November 30, 2017

Episode 113: All the great AWS re:Invent news

59 minutes

There’s no clever title this week, just straight to the point of covering the highlights of AWS re:Invent this week. They got the kubernetes now! There’s a passel of releases as well. We also discuss some other news like Meg Whitman leaving HPE (on good standing), net neutrality, WeWork buying Meetup, and Arby’s. For reals!

Pre-Roll SDT News

SDT got a new logo!

SDT got 1,000 logo stickers to give away!

You can get a sticker but completing this survey or sending us your address in Slack.

US Addresses only until Matt can come and get some stickers.

We’ll be doing a live show - probably - on Jan 16 at the CloudAustin Meetup.

Check out the Software Defined Talk Members Only White-Paper Exegesis podcast

Join us all in the SDT Slack.

Upcoming SDT newsletter.

Misc. news before re:Invent coverage

Changing of the guard at HPE.

WeWork buys MeetUp.

Net Neutrality - I realize this is naive, but I feel like things already operate this way.

EFF write-up

Stratechery & follow-up

This week in PE: OOOHH-OOOO! BARRACUDA! Also, Arby’s: eat all you want you’ll die anyway.

Work in tech? Time to ask for a raise.

Good overview of the end of OpenStack’s big tent theory.

AWS re:Invent

AWS Business Update

Amazon Web Services has an $18 billion revenue run rate and the business is growing 42 percent year over year

New AWS Services (100+ new total)

Loosely break into themes of Containers, Databases, AI/ML, and IOT

Amazon MQ - Apache ActiveMQ as a Service (lunches eaten?)

AppSync - GraphQL as a Service (lunches eaten?)

Aurora Serverless - burst database consumption

Comprehend - Natural Language Processing across 98 languages

DeepLens - video camera with AI embedded

DynamoDB Global - similar to Azure/Google initiatives

EC2 Bare Metal Instances - lots of competitors try to differentiate on this (lunches eaten?)

came out of the VMware work

i3.metal instance types

c5 AMIs can work too (new KVM-based instance type)

EC2 Instance types, up to 25Gbps networking

H1 - higher throughput to storage, replacing D2 instances

M5 - 1.15Gbps write to storage, encrypted at rest, multipurpose instances, new Nitro hypervisor

Deep dive on EC2 virtualization/bare metal

T2 Unlimited - good for microservices, bursty workloads with a credit system

Elastic Container Service for Kubernetes (EKS) - called it!

upstream K8s

automatically runs K8s with three masters across three AZs

monitoring/healthchecks built in, managed service

Fargate - Containers on demand, no host/orchestrator needed

similar to Azure Container Instances

apparently Google has App Engine Flexible which is similar (thanks JP!)

So, Matt: why would I use EKS instead of Fargate, etc.? Another write-up.

FreeRTOS - AWS bought(?) existing open source IoT operating system vendor

Glacier/S3 Select - run SQL-like queries against your buckets and storage (CSV & JSON)

GuardDuty - continuous security monitoring & threat detection (lunches eaten?)

IoT Analytics - MQTT processing, reporting & storage

IoT Device Defender - reporting, alerting & mitigation of existing IoT fleets

IoT Device Management - lifecycle, management & monitoring of IoT devices

Kinesis Video Streams - video ingestion/processing service

Media Services - YouTube as a Service, including monetization. Seems there should be an embeddable player somewhere.

Neptune - managed graph database service (lunches eaten?)

Rekognition Video - Rekognition now does video

SageMaker - framework for building AI services

Sumerian - VR/AR/3D IDE and platform?

Systems Manager - custom dashboards based off of tags, ties into AWS system management tools

Time Sync Service - AWS NTP

Translate - Google & MS already have this

Transcribe - speech recognition, we should use this!

More: The New Stack, The Register.

This kind of over-the-top analysis is kinda our thing. BACK OFF, MAN!

AWS Strategy Update

On Hybrid Cloud: “In the fullness of time — I don’t know if it’s five, 10 or 15 years out — relatively few companies will own their own data centers. Those that do will have a much smaller footprint. It will be a transition and it won’t happen overnight.” Link

More: ‘Is Multi-Cloud Real?: “We certainly get asked about it a lot. Most enterprises, when they think about a plan for moving to the cloud, they think they will distribute workloads across a couple of cloud providers. But few actually make that decision because you have to standardize on lowest common denominator when you go multi-cloud. AWS is so far ahead and you don’t want to handicap developer teams. Asking developers to be fluent in multiple cloud platforms is a lot. And all the cloud providers have volume discounts. If you split workloads across multi-cloud, you’re diminishing those discounts. In practice, companies pick a predominate cloud provider for their workloads. And they may have a secondary cloud provider just in case they want to switch providers.’

AWS re:Invent Preview Review

✔SaaS lunches will be eaten?

✔Amazon Kubernetes Service?

This Week in Kubernetes

All about AWS this week!

Well, GKS did get rid of billing for cluster managers

Coté finished up this pile of crap (get a preview!) and right after emailing it in was reminded that Ben wrote this up already, plus an update based on re:Invent this week.

End-roll

Conferences

Coté’s junk:

NEXT WEEK, FOOLS! SpringOne Platform registration open, Dec 4th to 5th. Use the code S1P200_Cote for $200 off registration. Coté and many others speaking.

Coté will be doing a tiny talk at CloudAustin on December 19th.

Matt’s (not) on the Road! Taking it off for the Holidays.

Recommendations

Matt Ray: Art of War, backlaid by Wu Tang Clan

Brandon: Hindenburg audio editor.

Coté: Programmed Inequality; drink after the kids go to bed; Mindhunter; Jim and Andy.