Architecting a Private Cloud - Cloud Expo
- 1. Architecting Your Cloud: Lessons Learned from 100 CloudStack Deployments Shannon Williams Vice President Market Development, Cloud Platforms CloudExpo New York June 13, 2012
- 2. Today’s goal: provide a basic understanding of different cloud architectures • Outline a process for defining a cloud • Describe the building blocks used to deploy a computing cloud • Look at traditional workloads and cloud workloads • Consider architectures that meet a broad set of requirements
- 3. Since 2008 CloudStack has powered hundreds of clouds • Secure, multi-tenant cloud orchestration platform – Turnkey platform for delivering IaaS clouds – Hypervisor agnostic – Highly scalable, secure and open – Complete Self-service portal – Open source, open standards – Deploys on premise or as a hosted solution
- 4. Each cloud drives unique requirements Service Providers Web 2.0 Enterprise 4
- 5. Architecture definition is a process Iaas Cloud Define target workloads Determine how that workload will be delivered reliably Determine the necessary functionality and performance Develop your technical architecture Implement your environment
- 6. Workload categories give us a starting point Traditional Enterprise Applications Software Disaster Development, Recovery Testing and Maintenance Social Media Managed IT Applications Services High Performance Batch processing Computing 6
- 7. Possible to categorize workloads into two sets Cloud Workloads Traditional Workload Cloud Workload Reliable hardware, backup entire Tell users to expect failure. cloud, and restore for users when Users to build apps that can failure happens withstand infrastructure failure Both types of workloads must run reliably in the cloud
- 8. Reliability & DR are Workload Specific RPO (Recovery Point Objective) • Recovery Point Objective (RPO) and $ 1 Recovery Time Objective (RTO) should be determined based on workloads $$ Regular 2 • Deployment and DR plan should be designed per RPO, RTO requirements $$ 3 Critical • Different types of workloads will achieve workload reliability in different ways Mission Critical RTO (Recover Time Objective)
- 9. Workload reliability drives unique requirements Traditional Workload Cloud Workload Link Aggregation VM Backup/Snapshots Storage Multi-pathing Ephemeral Resources VM HA, Fault Tolerance Chaos Monkey VM Live Migration Multi-site Redundancy Expect reliability. Back-up entire cloud. Expect failure. Design app for failure. Admin controlled failure handling Self-service failure handling Think Server Virtualization 1.0 Think Amazon Web Services
- 10. Other functionality will impact design as well Networking Storage Template Management VM Features Features Features Management Features • Resizing • Dedicated user • Persistent • Master • Delegated • High Availabity networks Storage Template Administration • Cloning • Integrated • Ephemeral Library • Live Migration • Monitoring Firewall Disk • User Template of VMs • Integrated • Automated upload • Live Migration • Windows Support Load Balancing Disk Snapshots • User ISO of Storage • IP Address • Cloud Storage upload • Usage • Linux Support Management access • Blank VM Metering • Naming • Multiple Guest • Disk creation • User Interface • Grouping Networks Monitoring • Private • Console Access • Security templates • VPN • Encryption • Multi- Termination • Template Hypervisor • Intrusion migration • Open-Source Prevention • Multi- Datacenter
- 11. Every cloud starts with basic building blocks Servers Networking Storage Server Server Server Clusters Clusters Clusters Networking Storage Hypervisor Resources Availability Zones Clouds
- 12. Two sample zone architectures - Traditional server virtualization zone - Amazon-Style availability zone
- 13. Designing a zone for a traditional workload Hypervisor Feature Rich– vSphere, vCenter vCenter Storage Enterprise Networking (e.g., VLAN) SAN ESXi ESXi ESXi Networking Cluster Cluster Cluster L2 VLANs Network Services Enterprise Storage (e.g., SAN) Load Balancing PV-LANs Multi-tier Apps Multi-tier VLANs OVF
- 14. Designing a zone for a traditional workload • Can achieve significant reliability for vCenter applications running in one zone. Enterprise Networking (e.g., VLAN) • Reliability of individual nodes is very high. • All zone storage is replicated to a second ESXi Cluster ESXi Cluster ESXi Cluster storage platform (synchronous or asynchronous) Enterprise Storage (e.g., SAN) • In event of failure, images are recovered from second storage array. • Existing workloads will run reliably. • Little cost benefit over existing approaches
- 15. Designing a zone for an Amazon-style workload Amazon-Style Availability Zone Software Defined Networks Hypervisor (e.g., Security Groups, EIP, ELB,...) Simple - XenServer Server Server Server Server Storage Racks Racks Racks Racks Local EBS Object store Server Server Server Server Networking Racks Racks Racks Racks L3 SDN based L2 Elastic IP Server Server Server Server Network Services Racks Racks Racks Racks Security Groups ELB GSLB Elastic Block Storage Multi-tier Apps L3 SDN based VPC CloudFormation
- 16. Object store is critical for Amazon-style cloud Amazon-Style Cloud Amazon-Style Availability Zone CloudStack Mgmt. Server Software Defined Networks (e.g., Security Groups, EIP, ELB,...) Server Server Server Server Racks Racks Racks Racks Server Server Server Server Availability Availability Availability Racks Racks Racks Racks Zone Zone Zone Server Server Server Server Racks Racks Racks Racks Object Storage Elastic Block Storage
- 17. Object store is critical for Amazon-style cloud Amazon-Style Cloud CloudStack Mgmt. Server • Workloads are distributed across availability zones • No guarantee on zone reliability • Applications designed to handle node Availability Availability Availability level failue Zone Zone Zone • DBs and Templates snapped to object store. • In event of failure, images are Object Storage recreated on new availability zone. • Dramatically less expensive
- 18. Cloud Transition – General to Workload specific Past Today General Traditional-Style Amazon-Style Architecture • General architecture for any • Workload-centric architecture workload • Workload-specific failure/disaster • Limited definitive failure/disaster recovery recovery strategy • Separate legacy and cloud app • Focused on legacy or cloud app architectures with interoperability architectures
- 19. Support for different styles will be required CloudStack Mgmt. Server Server Virtualization Availability Zone vCenter Enterprise Networking (e.g., VLAN) Availability Availability Availability Zone Zone Zone ESXi ESXi ESXi Cluster Cluster Cluster Object Storage Enterprise Storage (e.g., SAN)
- 20. Availability zones will be distributed globally CloudStack Management Cluster San Jose London Miami Hosted Dehli Tokyo Hosted Rio
- 21. Availability zones are becoming on-demand On Premise Hosted Private Cloud Managed Hosted Private Cloud Federated/Hybrid Public Private Cloud Cloud Services Cloud Services Enterprise Enterprise Enterprise Multi-tenant Users Multi-tenant Users Data Center Data Center 3rd party 3rd party hosted & operated operated • Dedicated resource • 3rd party owned and • Mix of shared and • Shared resources • Total control/security operated dedicated resources • Elastic scaling • Internal network • SLA bound • Shared facility and • Pay as you go • Security staff • Public internet • Dedicated resource • VPN access
- 22. Key takeaways • Begin with the definition of the workloads your users will run on the cloud. • Define a clear strategy for what level of reliability those workloads will require, and how you will deliver it. • Understand failure scenarios, DR scenarios, monitoring and backup deeply. • Don’t be afraid to support more than one workload category, it will be key to driving consumption. • Open platforms will provide the most flexibility to support a wide variety of workloads. • Hypervisor, Networking and Storage models are more important than specific vendor choices
- 23. Work better. Live better.