ITaaS is the new Holy Grail with 75 percent of IT managers saying ITaaS aligns with their organization’s philosophy and needs. Accustomed to living in a world where each application had dedicated servers, networking and storage, ITaaS eliminates this issue. It aggregates these resources into a common pool that is accessible by all virtual machines (VMs) and their hosted applications that may be owned by multiple different departments or even different organizations. These resources may then be allocated to them at any time.
Virtualization has provided an important first step toward delivering on the promise of ITaaS. By virtualizing server, network or storage resources, organizations get the flexibility to only assign the resources to the application on a VM that it needs while holding the rest of the resources in reserve to allocate to applications on other VMs as they need them. Virtualization also creates new opportunities for data mobility as the VMs on which these applications reside may be more easily moved within the infrastructure since they are no longer tied to a specific physical resource.
Yet virtualization alone fails to account for the specific challenges created by running multiple VMs in a shared infrastructure. For example, virtualization cannot natively distinguish between applications considered business- or mission-critical and those deemed “test” or “development.” This situation becomes problematic when applications compete for the same resources and the needs of less critical applications impact and/or take precedence over business- and mission-critical applications.
Further complicating the situation, multiple departments or organizations with their own VMs and applications may need to share this virtualized pool of resources. In this circumstance, a specific department or organization (known as a “tenant”) may need its applications to be given a higher priority to the underlying virtualized resources than its virtual neighbors.
Closing the Loop
To confidently implement ITaaS organizations must have a mechanism in place that effectively controls and manages both VMs and tenants in a shared infrastructure environment. Closing this loop requires that an ITaaS solution:
- Uniquely identify each application and tenant
- Give them priority classifications in the virtualized stack
- Gather performance data and associate it with the application and the tenant
- Be empowered to dynamically take the appropriate actions based on the gathered data
This last and most critical component is absent in most converged infrastructure solutions. While a number can uniquely identify an application, assign it a Gold, Silver or Bronze priority level and then even potentially gather the needed data to take action, most lack the capabilities to dynamically take the appropriate action(s) (i.e. – eliminate manual intervention) that is essential to making ITaaS a reality.
Putting a Priority on Performance
Priority Optimization software has been available on HP 3PAR StoreServ since June 2013. In that release, Priority Optimization capped the maximum amount of available resources that a particular LUN could access. In this way all applications residing on the same HP 3PAR StoreServ Storage system gets the resources they need while still leaving sufficient resources available for other applications such as Exhibit 1 illustrates.
Exhibit 1; Source: HP
However another issue that can arise is multiple applications, potentially from multiple tenants, competing for the same resources on the HP 3PAR StoreServ system. In this circumstance, organizations need assurances that their business- and mission-critical applications, as well as higher priority tenants, always have access to the resources that they need to meet the stipulations of their SLAs. Delivering on these SLAs may mean reducing the performance of other applications owned by that tenant or, alternatively, throttling back on the performance of applications of other tenants that have lower priority levels than they do.
The latest release of HP 3PAR Priority Optimization software (3.1.3) addresses both of these concerns. It provides Quality of Service (QoS) functionality at both the application and tenant levels by placing all of a tenant’s applications within an HP 3PAR StoreServ Virtual Domain.
Each application as well as each Virtual Domain is assigned a priority level. These respective priority levels include policies that define a maximum threshold as well as a minimum target level of performance. In this way, HP 3PAR StoreServ will only deliver a minimum target level of performance to lower priority applications and/or tenants when higher priority applications and/or tenants need more resources.
Priority Optimization satisfies the requirements of the higher priority application by first reducing the performance of other lower priority applications within a tenant’s Virtual Domain. If that re-allocation of resources is insufficient to meet the needs of that tenant’s higher priority application(s), Priority Optimization then reduces the performance of the Virtual Domains of other tenants to their minimum guaranteed SLAs. Using these steps, HP 3PAR StoreServ is able to re-allocate resources to meet the performance requirements of higher priority applications and/or tenants.