Cloud-Native Patterns and the Benefits of MySQL as a Platform Managed Service
0 likes1,103 views
The document outlines Pivotal's MySQL for PCF, a scalable, enterprise-ready solution designed to optimize database provisioning, management, and integration for cloud-native applications. It emphasizes improving operational efficiency, developer productivity, and reducing costs through automated processes and secure configurations. Additional features include high availability, simplified data management, and elimination of vendor lock-in, supporting both mission-critical and non-mission-critical workloads.
Cloud-Native Patterns and the Benefits of MySQL as a Platform Managed Service
- 1. © Copyright 2017 Pivotal Software, Inc. All rights Reserved. Version 1.0 Sept 2019 MySQL for PCF Enterprise-Ready, Platform-Managed MySQL & a Deep-Dive into HA Clusters
- 2. Pivotal’s Database Offerings Operational AnalyticalMixed Pivotal MySQL Pivotal Postgres Pivotal Greenplum ● High speed - tuned for OLTP ● Web scale ● Simple and easy to set up ● Thriving community of experts who share best practices ● Mixed OLTP-OLAP workloads ● Ideal for complex stored procedures and intricate designs ● Flexibility offered by extensions ● Common target for Oracle migrations ● Massively parallel for analytics at speed and scale ● Infrastructure-agnostic from bare metal to cloud ● Integrate structured and unstructured data in a single environment PCF PKS/on-premPKS/on-prem
- 3. The Emerging Need for Data Transformation Your data organization needs a path to be a part of the transformation to cloud native applications ● Certified DBAs spend wasted time running a ticket queue ● Growth in software practice and business bottlenecked by data/data team ● Developer productivity is hindered by data complexity and interdependence ● Held hostage by enterprise database vendors’ licensing tactics
- 4. Even if your Apps are cloud native...
- 5. ➔ Developers spend inordinate amounts of time for provisioning and setting up databases ➔ Cloud infrastructures are ephemeral requiring hardened availability at the software layer ➔ New distributed architectures result in an explosion of database instances ➔ Operators are overburdened with configuring, managing, and monitoring databases What We’re Hearing
- 6. Operational Efficiency __ Deliver highly available, resilient, up-to-date systems at a lower cost Developer Productivity __ Get apps to market faster by giving developers the tools and methods Secure By Default __ Apps that are easy to patch and incorporate best practices Multi-Cloud Success __ Make it possible to build and run effectively on any cloud Our customers achieve transformative results at scale MySQL for PCF enables these outcomes
- 7. ● Automate provisioning of databases on-demand using the platform, reducing DBA toil ● Simplify data monoliths into smaller & cheaper database backing microservices ● Integrate native platform tools & components for streamlined day 2 experience ● Enable secure & sane database defaults with per-instance configurations to match workload profiles ● Separate from vendor product lock-in, adopt OSS to reduce cost and leverage community expertise Value of MySQL for PCF
- 9. Automate Database Provisioning Actual customer app developer’s database request flow. ~8 weeks
- 10. Automate Database Provisioning Provisioning a database using MySQL for PCF. ~8 minutes
- 11. Break Down the Data Monolith
- 12. Break down the data monolith
- 13. Break down the data monolith
- 14. API access makes change management easier ● Define a API that projects a data model to match the needs of the consuming microservices ● API is the only permissible access point to a microservice’s data ● Data API provides a contract for accessing data ● Allows more flexibility for (and isolation from) changes to backing stores
- 16. Add on beyond the server process Opinionated Distro Integrate with Platform(s) Dev Enablement Day 2 Security Sane Defaults Logs Monitoring Backups Data import Service broker Dashboards Self-service APIs Fleet management Plan upgrades Upgrades Credential rotation Tuning Just the MySQL server Day 2 Enhancements
- 17. MySQL for PCF Upgrade Experience Source
- 18. Manage the Fleet
- 19. On-Demand Topologies Disk MySQL Disk MySQL Disk MySQL Disk MySQL Disk MySQL Disk MySQL App App App Single Node Leader-Follower Highly Available Clusters Jumpbox Proxy Proxy Proxy Backups
