Google Cloud Next ’17 and the Future of GCP

If you haven’t seen them yet, the sessions from Google Cloud Next ’17 are all on YouTube. They actually managed to post the sessions on the same days they were held, which is amazing turnaround time compared to most conferences.

As I posted on Twitter, the sessions and Google speakers were spectacular, and you should watch the videos and verify for yourself if you can. The talent that Google has is tremendous, and it should be a concern for AWS and Azure competitively.

That said, the Google Cloud Platform hasn’t had the fanfare that AWS or even Azure has had recently. Some of that is due to their perceived inability to understand the Enterprise market, which Google is working to solve with their hiring of ex-VMware CEO Diane Greene; some of that was due to their previous support models and processes, which they’ve turned around over recent years including announcing even better support model/plans at Next.

Nevertheless, the perception and corresponding reality of GCP’s place in the pecking order exists… but I think that’s going to change quickly. GCP is maturing, and it’s doing so fast. According to Eric Schmidt in the keynote, Google has spent $30 billion on GCP, so they’re committed.

I’ll start by thanking Google for their YouTube’s 2x speed playback functionality that made catching up on the sessions more enjoyable. After diving into the sessions and starting to tinker with GCP more, I wanted to share some of the big differentiators that GCP has.

The list below is going to include things that existed before Google Next ’17, as well as announcements at Google Next ’17. These are just a couple of things that I think are unique differentiation in the Cloud Wars.


Sustained Use Discounts

Google has had Sustained Use Discounts for awhile, which are the simplest approach to saving money via usage of any of the big providers, as it’s automatic and requires no upfront commitment. Sustained Use Discounts (SUD) basically work by providing a discount for instances that have been running for at least 25% of the time, with additional discounts the more the instance is used per billing cycle. For example, an instance running 100% of the month would automatically get 30% off. GCP also uses “Inferred Instances” to combine multiple, non-overlapping instances in the same zone into a single instance to maximum SUD benefits. The instance pricing for Google is also Per-Minute compared to Per-Hour.

Effective discount increases with use under Google’s Sustained Use Discount
Inferred Instances Example Calculation

Committed Use Discounts

Google announced the release of Committed Use Discounts at Next. Committed Use Discounts work by purchasing a committed amount of vCPUs and RAM. The vCPUs and RAM can then be used across instances at your discretion on all non-shared core machine types in a region. Discounts up to 57% can be had with Committed Use Discounts versus full price. This contrasts with AWS’ approach of Reserved Instances which have a bit less flexibility. Currently, none of the big competitors offer pricing discounts similar to this.

Always Free Tier

Google Cloud Platform also announced their expanded Free Tier that allows $300 free trial credit would extend to 12 months, but the best part is they have announced their “Always Free” tier. This “Always Free” collection includes things like 1 f1-micro instance per month, 5 GB-months of Regional Google Cloud Storage, 1GB of Google Cloud Datastore, and more. Full list available on the Always Free Usage Limits page.

User Interface & User Experience

Google Cloud Shell

Google Cloud Shell is not something new, but it’s something as an administrator I enjoy. The Cloud Shell provides CLI access directly inside the GCP UI to run “gcloud” commands and other utilities (e.g. mysql client, etc). It’s also a session you don’t have to pay for, and comes with 5GB persistent disk storage.

Google Cloud Shell listed all installed Cloud SDK Components

Equivalent REST or Command Line Display

Whenever you perform an action in the GCP UI, there is a link to see the equivalent REST and CLI commands that would be performed. These commands can be used instead of clicking “Apply” if desired, but it’s all formulated based on your settings defined in the GUI. Very neat compared to having to figure out the command line or API call on your own.

Generate Equivalent REST of Command Line Buttons on Wizards
Equivalent REST Request Example for Creating GCE Instance

Google Cloud Deployment Manager

GCP’s equivalent to AWS CloudFormation is Google Cloud Deployment Manager. It uses YAML versus JSON, so I am naturally a fan based on the readability of YAML comparatively.


Google groups resources into Projects instead of tying everything to Accounts like AWS. Depending on your perspective, this could be a benefit. Cross Project Networking (XPN) is still in beta which allows cross project networking as the name implies, and while the documentation breaks it down it comes off as a not as nice as VPC Peering among accounts with AWS.

Technical Goodies

Live Migration

Not a new feature, but if there is maintenance required on the host where your GCE instance runs, Google can Live Migrate it without disruption. If that happens in AWS, you need to restart your instance on a new host as there is no live migration.

Custom Instance Sizes

Google Compute Engine is not bound to pre-configured instance sizes, although they do have them as options (e.g. f1-micro). Instead, you get a slider to select CPU and RAM. This isn’t completely customizable, because you can only do 6.5GB of RAM per vCPU to avoid tilting too far in one direction, but the customization is nice.

GCE Instance Size Sliders

Open Standards

Google is big on Open Source and Open Standards. The release a lot of research as a company, and there are A LOT of popular open source projects released by Google.

Cloud Spanner

Google Cloud Spanner is a fully managed, mission critical, relational database service that offers transactional consistency at global scale, schemas, SQL, and automatic, synchronous replication for high availability.” That’s the formal definition of Cloud Spanner, but the informal definition is that it’s a way to do massive relational databases at global scale easily. It’s a unique solution.

The Entire Google Cloud Platform

I’m not going to go nearly as deep as one should on the technical goodies in GCP, so you’ll need to dig for yourself into awesome solutions like App Engine, Cloud Functions, BigQuery, Machine Learning, and more. Big Data and Machine Learning are areas where Google is naturally very strong, and will lead the way.

There is a lot to the platform: in the words of Google, “Cloud Platform is what brings Google’s innovation and infrastructure to you.