Guest Column | January 14, 2014

What You Need To Know About Converged Infrastructure Platforms: Part 3

By Jeff Kennedy, Business Value Marketing Manager, Eaton

How advanced power protection technologies provide businesses greater flexibility in ensuring converged infrastructure resiliency

In addition to the advantages described in Part 1 and Part 2 of this series, converged infrastructures also give companies an effective new way to increase the resilience of their information technology (IT) environment.

Traditionally, strategies for maintaining high availability have focused on an IT infrastructure’s hardware layer, emphasizing the use of durable, redundant servers, networking, storage, and power protection systems. Such approaches have a long record of success and remain strong options for organizations that need uninterrupted availability 365 days a year.

However, hardware-centric resilience schemes have several drawbacks. For one, they’re often difficult to modify, making them a poor fit for today’s dynamic and highly virtualized infrastructures. For another, they can be expensive, as they require companies to purchase and maintain more IT equipment than they need purely for purposes of redundancy. As a result, many businesses now utilize resilience strategies emphasizing other layers of the technology stack, such as:

  • The user layer: User-centric approaches to resilience identify user populations that can tolerate small amounts of data loss or brief service disruptions, and then deliver the redundancy that populations require.
  • The application layer: Large organizations, including major cloud service providers such as Google, often increase resilience by building maximum fault tolerance into their application code. However, such schemes rely on massively distributed — and therefore costly — architectures, as well as significant investments in programming time.
  • The cloud/virtualization layer: Rather than prevent equipment failures and power loss, resilience strategies focused on the cloud and virtualization layer mitigate the effects of such problems by either automatically restarting virtual machines or swiftly migrating them to unaffected hosts elsewhere on the network or in the cloud. As a result, they enable companies to maintain continuous or near-continuous uptime without enduring the expensive burden of deploying redundant hardware or writing fault-tolerant code.

Not surprisingly, building resilience into the cloud/virtualization layer is quickly becoming the most popular approach to preserving availability. And since converged infrastructures are heavily virtualized and automated, solutions ideally suited to cloud/virtualization resilience strategies, they too are rapidly gaining favor among enterprise IT managers.

Using advanced power protection technologies give businesses greater flexibility in how they ensure the resilience of converged infrastructures. Specifically, companies that deploy state-of-the-art uninterruptible power systems (UPSs), power distribution units (PDUs) and power management software in conjunction with converged infrastructures have these three options for protecting data and preserving uptime automatically during utility failures:

  1. Transfer all virtual machines to an infrastructure or site where power remains available.

Once that process is complete, the converged infrastructure’s power management system can perform a graceful shutdown of the hypervisor and a controlled power down of the physical servers and storage systems. With this option, full IT service remains available throughout the utility failure, though users are likely to experience somewhat slower performance than usual during the virtual machine migration process.

  1. Transfer mission-critical virtual machines to an infrastructure or site with functioning power, gracefully shut down the remaining virtual machines and then power down the physical host servers.

In this scenario, services provided by non-critical virtual machines will be unavailable until utility power is restored. For some companies, however, that’s an acceptable price to pay in exchange for a shorter and simpler migration process that also costs less because it doesn’t require as much disaster recovery capacity. For example, manufacturers must typically halt production lines during power outages anyway, so shedding production-related workloads poses no additional threat to profitability.

  1. Maximize standby power by gradually shutting down virtual machines rather than migrating them offsite.

This option utilizes the power management system’s load shedding functionality to shut down virtual machines sequentially, starting with the least mission-critical ones and ending with the most. Throughout that process, the virtualization management system automatically consolidates remaining workloads onto fewer and fewer physical devices, allowing companies to extend their backup power by progressively shutting down unused host machines. In addition to shedding the lower priority loads, the IT manager may also invoke power capping to regulate power being consumed by the servers and further extend battery run time. If utility power is still unavailable when standby power reaches critically low levels, the power management and virtualization management systems can execute a controlled shut down of the few remaining virtual and physical servers. This option leaves organizations vulnerable to service disruptions, but also spares them the expense of maintaining offsite disaster recovery resources.

Note that all three of these options hinge on tight integration between the converged infrastructure and a comprehensive power protection solution that includes networked UPSs and intelligent PDUs, as well as sophisticated power management software.

While converged infrastructures are inherently resilient, robust power protection is an essential element of successful converged infrastructure deployments. When developing packaged converged infrastructure solutions, therefore, resellers should treat power protection as the all-important “fifth element” of a complete solution, alongside servers, storage, networking, and virtualization software.

Jeff Kennedy is the Business Value Marketing Manager for Eaton’s Distributed Power Quality, Transactional Power Products division and is responsible for driving consistent strategy and messaging for the Eaton Transactional Power Products (<11kVA)  portfolio.  Kennedy attended North Carolina State University’s Poole College of Management where he earned his bachelor’s degree in business management with a concentration in marketing. Kennedy also attended North Carolina State University’s Jenkins Graduate School where he earned his master’s in business administration. For information on Eaton’s power management software, click here.