- 20. Tier 3 Tier 2 Tier 0 Tier 1 Non-mission critical Dev & Test Everything else Not used Enterprise DB solutions Mission critical workloads The new fleet of databases requires a new way of thinking
- 21. Tier 3 Tier 2 Tier 0 Tier 1 Mission critical workloads: - HA Clusters - Leader-Follower Mission critical and shared data sets: - Leader-Follower Enterprise DB solutions - Data Monoliths Manage a fleet of smaller databases that match your workload profile Non-mission critical, Dev & Test: - Single node - Leader- Follower
- 23. Recovery Time Objective In how much time you need to recover your IT infrastructure and services following a disaster in order to maintain business continuity? Availability topologies App App App Recovery Point Objective How much time worth of data loss you can tolerate in the event of a disaster in order to maintain business continuity? Good Better Best Good Better Best
- 24. Recovery Time Objective In how much time you need to recover your IT infrastructure and services following a disaster in order to maintain business continuity? Complexity & Cost How much time, effort, and technical expertise is required to troubleshoot in the event of a disaster in order to maintain business continuity? Availability Costs & Benefits App App App Recovery Point Objective How much time worth of data loss you can tolerate in the event of a disaster in order to maintain business continuity? Good Better Best Good Better Best Low Medium High
- 25. PCF Operator Workflow - L-F Failover Video: https://youtu.be/iOU41-kfjuQ
- 26. Video: https://youtu.be/PNHGz6YwNMw26 High Availability Cluster Failover Demo
- 27. HA Clusters: Deep Dive
- 28. Developer creates high availability instance and binds it to app VM dies, App fails over BOSH resurrector restores high availability, App fails back Disk MySQL Disk MySQL Disk MySQL App Jumpbox Proxy Proxy Proxy Backups Disk MySQL Disk MySQL Disk MySQL App Jumpbox Proxy Proxy Proxy Backups Disk MySQL Disk MySQL Disk MySQL App Jumpbox Proxy Proxy Proxy Backups HA Clusters - Failover
- 29. BOSH updates the 1st node, no impact to Applications BOSH continues to the 2nd, Apps fail over BOSH finishes the 3rd node, Apps fail back Disk MySQL Disk MySQL Disk MySQL App Jumpbox Proxy Proxy Proxy Backups Disk MySQL Disk MySQL Disk MySQL App Jumpbox Proxy Proxy Proxy Backups Disk MySQL Disk MySQL Disk MySQL App Jumpbox Proxy Proxy Proxy Backups HA Clusters - Rolling Upgrades
- 30. ● Multi-master: Any node is available for failover ● Out of sync nodes are automatically evicted ● In Leader-Follower, if one node crashes, the service is down till manual failover completes. BOSH DNSProxy MySQL server Proxy MySQL server Zone1Zone2Zone3 Proxy App MySQL server HA Deep Dive - The Cluster
- 31. MySQL server Zone1Zone2Zone3 ● Spreading three nodes across multiple AZs prevents against isolated IaaS failure. ● Worst case: Cluster is down, not split brainBOSH DNS Proxy Proxy Proxy App MySQL server MySQL server HA Deep Dive - Availability Zones
- 32. MySQL server MySQL server MySQL server Zone1Zone2Zone3 Proxy ● We control all the traffic to a single master with proxies providing super-fast failover times ● Failover is near-immediate when one node crashes. ● The proxies also help protect from split brain and from some Galera limitations Proxy Proxy BOSH DNS App HA Deep Dive - Proxies
- 33. MySQL server BOSH DNS Proxy MySQL server Proxy MySQL server Zone1Zone2Zone3 ● BOSH DNS (PCF 2.2+) directs apps to the active proxy ● “Zero downtime” rolling upgrades ● 99.999% uptime, planned and unplanned Proxy App HA Deep Dive
- 34. Up Next: Multi-DC Replication
- 35. Multi-DC Replication - Preview Active-Passive App Layer Active-Active *GSLB = Global Service Load Balancer Secondary DC Primary DC Secondary DCPrimary DC GSLB GSLB App A App A’ App A App A’ failover failover failover
- 36. Transforming How The World Builds Software © Copyright 2017 Pivotal Software, Inc. All rights Reserved.
- 38. Enterprise Ready Scale to Hundreds of DB Instances for Modern Microservices-based Applications ● Customize service plan definitions to support internal chargeback packages and enforce resource constraints. ● Highly available and resilient to node, availability zone, or site failure (beta). Zero downtime upgrade and patches ● Automate periodic backups and allow for restoring from a backup ● Data in motion is protected via transport layer security. ● Get visibility into details of service operation through standard monitoring and logging tools. Receive alerts for faster problem resolution