April 4, 2018 — AWS held its San Francisco edition of its summit today, and despite what some others would say the event actually did seem built for AWS practitioners. Compared to the mass Vegas wide chaos known as Re:Invent the event seemed to push what seems to be the ever growing triad of AWS (and much of the industry) concentration that would be containerization, serverless, and machine learning.
Some of the announcements of the day are:
Amazon S3 One-Zone Infrequent Access
Werner Vogels got on stage and actually released something that every startup developer didn’t know they really wanted. “S3-ZIA” pricing allows developers to select for single area zone object storing that comes in at $.01/gb storage. Given that the < 500gb price on s3 is still .021, this means that with cross region replication, amazon has released a product that allows you to get cross region replications on your assets for less than standard s3 storage, and “I just need to back things up for a bit” storage for 4/5ths the cost of regular IA storage.
Previewed starting last November S3 Select “enables customers’ applications to retrieve only a subset of data from an object by using simple SQL expressions.” For large ETL users, this can mean a massive boost in performance due to it being the s3 equivalent of a subquery select (get it?) in mysql.
Greengrass lets application developers to easily deploy and run machine learning inference locally on connected devices, which I am excited for because I really want Jarvis to come closer to reality and to heck with my privacy getting there. (Kidding, but basically this helps you get connected devices to be more connected™)
Sagemaker sits a layer above your machine learning models. Today they released a local version, along with support for open sourced containers along with the newest in tensorflow and mxnet. In what is either a gift or a curse depending on if you optimize your ML workloads, they also now allow the p3 instance types, which could cost you a car payment a day if you want to run the largest of.
In an ever vigilant push from the cybersecurity community, Amazon has released another policy manager into the world. I see this as a good thing for the AWS ecosystem, but I look forward to a potential v2 possibly having add ons that leverage other ecosystem systems and possibly machine learning.
So personal appeal here, unless you have a system for dealing with this already, please leverage secrets manager. My personal information has been spilled on the internet enough and this makes it really easy to rotate secrets whenever you really need. Baking your passwords into your containers / Cloudformation is just not needed anymore.
For those of you with complicated multi account — multi region deployments, getting a cross org view can sometimes be a pain in the… Administration. Config tools finally releases a system to get a real look at what your infrastructure is in a singular view. So we have that going for us, which is nice.
As someone who has done a roll your own version of this with Lets Encrypt, I am excited to see ACM releasing a Private CA. For those of you who aren’t encrypting your traffic, please do that. For those of you who are, this might save you some time rolling forward.
Thats all for today, shout out to this weeks sponsor, PubNub, who is here too!