By Gerald Hinton, Solution Architect, ClearPoint

As dedicated technologists at ClearPoint, we ensure we design and build what’s right for our customers. To do this, we stay ahead of the game by working closely with key partners such as Google Cloud. In September, the ClearPoint team attended the Google Cloud Summit in Sydney to hear what’s new from Google Cloud, meet with our customers and catch up with the wider Google community. 

Here are some key outtakes from the summit:

  1. A Multi-Cloud & Open Source world
  2. Market positioning
  3. Anthos
  4. Data
      1. Big Query GIS
      2. Machine learning within BigQuery
      3. BigQuery BI engine
      4. Cloud Data Fusion
  5. Machine Learning & AI
      1. AutoML tables
      2. AI Hub
      3. AutoML Video
  6. DevOps
      1. CloudBuild
      2. CloudRun

1. A Multi-Cloud & Open Source world

Google Cloud was very pragmatic that one cloud provider cannot meet the needs of all companies and this really resonates with me. Customers are choosing to go multi-cloud in a couple of different scenarios:

  1. When they want to get the best-of-breed tools to solve specific problems, they may choose tools from different cloud vendors.
  2. If they are an enterprise that wants to withstand an outage at any one cloud provider, having more than one powering your solution gives them that resiliency.

Google is very much positioning itself as playing nicely in a multi-cloud world and the launch of Anthos encapsulates this. We’ll talk more about Anthos shortly.

Another theme sitting alongside multi-cloud is open source. Since Google began 21 years ago, they have taken their own internal tools and offered them to the open-source community – with Borg becoming Kubernetes and Tensorflow, Android and AngularJS to name a few.

 

 

 

 

 

 

 

2. Market Positioning

During the keynotes and sessions, Google Cloud talked about their continued expansion through their offering to market, run rate and recognition in the industry. Google Cloud is expanding with ongoing investment. They have invested US$47bn over the last 3 years and US$13bn this year. There are now 61 zones in 20 regions globally for GCP, so any regional limitation that may have existed in the past is no longer a concern.

 

 

 

 

 

 

 

 

Google is now also receiving recognition from the likes of Forrester and Gartner. The recent market analyst’s surveys are recognising Google Cloud Platform as a cloud leader in a number of areas.

A few examples are below:

The Forrester Wave™: Cloud Data Warehouse Q4 2018 report names Google Cloud a Leader.

The report evaluates the data warehousing solution offerings of public cloud providers.
Google Cloud receives the highest score possible in 11 categories such as performance, data protection, and elastic scale.” 

Forrester Wave: Continuous Integration Tools
Google Cloud Build has strongest leadership compared to the other cloud vendors. Google Cloud Build is relatively new when compared to the other public cloud CI offerings, they had a lot to prove, and they did so.

3. Anthos

Since April 2019, Google Cloud has been shouting very loudly around Anthos and it was no exception during the Summit. 

Anthos is an offering that allows you to manage and govern multiple Kubernetes clusters across Google, on-prem and other cloud providers from a single pane of glass. Anthos brings together managed Kubernetes with a service mesh based on Istio and a custom policy management layer. As the details of Anthos continue to be fleshed out, I am impressed with the offering and think it has a place to reduce the overall complexity in large environments and improved governance in the enterprise space.

Whether Anthos will appeal to the wider market is dependent on what the commercial model will be. Only provisional pricing has been announced to date and that pricing is targeted at the enterprise space.

 

4. Data

Powering business intelligence and data-driven decision making are key strengths of the Google Cloud offering.  Google has key strengths in the data space, particularly around their data service offering with Google Big Query and the complementary services around it. Google have made announcements recently in a number of areas:

4.1 Big Query GIS

Big Query now supports storing GeoSpacial data types (GIS) which allow for queries and analysis to be done within Big Query. A visualisation layer is also available in BigQuery for use during development.

geoviz

4.2 Machine Learning within Big Query

Embedding Machine learning capability within BigQuery is now available. You can build and train machine learning models and do the prediction with a pre-built model all within the Big Query SQL like language. Today, you can use linear regression, binary logistic regression, multiclass logistic regression, K-Means clustering and Tensorflow model types – more are being added frequently. See an introduction here

 

4.3 Big Query BI Engine

Big Query is a fantastic data warehouse product where you can query petabytes of data within a few seconds. For business intelligence users who want a GUI that can query the data and get near-instantaneous response, Google has announced BI Engine. This allows data to be cached in memory with underlying storage in BigQuery for much faster querying on the fly. 

