OHECC 2014 KSU Cloud Integrated Datacenter
3 likes490 views
The document outlines a strategy for building a cloud-integrated datacenter at Kent State University, focusing on integrating compute, storage, and virtualization both on-premises and in the cloud. It emphasizes key steps such as virtualizing non-cloud-ready applications, adopting a cloud-first approach, and ensuring vendor diversity while also discussing the related economics and costs of cloud versus traditional infrastructure. Recommendations are provided for defining objectives, determining pain points, and effectively managing cloud services deployment.
OHECC 2014 KSU Cloud Integrated Datacenter
- 1. Next Generation Datacenter Compute, Storage, Virtualization on Prem and in the Cloud Walt Bainey, Systems Design Architect, Server Support & Engineering Joe Friscone, Sr. Systems Administrator, Server Support & Engineering Todd Ryan, Lead Systems Administrator, IT Infrastructure Services & Support
- 2. The Cloud-Integrated Datacenter Todd Ryan Lead Systems Administrator, IT Infrastructure Services & Support Email: tmryan@kent.edu Phone: 330-672-1652
- 3. Building the Cloud-integrated Datacenter Next Steps for Kent Recommendations Agenda The Cloud-Integrated Datacenter Integrating Compute Integrating Storage The Case For the Cloud
- 4. 1. Virtualize non-cloud-ready apps 2. Identify, prioritize, and migrate cloud-ready apps 3. Cloud-first approach 4. Hybrid services where appropriate 5. Ensure control over strategic applications 6. Maintain cloud vendor diversity and flexibility 7. Vendor selection is critical TheCaseFortheCloud KSU Cloud Strategy
- 5. They ask for… TheCaseFortheCloud Agility We give them…
- 6. • More Disaster Recovery options • Simplified availability across datacenters/regions • Snapshot backup without additional software • Opportunities to leverage more “aaS” for databases, development, etc • Software-defined everything TheCaseFortheCloud New Capabilities
- 7. TheCaseFortheCloud Economics When will we need to buy more compute or storage capacity? 0 100 200 300 400 500 600 700 800 1 5 9 13 17 21 25 29 33 37 41 45 49 Months
- 8. TheCaseFortheCloud o Application • Software licensing • Supporting software (DB, platform, etc) o Compute • Servers • Chassis • Virtualization • Operating systems o Storage • SAN controllers • Storage network • SAN disk o Backup/DR • Backup servers • Backup software • Local backup storage • Remote backup storage • Internet bandwidth o Datacenter Network • Core switches • Rack switches • Firewall • Load balancers • Cabling • Management SW o Network • Core switches • Firewall • Cabling • Management SW • Internet bandwidth o Support Services • Enterprise Monitoring • IDS/IPS • Antivirus • Application Firewall o Support • Implementation Services • Infrastructure Support • Application Support • Training Economics What does our infrastructure really cost?
- 9. Price per instance per year: $464 Price per VM per year: $430 TheCaseFortheCloud • 16 blades/chassis/SAN fabric/Ethernet/etc • SAN with 2-3 shelves per blade chassis • VM Licensing for all blades • Hardware and license maintenance • Hardware support Economics Typical Premise VM Cost • 2:1 ratio of m3.large and m3.medium • M3.large: 2 vcpu, 8gb, 100gb disk • M3.medium: 2vcpu, 4gb, 50gb disk • 3 year heavy reserved capacity Typical AWS Instance Cost How do I compare my costs to the cloud?
