Workload Placement Analysis & Policies

Workload placement is a critical component of infrastructure planning and design, and is the key element to achieving higher efficiency within consolidated infrastructure. Without strategically placing workloads, organizations not only increase operational risk, but they lose significant opportunities to maximize efficiency by increasing densities and designing infrastructure according to required service levels.

Workload placement analysis focuses on where workloads should go, both at the environment level and individual server level, and is guided by operational, business, risk, and performance policies. CiRBA enables organizations to validate that workloads are currently in the appropriate and/or optimal location, direct new workloads into the most suitable running environment, and ensure that controls are in place to prevent workloads from “drifting” onto the wrong servers.

Validating the Current Placement of Workloads

CiRBA enables the assessment of current placements of physical and virtual workloads. Leveraging established placement policies, CiRBA verifies that:

• Aggregate resource utilization levels are not exceeding set limits
• Business policies, such as data segregation policies, regulatory compliance rules, and export control restrictions, are not being violated
• Application placement affinities are being considered so that workloads that should remain together, are kept intact, such as combining certain VMs on common hosts to optimize memory sharing
• Placement anti-affinities are being enforced so that workloads that should be kept apart are separated, such as separating app clusters across physical boxes, supporting failure mode, and data replication requirements

Placing New Workloads within Existing Infrastructure

CiRBA enables organizations to determine which IT environment or hosting model is best suited to a particular workload by evaluating all the alternative placements such as a physical environment, a virtual environment, an internal Cloud, an external Cloud or in an outsourced location. CiRBA analysis also evaluate whether sufficient capacity exists to host the new workload, and that the workload is placed in accordance with prevailing policies. Key aspects of this analysis include:

• Modeling of new workloads, using hypothetical models, standard models chosen from a predefined catalog, or empirical measurements of pre-production environments
• Finding the most suitable infrastructure type (servers, hypervisors, Clouds, other hosting models)
• “Gating” of new workloads into the selected environment to ensure technical, business, and performance policies will not be violated
• Establishing “Virtual First” policies to ensure that physical servers are procured only if all virtual, Cloud, and other shared options are deemed unsuitable

Controlling VM Motioning through Active Placement Governance

In spite of efforts to carefully design environments and to manage workload placements over time, there can still be a risk that workloads will be moved to the wrong place. This can arise from manual system administration activities or from automated “load balancers," such as VMware DRS. Organizations can prevent this by establishing run-time governance that is consistent with business policies and technology constraints. CiRBA assists with this governance by providing the following:

• Automated generation of affinity and anti-affinity pairings of workloads based on all constraints and policies
• Integration with VMware DRS to populate rules that control VM movements

Establishing and Managing Tiered Services to Increase Cost-Effectiveness

Strategic placement of workloads according to SLA requirements enables organizations to control costs by matching workloads with suitable infrastructure. Less critical workloads can be placed together and receive lower service and resource levels, while more critical production workloads are placed on more expensive infrastructure. Key to this is understanding how a given set of workloads will utilize the resources of a server when placed together, and whether or not they will contend for resources and impact application response times. CiRBA accommodates service level requirements, as well as advanced workload modeling and resource contention probability into its analysis to enable organizations to establish and manage Tiered Services. Through this analysis, different hosting environments can be established, each tuned to specific SLA criteria, and each potentially having different chargeback rates to reflect the operational risk being assumed.