I started my career with large scale NCR computers followed by Borrough’s accounting machines, then moving into a world of IBM mainframes. Over the years I’ve worked with IBM midrange computers, and Intel based servers. I recall that for many years I and everyone else subscribed to the theory that “server farms” were expensive, subject to failures, and complex to maintain compared to investing the same or less money in large scale IBM computers. Today, that is just not the truth. Even IBM is a leader in network computing operating over 40 data centers for the IBM cloud consisting of thousands of machines the vast majority being cheap Intel clones. Yes, they do have IBM Power and zSeries machines in their cloud data centers, but the vast majority are extremely cheap X86 clones. Today a big machine is not necessarily less expensive or reliable than a distributed network of computers. Interestingly enough, IBM is perhaps the largest contributor to the Apache Software Foundation and its HADOOP and related project offerings which provide the software to manage distributed networks and “Big Data”. A key characteristic of this software is distributing data as well as processing across multiple machines in multiple data centers and providing 100% failure proof reliability. You can take an entire data center or any portion of a data center from one to many machines off-line, handle disk crashes, etc. and the system will manage itself. Software developers need not be aware of the physical environment and simply write applications that conform to the guidelines for this type of environment. Mainly using JSON and being stateless. Google is the worlds leader in number of servers with over 2 million servers in their network. The Microsoft cloud has over 1 million, Facebook while secretive about their infrastructure is reported to have about 500,000 as does Amazon. There are many other companies with 50,000 plus machines and hundreds if not thousands of companies today with 500 to 1000 or more machines in their networks. For years major mainframe vendors like IBM have claimed that bigger is better and that you need less people to maintain and support a big machine than networked machines. This is just not a reality any longer. Today you can buy more computing power, support it, and deploy applications at a fraction of the cost of “big iron”. Additionally, you now have the option of utilizing major reputable cloud providers like IBM itself! Why own computers when you can buy processing and disk storage in a cloud from reputable vendors like IBM? You then have no employees supporting your data centers and computer equipment. The saving in labor cost alone is astronomical. Many people claim that security is a big risk when using cloud based computing. The reality is that if you choose a reputable vendor then they provide better security and integrity checking that you can afford on your own regardless of who you are! It is becoming very difficult in today’s world to justifying owning your own computers. If you are small business you can buy an immense amour of computing power for as little as $100 per month! You can scale up to more processing power (at an increased cost) when you need it and the cost goes down when you do not need it. Most pricing is based on a fixe monthly minimum that has set limits on various resources. Contracts are usually written to scale as demand requires and to revert to the original levels when demand decreases. As far as reliability? Choosing your vendor is the key. Make sure they have multiple data centers and that your applications and data are seamlessly replicated across multiple locations. The software now knows about other computers in the same rack, on the same electrical source, etc. Replication is routed through sophisticated algorithms to avoid outages. Additionally you can take hardware off line for failure, maintenance, or software upgrade and it resynchronizes when it comes back online. Look at reputable cloud vendors and compare the cost to what you are paying today. Look at the human resources you get with a cloud and what you have on your staff. Compare skills as well as cost.