- 10. TheCaseFortheCloud Economics Consider that unused capacity in your cost analysis 100 247 503 240 480 720 0 100 200 300 400 500 600 700 800 1 5 9 13 17 21 25 29 33 37 41 45 49 VMCapacityandServerGrowthIncrements Months Total VMs 16 Server Increments
- 11. TheCaseFortheCloud Economics How SHOULD typical organizations buy capacity? 124 233 386 488 120 630 780 0 100 200 300 400 500 600 700 800 900 1000 1 5 9 13 17 21 25 29 33 37 41 45 49 VMCapacityandServerGrowthIncrements Months Total VMs 2 Server Increments
- 12. TheCaseFortheCloud Economics What does that excess capacity cost? $430 $842 $430 $638 $430 $569 $430 $535 $464 $300 $400 $500 $600 $700 $800 $900 $1,000 110 160 210 260 310 360 410 460 510 560 610 660 710 760 810 860 910 960 PriceperVM Total VMs Annualized Cost/VM Annualized Cost/Cloud VM
- 13. TheCaseFortheCloud Economics Cloud Pricing Trends – S3 Storage 15 14 12.5 9.5 8.5 15 12.5 11 8 7.5 13 11 9.5 7 6 10.5 9.5 9 6.5 5.5 8 8 8 6 5.1 5.5 5.5 5.5 5.5 4.3 2.75 < 11/2010 11/2010 < 12/2012 12/2012 2/2014 4/2014 First 1TB Next 49TB Next 450TB Next 500TB Next 4000TB Over 5000TB
- 14. BuildingtheCloud-integratedDatacenter Integrating Compute Comparing Cloud Providers Compute Services AWS Azure Google Oracle VMware Compute - Linux EC2 VMs Compute Engine Yes (Preview) Yes Compute - Windows EC2 VMs Compute Engine No Yes On-demand scalaing AutoScaling Auto scale No No Only with vCloud Director Licensing Security boundaries Multiple: ELB, VPC, Subnets, Sec Grps Server-level, Private Network No Per Instance Private Network Load Balancers Elastic Load Balancing Traffic Manager Yes No Yes VPN Tunnel VPC Gateway Azure Virtual Network No Yes (Preview) Yes Direct Leased Connection DirectConnect ExpressRoute No No Yes VM Import/Export VMware or Hyper-V, internet, ship disks For Hyper-V easily, Vmware import only No No For VMWare images only - via Internet Private Networks Yes-VPC Yes No Yes Yes Storage AWS Azure Google Oracle VMware Object Storage S3 Yes Cloud Storage Yes Yes Archive Storage Glacier No No No No Block Storage EBS Yes Yes Yes (Preview) Yes High performance Block Storage Provisioned IOPs EBS No No No No Storage Gateway AWS Gateway or 3rd party appliances StorSimple physical appliance 3rd party physical appliances No No Data import/export Via internet or ship disks Via internet or ship disks No No No Database as a service AWS Azure Google Oracle VMware NoSQL DBaaS DynamoDB Tables and Queues Cloud Datastore No No Oracle (compute-based) RDS No No Yes No MSSQL (compute-based) RDS SQL Database No No No - but on roadmap MySQL (compute-based) RDS No CloudSQL No No Data Warehousing Redshift No No No No In-memory cache ElastiCache Azure Cache Dedicated MemCache No No Big Data/Analytics MapReduce HDInsight BigQuery Possibly with Bus. Intelligence (Preview) No Application Services AWS Azure Google Oracle VMware Web sites No Web Sites No No No Mobile Services Only mobile push Mobile Services Cloud Endpoints Yes (Preview) No Code/Release Management No Visual Studio Online No Yes (Preview) No Application Streaming AppStream No No No No Search CloudSearch No No No No Java App Deployment Elastic Beanstalk No App Engine Yes No WebLogic Deployment No No No Yes No .NET App Deployment Elastic Beanstalk Cloud Services No No No Node.js App Deployment Elastic Beanstalk No No No No Python App Deployment Elastic Beanstalk No App Engine No No PHP App Deployment No No App Engine No No Go App Deployment No No App Engine No No Ruby App Deployment Elastic Beanstalk No No No No Data Processing Data Pipeline Scheduler/ BizTalk No No No Workflow Simple Workflow Service Automation (Preview) No No No Messaging Services Simple Email Service Azure SMTP Service No No No Mobile Messaging Simple Notification Service Notification Hubs No No No Message Queueing Simple Queue Service Messaging No No No Video transcoding Elastic Transcoder Media Services No No No Key Management CloudHSM No No No No Web traffic analysis Alexa No No No No Manual human processing Mechanical Turk No No No No Directory Services No Azure Active Directory No No No Template-based application and infrastructure deployments CloudFormation No No No Possibly with vCloud Connector or vCAC Operational templates for scaling and managing applications OpsWorks No No No Possibly with vCloud Connector or vCAC Application Marketplace Yes Yes No Yes Yes Other AWS Azure Google Oracle VMware Content Delivery Network CloudFront Azure CDN No No No DNS Services Route 53 No No No Service on the vCloud Edge Gateway Virtual Desktop WorkSpaces No No No No Data Streaming Kinesis No No No No Data Backup to cloud Not native. Requires 3rd party tools Azure Backup No Yes - For Oracle database backups only No VM DR to cloud