Containerization Strategy
0 likes168 views
The document discusses the challenges and inefficiencies of manual application deployment processes, highlighting the benefits of containerization as a solution to improve deployment speed, scalability, and cost-effectiveness. Containerization is presented as a new deployment model that automates and simplifies the process, reducing reliance on human intervention and accelerating the time-to-market for applications. Additionally, it addresses common misconceptions about containers, outlines best practices for implementation, and emphasizes the importance of orchestration and scheduling in deploying applications effectively.
Containerization Strategy
- 1. • Value proposition • Requirements • Application mapping • Approach • Comparison Containerization Strategy Balaji Mariyappan
- 2. Problems focus & Empathy Present quality process requires more environments to deliver to prod Hard to document each new step in the implementation in the manual process Application deployment steps are 50-70% manual Patching & fixes are manually applied and lot of human resources involved until it is done Differences in the environments, increases the QA process time in each environment Hardware augmentation can take up to 10-15 days with the mix teams & approvals Application infrastructure deployment is slowing down the process in many cases Not many repeatable process slowing down the implementation Mutability in Infrastructure & deployments requires more ITIL processes to be approved Multiple gate-controls before code deployed in all-levels of environment Multiple ITIL steps during the process deployments is slowing down Human intervention in every step and requires multiple stages to deploy Application scalability is managed manually – in cases it takes 2-3 days in lower environments Impacting Velocity Impacting Agility Impacting Governance Impacting Quality
- 3. What is Containerization? • The process enhances the deployment model, which reduce the cost and time to go to market. A new way of application deployment which includes new way of packaging the OS, tools, and applications and rolls out to and new class of environment • It leverages the concept of fully contained deployments so that underlying environment variables will be nullified at each implementation which will reduce the time to test & debug the application changes every time • It effectively leverages automation, so that building the new environment will be quicker at a low cost
- 4. Containerization - Unique Value Proposition … is to provide an application deployment platform which can reduce the operating cost to the organization and help business & IT effectively increase the frequency of adding new features in applications without negatively impact the Quality, Revenue and Business requirements
- 5. Why Containerization matters? Less Capex needs Reduction in Opex by time Faster to market by Agility in deployments Smaller the deployments will increase the deployment density per hardware or cloud instances needed Reduction in deployment will increase the time to support so more deployments can happen using less human involvements Faster the deployment more flexible for business plan their feature rollouts
- 6. Containerization efficiency ContainerVM Infrastructure OS Bins & Libs App/s Infrastructure Hypervisor Bins & Libs App/s Guest OS Infrastructure OS / Guest OS Container Engine Bins & Libs App/s Containermanager VM Bins & Libs App/s Guest OS Container Bins & Libs App/s Container Bins & Libs App/s Container Bins & Libs App/s + Isolated apps + CPU perpetual isolation - Anti-cloud / Pattern - High admin overhead - High Capex / Opex - Less Resource efficient + Isolated apps + Partially CPU perpetual isolation - Less cloud-friendly - Moderate Capex / Opex - Less Resource efficient - Less Automation + Highly automated + Resource effective + Small & portal + Faster deployment + Faster booting + High CICD friendly + Reduces QA time - Plausibly Less Secure & additional tools needed for Security - Less isolation - Networking & learning curve - Licensing impacts Standalone HW Virtualized Infra Containerized Infra
- 7. Crawl, Swim & Run …. Innovate Check and Market Production to scale of standard applications Check and Market Test one deployment in Prod POC / Discovery Discover an app based on complexity and MVP Not wait long in the Pilot, go to prod for proof of deployment Show and market to other teams for feedback Start deploying other applications Show and market to other teams for feedback Start modernizing the apps Execution Planning
- 8. Application Mapping Medium (Migrate as-is) Easy (Migrate as-is) Hard (Refactoring needed) High Efforts (Refactoring needed) Microservices PHP Apps Simple apps Custom Java Web Apps Stateful services Stateless services ClusteredStandalone Web Cache ESB Services Portal Web Portal services ODS MS SQL DBs No SQL Mainframe SOLR ETL Mobile services Node Apps Service Cache Message queues First in Production Reverse proxy Web-proxy Monitoring agents Map your application by • Effort • Easy • Medium • High Effort • Hard • State of the application • Stateful services • Stateless services
