Machine Learning stole the show across three days of service rollouts at Google Cloud Next 2018 in San Francisco. Despite already boasting the most powerful ML platform in the Cloud, Google’s next charge was to make its ML tools more robust and accessible: “What we found, working with hundreds of enterprises, is that using AI comes down to simplicity and usefulness,” says Rajen Sheth, a Director of Product Management at Google.
While much of the discussion centered around AI and ML for the enterprise, Google also announced significant updates to microservice management through Istio, hybrid cloud deployments in GKE and increased opportunities for cost savings at the resource level. Let’s break down the main Google Cloud Platform takeaways from Next.
Easier Customization of ML Models and More Out-of-the-Box Services Powered by ML
By releasing Cloud AutoML Vision, AutoML Natural Language, and AutoML Translation to beta, Google provides more stepping stones to useful integration of AI in the enterprise. These services differ from Google’s pre-trained ML APIs such that models created with AutoML take advantage of user-defined labels that are geared towards unique business needs. Each service presents an easy-to-use GUI with drag-and-drop functionality that places machine learning customization in the hands of those that don’t even know how to code. Solutions can then be deployed directly to Google Cloud Platform.
Along the same vein of democratization of AI, Google also announced the beta release BigQuery ML, a tool which allows SQL users of BigQuery to create custom regression models using only their working knowledge of SQL.
In regards to out-of-the-box Google Cloud Platform services that are powered by machine learning, we learned of a plethora of updates to Dialogflow Enterprise Edition and Text-to-Speech tools.
We’ve known for some time about Google Dialogflow’s aims to enhance the contact center experience by eliminating wait times, connecting virtual agents to knowledge, and making AI-assisted conversation more natural and conversation. At Next 2018, Google announced updates to Dialogflow that will make virtual agents easier to create, more accurate and more knowledgeable:
- Phone Gateway now allows you to configure an agent to begin taking calls (without worrying about underlying telephony infrastructure) in under a minute.
- Knowledge Connectors, to say the least, is really really cool: This allows your virtual agent to understand unstructured internal documents (you know, all of those documents already written for live agents) for use in conversation coupled with your model’s pre-built intent.
- Sentiment Analysis incorporation with Dialogflow scores user sentiment for one-off responses to analyze instances (perhaps, if there is an angry customer on the other end) when handoff to a live agent may be necessary.
- Agent Assist provides a live agent with support content (such as articles and resource documents) in real time during customer contact.
The end goal is to provide better, more timely customer service using AI-augmented contact center models. The latest updates to Dialogflow (a full rundown can be found here) get us one step closer to the goal.
To round out updates to ML for GCP, Google also announced increased availability of TPUs for model training in the Cloud, as well as Kubeflow v0.2, a more streamlined interface for supporting machine learning stacks with Kubernetes. This keeps in line with the theme of enterprise accessibility by making ML models faster to train and easier to deploy.
Google Cloud Services Add New Advantages to Cloud Computing with Managed Istio and GKE On-Prem
On the first day of Next, Google announced a new suite of offerings under the moniker Google Cloud Services, a managed container service that integrates with Istio service mesh to facilitate hybrid cloud environments. GKE On-Prem gives users the experience of running GKE in the Cloud, except within their own data centers. This allows customers to containerize their environment, abstract away cluster management and prepare themselves for full migration to the cloud (or to maintain a hybrid environment if they very well please).
Google anticipates widespread adoption of the managed version of Istio just as they have seen adoption of managed Kubernetes over the past few years. And similar to how Kubernetes simplifies the process of deployment, Istio allows for the management of microservices as they run, no matter which environment they are hosted on. SADA’s own Director of Cloud, Simon Margolis, covered rollouts to Microservices and Hybrid Cloud on GCP in his panel “Google Cloud Platform – The What, The How, and The Why.” View the video below to explore the following topics:
- Scale fast while keeping costs down
- Analyze massive datasets on a small team (as little as one person!)
- Decide on the right layer of abstraction on GCP (GAE v GKE v GCE)
- Leverage microservices oriented architecture on GCP as opposed to traditional monoliths in the cloud or elsewhere
On a final note to Cloud Services, Google also introduced a new managed CI/CD solution, Cloud Build. Though we’ve recently discussed CI/CD solutions in Google Cloud, Next ‘18 positioned Cloud Build as a particularly compelling contrast to third party CI/CD tools compatible with GCP, especially in light of its integration with Github (among other version control platforms).
The tools available in Google Cloud Services allow enterprises to modernize and standardize their workloads even before they are ready to move them into the Cloud. Of course, when these companies are ready, GCP will be expecting a knock on the door.
On a Final Note: It’s Always Nice to Save More Money
While there were over 100 significant updates to Google Cloud announced at Next 2018, our attention was particularly drawn to the introduction of resource-based pricing to sustained use discounts. Overall, it’s a pretty simple and intuitive change to the pricing model – customers should be charged at the core level regardless of what type of VMs they scale up and down to during the month – but it’s a welcome change for Cloud adopters that need more from their sustained use discounts. The reality is that cloud costs can account for a significant expense in the bottom line (especially at scale), and effort-free savings demonstrate good will between GCP and existing cloud customers.
We at SADA Systems have a lot to hit the ground running with for our enterprise clients following last week’s announcements. Curious about how your enterprise can adopt new features announced at Next ‘18? Contact SADA System’s team of Google Cloud experts to discuss the logistics of bringing cutting-edge cloud capabilities to your organization. Contact SADA System’s team of Google Cloud experts to discuss the logistics of bringing cutting-edge cloud capabilities to your organization.