The Costs of Cloud Computing

Cloud computing is often touted as the “future;” increasing speed, revolutionizing the world, and enabling easy scalability.

But wheres the beef?  Examples of the cost breakdown between traditional server farms and cloud services are hard to find.  Older companies that switch from one to the other don’t share their cost analyses. New companies often don’t even consider traditional solutions, since the initial buildout can be unaffordable.   In the spirit of openness, here is our summary analysis of traditional servers versus cloud computing for our startup.

Most businesses have IT challenges specific to their industry. Being a financial advisor in the peer-lending space creates specific challenges:

• LendingRobot handles orders for thousands of clients. Each client can have up to 20 separate investment rules – which means potentially thousands to tens of thousands of transactions must be handled individually
• Speed is vital – notes are released at specific intervals, and there are tens of thousands of investors looking for a limited quantity of notes
• All data transmissions must be secure
• LendingRobot has fiduciary responsibility towards each of its clients – meaning that there can be no preferential treatment towards any specific client

The Traditional Server Farm:

A traditional server farm would solve this problem by starting with one or two servers, then expanding capacity as additional clients join the firm. Each additional server has hardware, maintenance, electricity, software licensing, and setup costs.

Every server is capable of serving 80 clients, which means each client’s “share” of the annual server cost (or a client’s marginal cost) is $43 per client per year.

On top of the server costs, bandwidth, latency, reliability and security are all important aspects of loan procurement. These required features don’t come cheap; it costs about $6,000 per month for a secure DS3 connection.

The marginal and fixed costs can be added together to achieve an estimate for the yearly cost of 1,000 clients:

The Cloud Computing Alternative

Cloud computing is on-demand processing power. Servers are stored in gigantic warehouses in the tens- and hundreds-of-thousands, which are then rented out to consumers and businesses on an hourly basis.

The cloud solution to the peer lending IT challenge is to calculate specific server demand on an ongoing basis and meet that demand by renting the appropriate amount of processing power.

This remote solution creates logistical challenges not faced by traditional server farms. Every server within a cloud computing service is standardized – which means that the software must be specifically written to run on hardware that may not be optimized for the task at hand.

In addition, software is installed onto each instance (a virtual server) at the beginning of the rental period and is removed when the server is decommissioned. This means that any custom software must be quickly deployable and scalable. Additionally, any issues with software affect every server – so it is important to nail down potential issues quickly.

The big upside is the price tag: it costs us an average of $0.0283 per hour to rent an instance. This price includes the instance, reliable and secure high-bandwidth internet, power, space and automatic backup, which eliminates the need for in-house IT infrastructure.

However, due to the complexity of the software, the up-front developer costs are greater, and it requires an on-going commitment to updating and maintaining the cloud management software.
It took our engineering department a month and a half of developer time, as well as monthly software updates to the cloud service.

On the plus side, once the software is customized it is not difficult to vary computing power based on a client’s specific needs.  This means that a smaller client will be included in fewer investment rounds, larger clients will be included in more investment rounds, and fewer computer cycles will be wasted.

Our typical breakdown of 1,000 clients look likes so:

A Comparison

As mentioned above, developing software that is quickly deployable across a scalable infrastructure is a more complex than developing software one server at a time. The return is that both the fixed costs of a cloud server platform and the marginal cost of a client are almost a tenth of the cost of a traditional server farm.

Conclusion

For LendingRobot, cloud servers live up to the hype. The cloud enables us to offer a solution that matches client demand at a much lower price point, while still serving small and large clients identically.

The old method of building and maintaining a server farm would change the business model to something more traditional.   In order to continue business, we would need to eliminate the freemium threshold and mandate a minimum account balance. This is very similar to how older-style financial service companies operate.

Our experience with the cloud is replicated across many startups, and the cloud is changing more than just cost – cloud computing is also utilized in every aspect of the business. The cloud enables services like ours to exist. The culmination of thousands of startups trying to find their personal niche is fundamentally evolving society and resulting in high technology brought to the masses at an affordable rate, raising the standard of living for everyone.