Categories: architecture, cloud, devops, events, kubernetes

Google Next London 2018 Key Takeaways

Reading Time: 4 minutes

Google Next took place on the 10th – 11th October 2018 at the London Excel arena. The key takeaways of this article will be focused on application development, DevOps and AI as these are the areas of talks I attended.

The talks which I attended can be seen below:

Continuous Delivery Best Practices with Jenkins and GKE

Kubernetes Serverless and You

At Scale Financial Processing and Analysis with Thompson Reuters on Google Cloud

How to run your workloads on Google Kubernetes Engine (GKE) using Enterprise Ready Capabilities

Agile Application Development with Serverless Computing on Google Cloud

Multi-scale CI/CD from Serverless to Hybrid

This article doesn’t aim to be a thorough analysis of each point, just certain points that stood out to me during the talks.

Continuous Delivery Best Practices with Jenkins and GKE

The main goal of this talk was how Jenkins (one of the most popular CI systems in the world) can integrate with Google’s cloud platform within their CI/CD workflows. Key takeaways from this talk were:

  • Jenkins workflow can be integrated into Kubernetes using Helm Chart
  • Jenkins can intercept a pipeline process leading to deployment directly from Google Cloud Build (120 mins free per day) & Google Container Registry
  • High spec preemptible nodes can be used within a CI/CD process to provide high performance at low cost
  • A security mesh using Istio can be integrated during deployment providing a method to secure, control, and observe services
  • Knative a Kubernetes based platform to build, deploy, and manage modern serverless workloads is now supported and can be integrated into a CI/CD workflow
  • Default Ingress for Kubernetes is the LoadBalancer. NodePort service not recommended even for CI/CD workflows
  • The readiness probe value should always be analyzed when deploying a new version of a pod (Rolling Updates via deployments)
  • Cluster auto scaling and horizontal pod is built into Kubernetes however vertical scaling is also possible
  • Logging and application tracing is directly integrated into google cloud using StackDriver to analyze your deployments
  • Recommendation to listen to the weekly Kubernetes podcast!

Kubernetes Serverless and You

This talk focused on providing information on what Google and industry leaders think about the future of serverless

  • Companies such as the New York Times have completely gone serverless
  • Kubernetes integration with serverless is possible with Knative which provides a set of middleware components deploying to Google Cloud Functions
  • Knative provides serverless containers (have to enable it via the add-on)
  • Rely’s on an event driven architecture
  • Sample flow could be an IoT (Internet of Things) device triggering an event which is routed by Kubernetes to the cloud function

At Scale Financial Processing and Analysis with Thompson Reuters on Google Cloud

This talk focused on explaining how Thompson Reuters uses Google Cloud to conduct financial analysis with enormous data sets ingested into BigQuery.

  • Thompson Reuters combined and injected historical / real time financial data into BigQuery
  • Once the data was inside BigQuery the data could be run through machine learning algorithms (trained) and visualised by software such as Tableau
  • Tensorflow was used to train the models within Google Cloud
  • Blockchain data is also now being combined to improve buy in / buy out stock predictions
  • Jupiter workbooks used for the Python code 

How to run your workloads on Google Kubernetes Engine (GKE) using Enterprise Ready Capabilities

This talk focused on the growing interest and rapid adoption of containers amongst organizations. Many are hesitant as it can be demanding. The result of the talk involved how using the enterprise ready features of Google Kubernetes Engine (GKE) and cloud networking can resolve these problems.

  • Cloud armour delivers defense at scale against ingrastructure and application Distributed Denial of Service (DDos) attacks using Google’s global infrastructure and security systems. This includes HTTP flood attacks.
  • Istio can be an extra layer of protection for Pods within a Kubernetes cluster
  • Knative provides additional add-on support for Kubernetes
  • Google cloud marketplace allows open source and licensed software to deployed with no hassle onto Kubernetes ex. Elasticsearch and Kibana
  • Different teams can have their own Kubernetes cluster all on the same VPC
  • A new efficient LoadBalancer (Network Endpoint Group) was released today in beta. It is an improvement over the instance group LoadBalancer by improving latency providing less hops via iptables to reach pods within kubernetes nodes
  • Ingress (default LoadBalancer) on Google Cloud Platform provides support with Cloud Armor (Denial of Service defence) Identity-aware Proxy (identity based access control) and Cloud CDN (Global caching infrastructure)
  • Cloud NAT is now in Beta

Agile Application Development with Serverless Computing on Google Cloud

This talk explains how GCP provides a set of tools in order to create and deploy serverless workflows. The benefit? The avoidance of managing servers.

  • Cloud build and spinnaker can be used for a CI/CD workflow
  • Stackdriver will shortly be supporting grafana integration
  • Cloud scheduler can be used as a cron to trigger cloud functions
  • API management tool such as apigee can be integrated in the workflow
  • Cloud functions are private by default for security
  • Dialogflow (human–computer interaction technologies) integration with cloud functions was shown
  • A sample workflow is: Front (Firebase) -> API (ApiGee) -> Webhook to serverless function -> Response
  • Serverless containers are now in alpha

Conclusion

Pros

  • Interesting talks
  • Wide variety of subjects
  • Great demos and side activities such as talking to experts and companies who use Google Cloud products
  • Good food and coffee 😉

Cons

  • It would have been good to have a few talks on how different messaging brokers were easy to install on the Kubernetes engine eg. Kafka / Rabbit MQ instead of talking about PubSub for everything. Same for databases with MongoDB instead of Google’s Datastore. (understandable for marketing purposes but a little more OpenSource spirit would have been great). There are many companies out there who just want to leverage Google Cloud for their infrastructure and network, not their SAAS solutions.
  • Too many people (you had to arrive at a talk at least 15-20 minutes before a talk to guarantee a spot)

Until the next one!

BCG Platinion

Please follow and like us:

Leave a Reply

Your email address will not be published. Required fields are marked *

*