I recently attended the Google Cloud Next ‘19 conference in San Francisco, thanks to my generous Leading EDJE training budget. I went because I wanted to learn more about Google Cloud Platform as most of my cloud experience has been with AWS. I’ve been to Amazon’s re:Invent conference a couple of times and thought it would be interesting to compare the conferences and the platforms.
There’s no doubt that part of the conference experience is the city where the conference is located, and walking through the Las Vegas casinos (many times) is an integral part of re:Invent. As re:Invent expands to more locations getting around has become a real problem. Google Cloud Next took place at the Moscone center in downtown San Francisco with a few other venues close by, making the conference seem smaller and more compact, even with nearly 30,000 attendees.
While Google Cloud Next is a shorter conference (3 days instead of 5), the format of both conferences is similar with a combination of keynotes, sessions and workshops, with a giant expo from vendors.
My biggest disappointment with Google Cloud Next was the lackluster keynotes. The main reason (for me at least) to attend keynotes is the expectation that there will be big announcements. Two of the three keynotes did contain announcements (the developer keynote did not), but they were lacking in build up and technical details, and I would have been better informed by reading the Google Cloud blog from my hotel room. There was none of the excitement and technical insight that the re:Invent keynotes contain.
I did find the sessions I attended very informative, especially with my limited GCP experience. I selected a few introductory sessions to give a grounding in the services available, but largely stuck with higher level technical sessions.
The expo area was open all three days most of the day, unlike re:Invent where it seemed to be closed more often than not. As well as vendors, there were displays from Google, and a large labs area where I did some late night Qwiklab labs and took part in the “Cloud Hero” competition.
From Google’s perspective I think that Anthos was the biggest announcement, but with pricing starting at $10,000/month it’s a little hard for me to get excited about. The idea of a consistent control plane and APIs for Kubernetes regardless of where it’s hosted is interesting and may eventually be transformative, but the announcement and documentation are currently light on technical details, especially around how they plan to support other cloud platforms. I also listened to a podcast about the Anthos Migrate service that will transform VMs into containers and all I got out of it is that it uses “streaming” to migrate.
I was much more excited by Cloud Run that provides serverless deployment of containers based upon the open source Knative runtime. I’m pretty sure this didn’t get announced at the keynote, but the sessions afterwards expected us to be aware of it. The big differences between Cloud Run and other function based serverless cloud functions (e.g. Google Cloud Functions and AWS Lambda) are the ability to provide a full container, and the concurrency model that allows for multiple requests to the same instance. The demonstrations showed fast startups (which I’ve confirmed myself) and rapid scaling under load. Scaling is based upon Knative’s throughput model rather than the less useful CPU based scaling typically used in Kubernetes.
Google also announced an interesting partnership model with a number of open source providers including Elastic, MongoDB and Redis Labs. You’ll be able to provision services from these partners from within GCP with a consistent interface, unified billing and initial support from Google. This compares to the AWS model of taking and hosting open source projects and sometimes forking them (e.g. ElasticSearch).
There were a lot of other announcements, and I encourage you to look through the full list.
So based upon all this new information I’ve absorbed, what are my impressions of the Google Cloud Platform?
- More of a focus on global and multi-region services than AWS - for example global load balancers and VPCs, but this seems to address the compute side of things more than data.
- The strongest Kubernetes implementation (unsurprisingly) and a range of services that build upon this ecosystem.
- The Cloud Shell (which Azure also has, but AWS lacks) is very useful.
- Managed SQL Server and Active Directory shows a new focus on enterprise migrations that was perhaps lacking before.
- Interesting machine learning integrations into other services for data analysis.
- A much smaller selection of services than AWS, but perhaps enough for most projects.
Without actually using GCP for a real project I reserve the right to change my mind, but I’d be interested to work on a GCP project.