Yesterday was the start of the 9th annual Defrag conference! What is Defrag?
Defrag describes itself as: “We explore the frontiers of technology’s intersections with society, government, education, healthcare, and commerce through discussions and sessions rooted in topics like cloud computing, APIs, mobile technologies, big data, devops, the internet of things, and next generation human computer interaction.”
Neat! Read on to hear highlights from the first day of festivities.
Think this is only Sci-Fi? Not so fast. Ramez shared dozens of real world, empirically tested, examples of brain augmentation as it happens today.
The simplest is the cochlear implant. It’s a device implanted into people with a certain types of deafness to directly stimulate a nerve to produce sound. A more showy example appeared when he mentioned a University of Washington study where researchers were able to remotely control each other’s arms.
Naam then took a step back and asked the hard questions: what are the downsides to technological advances like this? Well, for one, in a totally linked IoT world, whether it’s our brains or our shirts and shoes, means we could be constantly tracked.
Naam looks to two prior technological revolutions for hints as to how this might play out. When you look at the printing press or mobile phones, two trends appear. First, the rich certainly are the early adopters. But, secondarily, that fades rather quickly.
Both technologies enabled new forms of communication. The new channels have the potential for abuse, but in the long run magnanimous uses prevail. Look to the publishing of John Locke’s book “Tolerance,” which inspired many subsequent declarations of basic civil rights. Look to how mobile phones have enabled disparate places like Ferguson, and Libya to make the world aware of their own important events.
Naam closed his talk with a conjecture that most large social shifts are enabled by new forms of communication. Communication enabled by new technology.
From Plaything to Production
He started with a look at the current lay of the land. Kubernetes, Mesos, and Docker are all hot topics. Johnston-Watt then broached the topic of the infamous “hybrid cloud.” As recent as a year ago, it’s a thing many were claiming couldn’t exist. “There’s no such thing as an on premise cloud.” Yet, we’re seeing more and more hybrid solutions (and happy hybrid customers!) these days.
This is Duncan’s hint to what makes a piece of Tech truly revolutionary. It enables a new way of doing things, but remains inclusive to the old-world at the same time. In short, disrupt, but be inclusive of the old world. Our old world is the Enterprise.
The rest of Johnston-Watt’s talk was a look forward. What’s on the horizon as a disruptive force? He coined the term VLS, for Very Large Scale distributed systems. Examples of VLS can easily be found in the internet of things and big data processing. For example, in a few short years there will be about one trillion sensors in the world.
That’s a lot of gizmos running apps. That’s even more data to process. Johnston-Watt believes Autonomic Computing, an idea from the early 2000’s, will galvanize and rise to meet the challenge. That is, self-configuring, self-optimizing, and self-healing systems. These need not be large. In fact, a lot of the philosophy peeks through and shows up in the world of microservices already.
The future is a system that monitors something through a sensor, and reacts intelligently. Duncan recommends you give the original paper on Autonomic Computing a read.
How to Do Massively Parallel Job Processing on the Cloud Using Docker
Iron.io’s very own CTO and Co-founder Travis Reeder led a talk on how workers and Docker are making a serious dent in the cloud.
First things first, Reeder began his talk with a little back story. What is a worker? A worker can be any code that runs in the background and takes more than a fraction of a second to run. This could be be image processing, ETL, genome processing, etc. The important piece is that a worker should do one small very specific thing.
As an example, Reeder took a look at a monthly billing job. There are a few moving parts here: generating an invoice, generating an email from a template, and finally sending the mail to the customer.
Try doing this in the old world and it’s dog slow. To illustrate the point Reeder offered the what-if imagine you have 50,000 users to bill, and building and sending each email takes about 5 seconds. At that rate it takes about three days to finish!
Also, what happens if the task fails mid-way? Without an atomic process, you’re stuck figuring out which users got billed. Which didn’t? You certainly can’t bill twice!
Contrast this with the worker model. Each email could be sent by a pipeline of ephemeral workers. Given an unlimited number of processes, the job could theoretically complete in five seconds.
Take the case where a job fails. Other emails can happily continue sending, and it’s easy to see exactly which emails failed.
Reeder returned to the world of concrete by adding Docker to the equation. Docker offers an easy way to partition compute resources, avoid dependency hell, and distribute your code. Remember the earlier note about “given an unlimited number of processes?” Docker makes that a near reality.
Reimagining Insurance: Opportunities in a $5T Market
Dan Reed is the Managing Director at American Family Ventures. They’re a venture wing of a traditional insurance company. Dan took to the stage to share how the internet of things has already made a big dent in the world of car insurance.
Dan kicked off his presentation with a video clip of 16 year olds nearly crashing their car. How did he get the clip? In 2007 their Teen Safe Drivers program introduced a combined dashcam and accelerometer. The camera was programmed to activate whenever the accelerometer picked up odd readings. For 10 seconds the camera, plus accelerometer readings were recorded.
The clever bit here is what happened to the data that was captured. Instead of sending the data to the insurance company, the video was sent directly to the teen’s parents!
The outcome of the program? Well, it turns out that 16 year olds are nine times more likely than their parents to get in an accident. After four months with the dashcam a teen’s risk of accident dropped by 90%. A win for informed parenting and IoT!
Another interesting use case is collecting stats on how people choose to drive. AmFam asks customers if they’re willing to attach a dongle to their car. The dongle reports statistics in exchange for a rate reduction for drivers who are found to fit the profile of a safe driver.
This is neat, but it also exposed some looming challenges for IoT. There are a lot of soft human problems around data collection, it’s essential to both provide the right incentives and assure customers you’re trustworthy.
Closing out the discussion on IoT, Dan noted the following open problem: If IoT data affects a customer’s price, then how can a customer comparison shop? Customers need to own the data. Lastly, finding a standard format and a simple and secure way to transfer it is a big, open, opportunity.
Tune In Tomorrow
We’ll be back tomorrow to share thoughts and highlights of Defrag Day Two. Stay tuned!