1 June 2020
So you have this idea for a killer app, you tested the market and found solid demand. Your design team has created amazing UI mocks and your dev team has coded a very compelling MVP. So far, you’ve managed to operate on a shoestring budget. But you’re ready now to launch your startup
Time to go for it! You and your co-founders meet to go over what it will take to launch and get to the next level. How much money do you need to raise? This familiar path taken by so many startups was one we at at HTML5VideoBank ourselves took.
Not so long ago we were sitting in Chuy’s and over their delicious Creamy Jalapeño Dip, talking about launch plans. Being an application company we didn’t have any physical products to worry about. There was no manufacturing, distribution, or product returns. But since our product was going to be a service, our IT infrastructure would have to be world class right from the start. It would be the one key thing that could make or break our company.
MVP is an Agile Programming term. It stands for minimum viable product. It’s your service in its minimal form, without all the bells and whistles that a final release might have. The idea is that the MVP is given to a set of potential customers and their feedback is used to improve and fix the MVP. Several iterations of this test/fix and you not only iron out most of the major issues, you also get a sense of what users love and what they hate about your product.
Fresh from the success of our MVP we were eager to plan the rest of the business. Product Managememt wanted global 3-nines (99.9%) uptime. Operations and development were asking for CI/CD, automatic failover, and scalability to millions of users. Oh, and full dev, uat, prod pipelines, expandable petabyte of fast storage as well. Marketing insisted they had to have a fast, responsive UI, and state of the art security.
All of them were reasonable requests for a modern, global app. And since this was going to be a video processing service there would definitely have to be high end CPUs in the mix. If you’ve ever edited movies on your laptop, and seen how hot it gets and how hard the fan runs, you know what I mean. Now imagine a service transcoding videos for potentially thousands of users. We founders glanced at each other. We were all thinking the same thing. How much?
Someone at the table happened to have an IT infrastructure sizing spreadsheet on their laptop. It was pretty old and thoroughly obsolete but we entered our requirements in anyway, just to see what would come out. It was a bit of a surprise that the estimated cost to build our IT infrastructure back in 2007 would have been $5.3 million. The architecture would have been based on physical servers in redundant data centers. Public cloud computing was only just coming out and wasn't included in the spreadsheet.
We already knew physical servers and datacenters wasn’t the direction we wanted to take. HTML5VideoBank would be built on micro services running on a Function as a Service or FaaS platform. FaaS is sometimes called ‘serverless’ but I think that term is a bit confusing. There are servers involved in running the workloads, just that the ops and admin aspect is automatic and hidden. Our architecture would also be Multicloud based. This decision is another no-brainer. We wanted to take advantage of best-in-class cloud services and one single public cloud provider does not have all of them. AWS might have some that are best, Google Cloud or Azure might have others. You don’t go into a grocery store and buy the same brand of everything. Cloud services are no different.
Putting together our requirements and getting started on multiple public clouds cost us a grand total of $273. A startling difference between the 2007 cost to start - $5.3 million, versus $273 now! Granted operational costs on the cloud will be higher but it does show how the economics of IT have shifted with cloud computing. IT costs have changed from being a capital expense to an operational expense. It costs next to nothing to get started. All you need is a great idea and the know how.
Time to add another key metric - cost, to Jason McGee’s serverless infrastructure chart that I first talked about in Economic Theory and Cloud Computing.
If you're still putting up MP4 videos on your websites, it's 2020! Please check out HTML5VideoBank and learn how the pro's do it using adaptive bitrate (ABR) streaming such as HLS, MPEG-DASH, MSS and the latest, CMAF.