Bringing together the best of PaaS and IaaS

MAR 27, 2014
By Navneet Joneja, Cloud Platform Team

Cross-posted from the Google Cloud Platform blog

Editor’s note: This post is a follow-up to the announcements we made on March 25th at Google Cloud Platform Live.

For many developers, building a cloud-native application begins with a fundamental decision. Are you going to build it on Infrastructure as a Service (IaaS), or Platform as a Service (PaaS)? Will you build large pieces of plumbing yourself so that you have complete flexibility and control, or will you cede control over the environment to get high productivity?

You shouldn’t have to choose between the openness, flexibility and control of IaaS, and the productivity and auto-management of PaaS. Describing solutions declaratively and taking advantage of intelligent management systems that understand and manage deployments leads to higher availability and quality of service. This frees engineers up to focus on writing code and significantly reduces the need to carry a pager.

Today, we’re introducing Managed Virtual Machines and Deployment Manager. These are our first steps towards enabling developers to have the best of both worlds.
Managed Virtual Machines

At Google Cloud Platform Live we took the first step towards ending the PaaS/IaaS dichotomy by introducing Managed Virtual Machines. With Managed VMs, you can build your application (or components of it) using virtual machines running in Google Compute Engine while benefiting from the auto-management and services that Google App Engine provides. This allows you to easily use technology that isn’t built into one of our managed runtimes, whether that is a different programming language, native code, or direct access to the file system or network stack. Further, if you find you need to ssh into a VM in order to debug a particularly thorny issue, it’s easy to “break glass” and do just that.

Moving from an App Engine runtime to a managed VM can be as easy as adding one line to your app.yaml file:

vm:true

At Cloud Platform Live, we also demonstrated how the next stage in the evolution of Managed VMs will allow you to bring your own runtime to App Engine, so you won’t be limited to the runtimes we support out of the box.

Managed Virtual machines will soon launch in Limited Preview, and you can request access here starting today.

Introducing Google Cloud Deployment Manager

A key part of deploying software at scale is ensuring that configuration happens automatically from a single source of truth. This is because accumulated manual configuration often results in “snowflakes” - components that are unique and almost impossible to replicate - which in turn makes services harder to maintain, scale and troubleshoot.

These best practices are baked into the App Engine and Managed VM toolchains. Now, we’d like to make it easy for developers who are using unmanaged VMs to also take advantage of declarative configuration and foundational management capabilities like health-checking and auto-scaling. So, we’re launching Google Cloud Deployment Manager - a new service that allows you to create declarative deployments of Cloud Platform resources that can then be created, actively health monitored, and auto-scaled as needed.

Deployment Manager gives you a simple YAML syntax to create parameterizable templates that describe your Cloud Platform projects, including: Templates can be re-used across multiple deployments. Over time, we expect to extend Deployment Manager to cover additional Cloud Platform resources.

Deployment manager enables you to think in terms of logical infrastructure, where you describe your service declaratively and let Google’s management systems deploy and manage their health on your behalf. Please see the Deployment Manager documentation to learn more and to sign up for the Limited Preview.

We believe that combining flexibility and openness with the ease and productivity of auto-management and a simple tool-chain is the foundation of the next-generation cloud. Managed VMs and Deployment Manager are the first steps we’re taking towards delivering that vision.

Update on operating systems

We introduced support for SuSE and Red Hat Enterprise Linux on Compute Engine in December. Today, SuSE is Generally Available, and we announced Open Preview for Red Hat Enterprise Linux last week. We’re also announcing the limited preview for Windows Server 2008 R2, and you can sign up for access now. Windows Server will be offered at $0.02 per hour for the f1-micro and g1-small instances and $0.04 per core per hour for all other instances (Windows prices are in addition to normal VM charges).

Simple, lower prices

As we mentioned on Tuesday, we think pricing should be simpler and more closely track cost reductions as a result of Moore’s law. So we’re making several changes to our pricing effective April 1, 2014.

First, we’ve cut virtual machine prices by up to 53%: What’s more, you don’t need to sign up for anything, make any financial commitments or pay any upfront fees. We automatically give you the best price for every VM you run, and you still only pay for the minutes that you use.

Here’s what that looks like for our standard 1-core (n1-standard-1) instance: Finally, we’ve drastically simplified pricing for App Engine, and lowered pricing for instance-hours by 37.5%, dedicated memcache by 50% and Datastore writes by 30%. In addition, many services, including SNI SSL and PageSpeed are now offered to all applications at no extra cost.

We hope you find these new capabilities useful, and look forward to hearing from you! If you haven’t yet done so, you can sign up for Google Cloud Platform here.

Navneet Joneja, Senior Product Manager

Posted by Louis Gray, Googler