On Monday, December 10th, 2012, OLC attended Prince Building Tech Talks Meetup: Bootstrapping Your Startup in the Cloud with Andy Parsons of Happify and Meghan Gill of 10gen.
Andy Parsons, CTO and co-founder of Happify called himself a "startup junkie and an enthusiast." He specifically stated that he is not a "DBA or a heavy-duty sys admin." Parsons gave an in-depth presentation on scaling clouds for startups.
Parsons defined cloud as "a delivery of computing as a service rather than a product." Resources are "shared and provided over a network." The impact of cloud has changed economics. "It's democratized entrepreneurship," he said. Cloud has also changed the venture capitalist landscape. "You had to sell a vision before you got funding," Parsons said. "Raise money to build something before cloud. Now it's the opposite. You build something to raise money. You can now start a startup through bootstrapping."
He outlined the essential aspects of cloud. There is self-service on-demand, which gives provisions without humans and no ownership of servers. Next, pooled resources, which are multi-tenant and location independence, then rapid elasticity—provision and destroy—and measured and metered service, which provides user and cost transparency. "Know what you're paying for," he said.
"There are a couple of service models in cloud," Parsons stated. He presented SaaS (Software as a Service), "which is Google Apps," PaaS (Platform as a Service): "Drop code, watch magic," IaaS (Infrastructure as a Science), "building blocks." Parsons said that Google is pivoting to do what Amazon is, which is "doing, growing and understanding what you—the user—are doing."
He also outlined some development models. "Private cloud models mean that there is one single customer. Oracle is getting into this space. The cloud is owned or rented and it can be on premises or off. Parsons touched on public cloud: "This is a typical startup's situation in EC2 (Amazon Elastic Compute Cloud). It's available to anyone and it is serviced off premises." Finally, Parsons revealed a hybrid cloud, which is "two or more cloud infrastructure providers bound together. It can also be hybrid multi-tenant and dedicated resources."
Parsons asked the audience what is essential to cloud. He answered for the audience. "Here's an example," he said. "Web servers scale up and down to meet the demands of offline data crunching. Cloud needs recommendation engines, social graph applications and giant search indexing."
"Now, how well does cloud work?" Parsons asked. "This all works really well until something goes wrong. Failure of some kind is assuages and performance drops. It could be an instance failure—could be server, or someone tripped over a wire, you never know. There's no way to diagnose the problem, like a vanishing server. It could be network outages or EBS RAID bottlenecks. The moral here is, architect yourself properly."
Parsons gave an example of a massive Amazon Web Services outage. "The majority of servers are located in Virginia. Not a lot of people know this. On April 21, 2011 there was a massive AWS outage. There was some data loss and it revealed details on how Amazon's Elastic Block Store (EBS) worked at the time. This was all triggered by a router upgrade!"
Parsons tackled the problem of using Disk I/O. "It's unpredictable," he said. "There's no guarantee and I call it the 'Bane of Darkness.'" On improving it, Parsons suggested that the price drop and have the platform accelerated by the Google Compute Engine. "Add an SSD, which is already in AWS Dynamo DB. Improve virtualization. Provide ever more interesting Platform as a Service offering. Begin the convergence of Infrastructure as a Service and Platform as a Service."
On expenditures, Parsons said, "If you buy hardware it's a lot of money. Cloud-based software is much cheaper and attractive for startups. In the long run, however, it's much cheaper to use hardware than cloud. Elasticity is key here." He admitted, however, that price comparisons are tricky. "Just remember that billing is a means of differentiation," he said.
In closing, Parsons said,"Plan ahead for failure. Some don't do it and it becomes a catastrophe. Plan ahead for growth economics. Automate as much as possible and don't treat cloud like hardware."
Meghan Gill of 10gen briefly talked about the New York City tech community. "Startups prosper in some places," Gill said. "New York tech is growing, but there are very few enterprise software companies. There is a community, but it's very small." She asked if this segment could grow. Gill listed four opportunities: engineering talent—bring talent to the East Coast and make NYC a viable option for potential tech hires, early adopters—allows for scalability, partnerships—funding, and customers—source of income.
Early adopters are startups and web companies. "A critical mass of consumer and media startups in NYC are early MondoDB adopters (foursquare)." On partnerships, Gill admitted, "a lot of MondoDB partnerships are on the West Coast. It's nice to be able to just drive down the street to meet the partners. This is a big disadvantage of being on the East Coast, but New York City is a great place to sell platform."