Seven Reasons for Putting Business-Critical Workloads in the Cloud

Business leaders are starting to understand and realize the benefits of Cloud Computing for non-critical workloads. Those applications that do not have strict service level agreements (SLA), or do not process business-critical information are easier to justify for Cloud infrastructures. The reasons for putting a Tier-1, mission-critical application and data in the Cloud are different from their Tier-2/3 counterparts. This presentation will identify seven critical business drivers for putting Tier-1 workloads in the Cloud. Real-world case studies from Indiana University, American Tire Distributors, and Madison Community Hospital will be presented to demonstrate how these drivers are being applied in production Cloud environments. Best practices for realizing these benefits, and deployment options will be reviewed. The seven benefits of putting Tier-1 workloads in the Cloud include the following: 1. Cost Reduction — This comes in the form of both CapEx reductions from hardware and hardware-based software licensing, and OpEx reductions from improved manageability, shorter development cycles, improved QA, etc. 2. Near 100% Reliable Disaster Recoverability — Historical recovery from a DR image has been dicey at best. Virtualization and Cloud Computing provide near 100% recoverability. 3. Product Release Cycle Optimization — Fast and complete environment cloning; isolated test environments; quick server provisioning; improved development server management, etc. 4. High Availability (HA) — High availability for Tier-1 workloads is not a nicety, it is absolutely critical. VMware HA, Oracle RAC, and other HA solutions work better in a Cloud environment. 5. Application/DB License Optimization — Load up the hardware with more Oracle or other Tier-1 workloads without paying for more hardware-based licenses; comingle pre-production and production workloads on the same servers with performance isolation; use Cloud tooling to maintain recoverable DR copies in remote data centers without licensing the standby servers (until and unless they are needed). 6. Production Error Capture — When there is a production problem, create a snapshot (after a brief down) to capture the reproducible error state. That production capture can then be triaged and diagnosed in a safe environment. 7. Security Considerations — Complete isolation between systems on the same host; smaller attack target footprint; processor-based VM security assist; randomization of executable locations; Dynamic Binary Translation; zero-based memory; storage isolation between VM’s that prevents wide-scale data corruption; virtual switches less vulnerable to MAC spoofing, random frame, and other types of attacks; etc. This is not a theoretical presentation, but a discussion of real-world examples compiled over the last five years.

- by Nathan Biggs