This year’s AWS re:Invent was bigger than ever with ~32,000 attendees. A highlight at the start of each re:Invent is a presentation by James Hamilton, VP and Distinguished Engineer at Amazon Web Services, affectionately called “Tuesday Night Live at AWS re:Invent.” You can view the recording of Hamilton’s ode to data center geeks everywhere below.
I’ve also provided a recap of his talk in this blog post If you want a short digest of the presentation. And if you are interested, you can also read my recap of the first keynote here and my recap of the second keynote here.
This session is typically an opportunity for Hamilton to give the public a behind the scenes look at some of the inner workings of AWS and to show off their technical innovation chops. That was definitely front and center with updates on things like Amazon’s network architecture and transition to renewable energy in their data centers. However, Hamilton was also clearly targeting his message to enterprises and Fortune 500 companies, specifically those who are considering AWS or are just dipping their toes in the world’s largest cloud. The not so subtle message Hamilton delivered to enterprises were, “You and your vendor partners can’t even come close to what we can build and maintain; so stop trying to build your own data centers and come to AWS.”
Hamilton started on this theme immediately in his talk with a comparison that shows the scale of AWS.
The implication is that AWS has and is adding enough capacity to handle the workload of any number of Fortune 500 companies and then some. The pitch to enterprise are that by moving to AWS, they can relief themselves of the undifferentiated heavy lifting of data center management and focus on creating business innovation. AWs promises customers that they can go fast, scale their resources up or down as needed and do it for “a whole lot of heck cheaper” than if they did it themselves.
To further underscore their scale, Hamilton then went through a series of slides that showed the high-level architecture for AWS.
AWS currently has 14 Regions worldwide and plan to add 4 more. A Region is a distinct geographic location where AWS has data center presence. Within each Region are mulitpel Availability Zones which are discussed later. A growing number of AWS Regions across the global are critical for enterprise adoption because it can address issues such as latency and data sovereignty.
Connecting all these Regions is the Amazon Global Network with redundant 100GbE links that stretch across the globe. Building something similar for on-premises infrastructures would be very costly for any single enterprise.
Drilling down into the Regions themselves, Hamilton tried to convey the scale and capacity of AWS. Each Region is composed of 2 or more Availability Zones and there are currently 38 of those across those 14 Regions. An Availability Zone or AZ consists of one or more data centers, each with redundant power, networking and connectivity and hosted in separate facilities. The concept of AZs are important to note because applications can be architected for high availability by deploying them across multiple AZs.
An Availability Zone can span multiple data centers, with an AZ potentially spanning across as many as 8 data centers. Some AZs host as many as 300,000+ servers. This gives AWS a total footprint that is larger than the next 14 public cloud providers combined. It also demonstrates their ability to scale to meet the requirements of any enterprise.
At this level of scale, AWS claims they have no choice but to create custom technologies for their data centers. This runs the gamut from compute to networking to storage.
This is the only way AWS can gain the cost efficiencies, performance and scalability needed to successfully run their cloud. It It also allow AWS, claims Hamilton, to build more reliable data centers since they don’t have to rely on typical enterprise vendors who build overly complex products that are filled with bloat. Again, the point being made is that AWS is on the leading edge of technological innovation and data center management. So enterprises should not try this for themselves but rely on the experience and expertise of Amazon Web Services.
Having proven, Hamilton believes, AWS’s superiority in managing data centers at scale, we move on to one of the elephants in the room when it comes to traditional enterprises moving workloads to the cloud – “What about the mainframe?” Many enterprises still run mission critical applications on their mainframes and that is a workload you don’t just move to any cloud, including AWS.
To address this, Hamilton introduced Navin Budhiraja, Senior VP and CTO of Infosys, who talked about their new Mainframe Migration to AWS service offering.
This offering by Infosys helps customer solve three problem, claims Budhiraja, that enterprises are facing if they do not migrate off the mainframe to the cloud – escalation of costs, lack of agility and shortage of skills. Clearly, this was another way to tell enterprises they have no excuse for not migrating to the public cloud.
Going back to the theme of scale, Hamilton then introduced Tom Soderstrom, CTO at the NASA Jet Propulsion Lab. Soderstrom talked about their mission to answer the big questions of the Universe through initiatives such as the Deep Space Network.
Soderstrom explained that only a public cloud platform like AWS could satisfy the compute and storage requirements of the Jet Propulsion Lab. This is a scale that few enterprises, even the largest, would need or could build.
Moving on, Hamilton introduced Dr. Matt Wood, General Manager of Product Strategy at AWS. Dr. Wood talked about Machine Learning and its application for Artificial Intelligence and Deep Learning. He iterated that to address the challenge of Deep Learning requires scalable infrastructure.
The evening ended with an update by Hamilton on their progress towards becoming a 100% renewable energy business. The progress made by AWS in this area is noteworthy.