Thursday, July 2, 2009

eDoorways Part 4 - Technology Scalability

Don't Panic
This section is really more about preference than objectivity. I am personally less likely to invest in a company that has a higher operating cost per page served than its closest competitors.

For example, say Massive Motors spends $8,000 to produce a car that sells for $15,000. Their competitor Nimble Wheels spends $6,700 to produce a similar car that sells for $14,000. Who would you invest in based on that limited information?

The same principles hold true on the web. Google manages hundreds of thousands of machines to operate their global network. They rely on open source (free) operating systems, an open database product (home built), and uses a mix of free to use languages that do not need to be licensed (C, Python, among others). Google is able to deliver a cost per search that undercuts Yahoo and Microsoft. As such, they have more leverage in negotiate ad deals.

Potential competitors to eDoorways like Hunch, Yahoo Answers, Experts-Exchange, and GetSatisfaction all use open source server platforms like FreeBSD or a Linux variant. eDoorways will run on Windows, so every server in the farm will cost more to license and operate.

To make matters worse, the software for eDoorways is being written in a .Net flavor, so it will always require a Windows server underneath. The competitors listed above use Ruby on Rails, PHP, Perl, or Python, so they can shop any host and any platform that meets their needs.

It is often argued that .NET is cheaper to program in, but arguing about programming languages is a lot like arguing which Religion is the True Religion. It really has more to do with what your team is comfortable with, what your consumers need, and what is the most feasible selection for your company to operate. One thing is true for shops running a Microsoft stack: They burn through more cash than those companies using nimble alternatives.

So the investor takeaway on this shouldn't be anything more than being aware that some operational dollars are going to be eaten up by software licensing while most competitors do not have that built-in cost.

Reduce Costs during launch
Many start ups also burn cash by designing their infrastructure to meet the initial peak demand of hundreds of thousands, only to have the investment wasted when the traffic settles down to normal levels.

I would highly recommend that eDoorways look to augment their standard server farm with temporary traffic load measures available through cloud computing services. Amazon AWS offers on-demand servers (EC2) starting at $.10/hr for Linux and $.125/hr for Windows. They can also host content with S3 for $.10/GB hosted.

Mixing cloud resources into a launch is a massive cost savings. Say they need an extra 500 servers for the first month (~730 hrs) to meet peak demand. Traditional deployments would require them to go buy 500 servers and have them all on and running. Figure that each server will cost a minimum of $1K. That's really conservative considering that you'll also need racks, switches, routers, storage arrays, software licenses, etc. But let's just say $500K is what they need to meet that peak demand in the first month.

With the cloud, all they'd pay is 730 * .125 * 500 = $45,625. It gets better, since this is largely a US based service, most of these machines can be turned off at night. Also, traffic tends to rise as a bell curve throughout the day and throughout the week. And the best part about the Amazon cloud is that you can script the farm to expand and contract automatically based on load. So as traffic picks up, machines turn on automatically. When traffic declines, the capacity declines to match usage. All the while they just pay for the hours they use. EDWY could easily provide massive scalability for the launch for less than $30K.

Investor Takeaway
If I were investing in this company, I would certainly ask the management about their deployment plans for the launch. Are they going to build peak capacity in house? What happens when demand exceeds capacity? What happens when the initial burst of traffic dissipates into normal usage levels? Have they heard of cloud computing?

These comments here are targeted for current investors and are designed to help you pressure the company to be more nimble in their ongoing operations. The less cash they burn monthly means the more profits that you the investor can pocket.

No comments: