioning to keep it all cool.
The story doesn’t suddenly finish “then we discovered cloud computing and everything was perfect”. The biggest cloud computing providers, Amazon AWS and Microsoft Azure started small and were devilishly hard to configure and manage. They weren’t cheap either.
That wasn’t the biggest barrier though: Amazon is an online retailer. why on Earth would I go to a web console, create a virtual server and then rely on them to look after it for me? As an IT professional of (at that time) almost 15 years I was absolutely certain that physical boxes that I could touch were the only reliable way.
Although the internet is all-pervasive, Windows web hosting is still regarded as something of a black art. Whereas Linux is lean and mean and just works, Windows requires lots of love. Linux evangelists wonder why we bother, but anyone who started to harness the power of Microsoft’s .NET framework early in the noughties understands that it could offer the entrepreneurial developer the ability to create in just a day applications that would take months with PHP.
Those who had learnt the dark art of managing Windows web servers hardly wanted to risk it all in the cloud. But about 3 years ago, I started to experiment.
You see, Amazon allow you to run a virtual server instance for as little or as much time as you like. You literally pay for it only while it’s turned on. This was where they were clever, as it meant thousands of us started to try things with their virtual machines that we simply couldn’t do with hardware. Stakeholders generally don’t like spending tens of thousands on machines that are basically experimental toys, but they don’t mind risking a few dollars.
Amazon recognised the new ways people were finding to use their infrastructure and offered us more and more services. The key to luring us in though, was that as they achieved economies of scale, they gave them right back to us. They provided us with load balancing and redundancy we could only dream of.
The biggest challenge of moving to a virtual environment is understanding the performance of the equipment. Since every specification has a “v” in front of it, the units of measurement of performance are all arbitrary. Therefore there are risks. You will require more machines, you will have to learn how to get the most from them and you will love the bandwidth. You will also love that you never have to worry about upgrading your hardware again!
It has taken almost 3 years to fully settle into AWS but to my surprise, Amazon have listened and communicated directly with us. They had envisaged at the start that a few large corporations would use their virtual offering for massive batch processing. That does happen but what has really surprised them has been how many of us have created relatively large farms of virtual servers doing all manner of different tasks.
It was natural that firms like ours would be some of the first to take advantage of cloud servers but I still hear from too many other small and medium enterprises who have ungainly, outdated physical infrastructure that they don’t need. They are missing out on time savings, cost savings, risk reduction and the opportunity to focus on the things that really matter to customers.
I lose track of how often my wife tells me the whole of her company is having another coffee break because a server has failed. The server technicians then tell the senior management the same thing: “Buy new servers.” They would - in their view, their job depends on the employer owning infrastructure.
I’d see it differently, based on experience here. They could move to the cloud for most of their server purposes and focus their time on delivering even better service to their customers.