Workload placement - for optimizing capacity

Capacity approach for cloud environment must ensure maximum cloud infrastructure utilization. In virtual environments where provisioning of infrastructure is easy there can be issues of under provisioning and over provisioning. So there is a must requirement for an efficient capacity approach which will seek a balance between these two.

A capacity approach must be primarily focused on placement of workloads and resource allocation. Fragmented capacity may lead to inefficiencies and may even double the infrastructure required to host the workloads. Capacity procedures must be defined which ensure intelligent allocation of virtual machines as required by the applications.

Here, capacity management toolsets can be considered, that enable the service provider to define technical, business, and compliance rules for workload placement. These workload placement rules are configured in cloud lifecycle management toolsets in the management layer. Rule engines must be evaluated to guarantee the health and accuracy of the capacity management solution.

As shown in Figure below, the best approach for placing workloads or virtual machines on a virtual infrastructure must be used to ensure existing resources are consumed efficiently. Allocation with the type A workload placement would lead to inefficiencies and unused capacity issues. As shown in the allocation scheme A in the figure unused patches potentially signify unused capacity. On the other hand, the workload allocation in type B ensures the best possible usage of resources.

Allocation "A" Allocation "B"

A workload is a logical classification of work performed in a virtual infrastructure. Workloads may be classified by who is doing work, what work is being done and how the work is being done. For example, a service provider must be able to classify the workloads according to business functions like sales, marketing, and finance, etc. Business-relevant workloads are also useful when it comes time to plan for the future.

Business applications must be analyzed for infrastructure usage requirements. There can be two business applications which are equally critical to business and consume resources variably. Capacity management must be able to establish such needs and accordingly classify the service levels associated with applications so that the infrastructure is provided accordingly which can be disk space, compute capacity, memory and network bandwidth requirements. These measures will help in planning of future system requirements.

Figure depicts how planning for workloads is critical for efficient resource utilization. Techniques like estimation, modeling and load testing can be used for efficient workload planning. Workload plan A, when placed in the infrastructure definitely is not the best workload pattern for available infrastructure whereas workload plan B is fit for the available infrastructure leading to efficiencies and apt resource utilization schemas. There are several toolsets available which specialize in workload planning that help capacity planners in establishing the best approach for capacity.

Workload Plan “A” Workload Plan “B”

Based on the performance, service level targets and business vital functions, a high-level approach toward meeting the targets is chosen. For example: A stiff cost focus in performance targets would call for a just-in-time solution, whereas, service continuity at stiff performance targets might require solutions to utilize the availability of margin capacity. An approach must be decided upon to provide real-time and historical workload reports that can be used for on-the-fly resource optimization and problem diagnosis. Real-time workload reports must identify under- or over-utilized server capacity, which can be used to optimize the distribution of the workload across all the available hardware, as well as help prevent unnecessary hardware purchases.

While setting up the approach for capacity, service providers also must assess the workloads through discovery, and inventory of IT assets, because these are prime causes behind capacity bottle necks. This must be done function, location and environment wise. Statistical algorithms must be used to calculate workload growth and other related metrics. These reports can also help to diagnose problems, especially in situations where capacity limitations go unnoticed or where load balancing within server farms or clusters is not working properly. They identify bottlenecks and required additional capacity to support expected or desired workload growth while respecting thresholds on resource utilization and response times. The capacity approach also must address the spare capacity management and ensure defragmentation and other techniques are in place to ensure proper positioning of workloads in the infrastructure. Proactive capacity allocation methods must be considered based upon infrastructure event monitoring and analytics.

Whitespace management is the management of the spare capacity in an environment. Key to managing whitespace is analytics that determine optimum spare capacity to take care of demand spikes. Toolsets must be brought into place to calculate the trade-off between cloud capacity and the associated costs.

Besides finalizing these issues, the capacity approach is established in consensus with the set financial plans and budgets. Financial management will provide information on the current and forecasted costs of providing capacity. In turn, the capacity management process also will provide information on charging capacity related needs and other data for calculating budgeting and charging.