You can then use the data to power rich front ends from the data you already have loaded into Big Query. That data can be surface in Google Data Studio or other visualisation tools such as Tableau, PowerBI or Looker.

 

4.4 Cloud Data Fusion

Google has improved its capability recently in Extract/Transform and Load or ETL which is the method of getting data onto the GCP platform. There have been options from Google and third parties up to now but it was not a strength. 

Cloud Data Fusion is a new product that greatly improves the capability within the GCP platform where you have the ability to use a visual data wrangler to view & setup data manipulation, and then build data pipelines which Google executes for you on on the managed hadoop service dataProc.

Data sources can be file or stream-based with support for ever-increasing integrations that today include Google Cloud storage, Pub-sub, Amazon S3, and Amazon Kinesis.

 

5. Machine Learning & AI

Google continues to leverage its deep machine learning capability across GCP. The TensorFlow processing Unit (TPU) which is a custom chip designed to speed up TensorFlow based machine learning is available in compute instances which is unique across its competitors. This is accelerating many of its offerings under the hood.

 

5.1 AutoML tables

AutoML Tables is a way to perform machine learning in an automated way on tabular data. The data could be in BigQuery or a CSV file. You point AutoML tables at the source data which includes the fields on which to base the machine learning and the correct result for that row of the test data. AutoML then goes away and comes up with a model that it thinks can correctly predict the correct outcome for that sample data by trying a bunch of iterations of models and selecting the best one. Once the model is built, you can deploy it and make predictions over API or through the GUI.

 

5.2 AI Hub

AI Hub is a repository of machine learning models and usage. You can either make use of the publicly shared resources or have resources that are only available within your organisation. AI Hub has the ability to do one click deployments of the defined workload into your GCP account.

 

5.3 AutoML video

AutoML videoThe ability to do object recognition and classification through a managed service. AutoML is Google’s offering for building machine learning models and running predictions. You need to supply sufficient data to train the model, in this case, video, and then can use the model you build to make predictions.

Video is a recent addition to the AutoML offering and allows you to build custom models of what you want to observe, be that Kiwi fruit ripeness, cars or goals being scored without needing to worry about infrastructure or tools.

 

6. DevOps

DevOps capability is at the core of Google Cloud and aligns well with ClearPoint’s view on how to build and deploy software with reduced risk and increased velocity.

 

6.1 CloudBuild

CloudBuild is an application build service and CI/CD management layer from Google Cloud. You can build applications in any language and it supports Docker natively. It can initiate many types of steps including downloading dependencies, running the build, executing automated tests and then deploy it to a registry or directly to staging or production environments. We are already using CloudBuild on a customer project and think it’s a good fit for many applications.

 

6.2 CloudRun

CloudRunCloudRun is the Google implementation of the open-source knative initiative. Knative is designed as a way to run serverless code as containers on top of a Kubernetes cluster that mimics AWS Lambda or Azure functions.

CloudRun from Google Cloud can either be run as a fully managed service where the packages get deployed to Google-managed Kubernetes clusters, or it can be deployed to a Kubernetes cluster you are managing on Google Kubernetes Engine (GKE). Which method you would choose is based around your needs and often the cost model. The fully managed service allows you to scale to zero. So when there is no traffic coming in, it’s free as only as you pay per invocation which can add up when a lot of requests are coming in. When running on top of GKE, you pay for the cluster which means you don’t ever pay anything, but this is likely more cost-effective at a larger scale.

For any questions around Google Cloud and how to make the best use of it, please contact Gerald.Hinton@ClearPoint.co.nz or contact any of our team here