Not native. Requires 3rd party tools Recovery Manager No No vCHS-DR Management/Monitoring AWS Azure Google Oracle VMware Identity Management IAM No. Minimal roles but insufficient No No No Monitoring CloudWatch In preview No Yes (Varies by product) Available from vCOPs, addt'l Licensing Activity monitoring/Logging CloudTrail No No No No Best practice analysis and cost optimization Trusted Advisor No No No No Procurement model Pay-as-you go or Reserved instances Only Pay-as-you go No No Completely up front, on-demand promised SaaS Offerings AWS Azure Google Oracle VMware Marketing No No No Yes No Sales No No No Yes No Service No No No Yes No Human Resources No No No Yes No Talent Management No No No Yes No Financials No No No Yes No Procurement No No No Yes No Project Portfolio Management No No No Yes No Supply Chain No No No Yes No Enterprise Planning/Reporting No No No Yes No Social Network/Marketing/Monitoring No Yes No Yes No Business Intelligence No No No Yes (Preview) No Messaging/Email No Yes Yes No No Instant Messaging No Yes Yes No No Collaboration No Yes Yes No No File Sharing/Synch No OneDrive Drive File Sync and Share (Preview) No
- 17. • Seamless integration of premise and cloud compute • User won’t know where their infrastructure is deployed • Brokerage for cloud infrastructure services • Address some of our pain points • Time to delivery • Unpredictable Cap-ex • Unused capacity • True cost visibility • Take advantage of cloud benefits BuildingtheCloud-integratedDatacenter Integrating Compute Our Objectives
- 18. BuildingtheCloud-integratedDatacenter Integrating Compute How we did it (Summer 2012-Fall 2013) Provider Selection Evaluate Cloud providers Select AWS Perform Security review Negotiate Enterprise Contract Learn POC for one major app In-depth portfolio review Deployed sandboxes Deep dive and education Design Initial Production Design Sandbox Testing Refine design Deployment Deploy Test Build test apps Refine design Deploy Production
- 19. What it looks like today BuildingtheCloud-integratedDatacenter Integrating Compute
- 20. What it looks like today BuildingtheCloud-integratedDatacenter Integrating Compute AZ2 x.x.0.c/16 AZ1 x.x.0.a/16 ProductionUSEASTVPC KSU Datacenter AD LDAP File Share Storage Gateway DNS ISSVR Region: US-EAST-1 S3 Region: US-WEST-2 S3 EBS Snapshot Internet Cisco ASA #1 Cisco ASA #2 Security Group Availability Zone Region AD DC1 AD DC2 Monitor1 Mgmt Sec Group JUMP-AWS x.x.0.0/16 EBS Snapshot AZ1 x.x.0.b/16 Web Svr Sec Group WAF-1-ZONE1A S3 WAF Security Group Web Svr 1 Web Svr 2 AZ1 x.x.0.d/16 VPC GW Directory Services Sec. Group EE-group y.y.y.a/32 y.y.y.c/24 y.y.y.f/23
- 22. • Removal of pain points • Time to delivery • Unpredictable Cap-ex • Unused capacity • True cost visibility • Backup performance • Provide lower cost storage options • Brokerage for cloud storage services • Seamless integration of premise and cloud file storage • Provide any quantity of storage without buying hardware BuildingtheCloud-integratedDatacenter Integrating Storage Our Objectives
- 23. BuildingtheCloud-integratedDatacenter Integrating Storage How we did it (Summer 2012-Fall 2013) Learn POC with AWS Gateway Study use cases Research product capabilities Gateway Selection Identify vendors in space Detailed feature comparison TCO analysis of top products Product Selection Deployment Deploy device to test Perf and usability testing Refine config Deploy Production Migration Revisit use cases Identify high-value candidates Migrate
- 24. BuildingtheCloud-integratedDatacenter Integrating Storage What it looks like today Tier 2 (Cloud Storage) Tier 1 (SSD, SATA) CIFS/NFS File shares KSU Datacenter S3 Scripted Snapshots S3 Encrypted Compressed Storage Gateway NAS ISCSI Archives Backups AWS Storage Gateway Enterprise Storage Other App Servers Managed Snapshots/ Retention
- 25. • Develop reference architectures and templates • Better “cloudify” apps • More DBaaS • More PaaS • Auto-scaling • Look at integrated management/reporting tools • Continuously review other cloud providers • Deeper look at how we use storage • Where do we use SAN that NAS would suffice? NextStepsforKent Next Steps for Kent
- 26. 1. Define your objectives and pain points 2. Determine if Cloud Computing is a solution 3. Get executive sponsorship 4. Build a cross-functional team (compute, network, security, appdev) 5. Develop accurate cost model for premise services 6. Perform an in-depth review of applications 7. Pick your cloud wisely and develop a relationship 8. Perform a POC 9. Design and test your architecture. Does it meet your objectives? 10. Deploy production apps 11. Monitor spend closely 12. Repeat Recommendations Recommendations
- 27. Wrap-Up