- 9. Myths about Container adoption Myths Gartner research recommendation Containers are just for Microservices Containers are great distributing and scaling Microservices Container can be used for module and Microservices Docker engine is enough to build a complete container platform Role of Orchestration and scheduling is mandatory Scalability requires platform Simplify the deployment by using platforms Container will replace VMs Comparison is Apples to Refrigerators Container is better run on VM than bare metal VMs help using existing operational processes Container are just for Linux Docker on Linux and windows has same outcome, different implementations There are plenty of windows application images available in the public hubs Help future-proofing the legacy applications Container are faster than VMs Life cycle operations are faster in container vs VMs Guest apps performance is identical when compared in VMs Use Containers where faster startup & spin-down time matters Containers aren’t secure Threat vector is high due to multiple repetition / automation process Gartner paper: Technical Insights: Top Container Myths – Simon Richard Busted Busted Busted Busted Confirmed Plausible
- 10. Pillars Governance Platform & Environment Operations & CI/CD Application Technology Non- Technology Application deployment, container manifest, Container platform manifest, application config Code versioning, Image registry, CI, CD & tools & Release maintenance Hardware, Platform, Capacity, HA, Monitoring tools, Security scanning tool Team org, Service delivery, Reporting, training, marketing
- 11. Operating Model When /Operation Code Build /Containerize Code Deploy / Container deploy Test Propagate Now Containerization for new Apps / Microservices Containerization for Existing apps Future model For app deployment QADeveloper Web AdminDevOps Web Admin Developer DevOps Web Admin QA DevOps DevOps Web Admin QA DevOps DevOpsDevOpsDevOpsDevOpsDeveloper
- 12. Personas • Responsible for OS Images & updates • Custodian of OS & Security images • Responsible for cloud infra management Sys Admin (DevOps) Sec Admin (DevOps) Web Admin (DevOps) Developers (DevOps) • Responsible for scanning and certifying images • Run scans for cloud deployment vulnerability during runtimes • Validates the code security • Deploy images in container platform • Maintains code & image repositories • Responsible to maintain the deployment pipelines • Coding & maintaining the versions • Container Image creators • Addresses the code & build compliance QA • Application testing • Manages Automated testing pipelines • Manages the automated performance testing
- 13. Sample Container deployment flow Stories Coding Build Validate Image Deploy/Run Task tracker SCM Artifactory Kubernetes Clusters Orchestration & pipeline tools - Spinnaker / Harness.io /XL release/Jenkins-Ansible-Udeploy/Rancher Monitoring (NR) & Security Scanners (Aqua/Twistlock/Snyk/Jxray) Build tool Security Admin DeveloperBusiness Maven Gradle Nexus Bitbucket Kubernetes Docker Bitbucket SVN / TFS Jira Presentation – Akamai – Apigee – ELB - Haproxy DevOps DevOpsDevOps
- 14. Deployment flow now & then Developer Desktop Integration Env QA Env Production Env Artifacts Repo JAR Code Repo Developer Desktop Integration Env QA Env Production Env Container Repo Image Code Repo Currently, the developer check-in the code and code will be built for different environment and deployed on an existing platform where OS, patching & tools are expected to be similar in counts and versions In the containerization, artifacts will be built as an image which contains OS, Patching, tools other configuration and run as containers with application code. This process will make sure underlying components will remain the same in all deployment platform including developer desktop
- 15. App Deployment Options Code Build Push Pull Deploy Test Code Build Test Build Deploy Test Developer DevOps DevOps QA & Sec Build once Deploy many Build many Deploy many DevOps QA & Sec DevOps DevOps QA & Sec Developer DevOps DevOps
- 16. Infrastructure Guest OS/ OS Container Engine Container Control plane Deployment pipeline SCM Build repo / Container Image registry Configuration manager / Infra automation API manager Load balancer / Reverse Proxy CDN / WAF / API Security Cloud 1 LINUX Docker Kubernetes Jenkins Bitbucket Nexus / Artifactory Ansible APIgee Haproxy Akamai Cloud 2 LINUX Docker Kubernetes AWS Code pipeline Bitbucket ECR Cloudformation AWS API GW ALB/NLB Cloudfront On-prem Hardware LINUX Docker Kubernetes Jenkins Bitbucket Nexus / Artifactory Ansible APIgee F5 Akamai OpExWork(20%) CapWork(80%)OpExWork(80%) CapWork(20%) High vendor locking possibility Medium vendor locking possibility Low vendor locking possibility Legend Sample Containerization Vendor locking matrix Version 9/24/2019 - Balaji Mariyappan Stack option 1 Stack option 2 Stack option 3Stack / Layers
- 17. Container platform comparison (EKS Vs. Fargate) 1 2 3 4 5Cost attractiveness Ease of use 1 2 3 4 5Kubernetes compatible 1 2 3 4 5Customization possibility 1 2 3 4 5Deployment friendly 1 2 3 4 5Compatibility w/AWS svc Factors Importance 1 2 3 4 5Market share AWS FargateAWS EKS 1 2 3 4 5 *EKS /GKE/AKS are partially PaaS solution for Kubernetes *Fargate / CloudRun / OpenShift are fully PaaS solution for K8S