Success with virtualization is often dependent on successful management of the underlying hardware infrastructure. This may sound like a simple premise, but the goal can be increasingly difficult as data centers mix device vendors and models, and then compromise on the visibility provided by – often several – systems management tools. Converged infrastructure (CI) promises to overcome some of the issues by providing a bundle of servers, storage and network devices that have already been integrated, optimized and then tied together with unified management tools. But CI platforms are not necessarily right for every situation. IT professionals must understand the elements of CI, how CI interacts with legacy systems, and the best deployment scenarios for the technology.
What are the elements in a converged infrastructure, and what is actually virtualized?
The goal of converged infrastructure is to replace disparate, heterogeneous data center equipment with a complete suite of highly integrated and optimized hardware including servers, storage and networking devices. Virtualization is present in all three areas, allowing IT professionals to quickly pool and provision the storage and networking resources among a legion of virtual machines – each of which can easily be migrated between local and remote hardware. The high level of virtualization and flexibility found in CI deployments is often a precursor to private cloud deployments or even some type of software-defined data center.
CI platforms can be built from the ground up, such as IBM’s Flex System which includes an IBM Flex System Enterprise chassis, a range of x86 or POWER servers , PCIe expansion devices, IBM Flex System V7000 or Storwise V7000 storage units and a choice of networking including Ethernet and IBM Flex System fabric. The hardware choices are all managed by a single management framework, such as IBM Flex System Manager.
Cisco UCS is another example of the ground up approach. For example, Cisco UCS C-Series Rack Servers and Cisco Nexus Fabric Extenders for UCS Rack Servers can work together for a dedicated CI deployment while providing access to an existing SAN. Cisco UCS is supported through UCS Manager or UCS Director software.
CI platforms can also be bundled using gear from several different vendors. For example, NetApp’s FlexPod combines NetApp storage systems, Cisco UCS servers, and Cisco Nexus fabric into a validated CI product. Cisco’s UCS Director can provide centralized automation and management of the FlexPod’s components through a single, unified management tool.
Virtualization can be provided through almost any hypervisor including Hyper-V or vSphere.
How well does converged infrastructure interact with existing (non CI) systems?
The question of heterogeneity can be a challenging one for converged infrastructure environments. Generally speaking, CI can interoperate with non-CI devices whether virtualized or not. For example, there is no technical reason why a server of a different make or model can’t be added to a switch port on a CI platform. However, this can often defeat the purpose of CI.
Remember that the entire point of converged infrastructure is to provide a single ubiquitous hardware and software framework with comprehensive, granular management tools. Everything is tuned and tested and known to work together. When a foreign device is connected to CI, a new heterogeneous situation occurs which can result in reduced performance, poor management insight (limited management capabilities) or other interoperability problems. These are well-known issues in traditional heterogeneous enterprise environments, but an undesirable (and unnecessary) scenario once CI is introduced.
When CI is added to an existing data center environment -- perhaps as a platform for a private cloud -- the CI usually exists as an island of infrastructure with its own unique management tools and standards of performance. Any existing infrastructure is typically left to continue operating as it had previously, along with any other management tools. Supporting two “different” environments in this manner can pose a great deal of additional work for IT staff.
When considering a converged infrastructure, it is critically important to discuss support for heterogeneous or legacy environments with the CI vendor and perform extensive proof-of-principle testing in advance of any purchase commitment. This will tell you exactly how CI will interact with heterogeneous devices in your own data center.
We already have a virtualized data center, is it really worth moving to converged infrastructure?
Converged infrastructure has proven itself to be a powerful option for the modern data center. CI can potentially provide a more robust, flexible and better-managed high-performance virtualized infrastructure than typical heterogeneous data centers. But CI is still just a tool designed to tackle enterprise computing tasks. As with any tool, it’s important to understand when and how the tool is used most effectively.
CI is highly disruptive -- with CI bundles (servers, storage and networking) it essentially replaces the existing data center infrastructure as well as the investment made in that existing infrastructure, software licenses and service agreements. Few organizations are willing to throw away the major capital invested in a current data center in favor of an entirely new platform, and these financial implications have been a principal reason that CI has been slow to gain broad acceptance. In many cases, CI adoption is deferred until a major technology refresh is needed, or until a new "greenfield" facility becomes available (such as a new remote data center). Admins should look for ways to repurpose any gear displaced by CI platform deployments.
CI can pose growth and vendor challenges. Even though CI platforms are designed for scalability, it’s important to realize that there are only a limited range of products that are compatible with the overall CI bundle. Limited product choices (and dependence on CI vendor product lifecycles and roadmaps) often translate into higher product costs and potential availability challenges later in the product lifecycle. Organizations with unpredictable or erratic spikes in computing growth may have trouble obtaining adequate gear upgrades in a cost-effective manner. This is very different than traditional heterogeneous data centers, where any x86 white-box server or Ethernet switch might work in the environment.
CI deployments are also being slowed by alternative infrastructure paradigms like colocation and public cloud providers that can offer effective computing resources (and all necessary support) as a monthly operating expense rather than a major capital expense. This conversion of CAPEX into OPEX has become increasingly attractive as organizations seek to exert better control over IT expenses.
Converged infrastructure platforms can bring greater efficiency in IT operation and systems management, but the technology is highly disruptive and may not fully interoperate with existing data centers. CI deployments should be clearly justified with a return on invetment analysis and vetted with a thorough proof-of-principle project. In addition, organizations can reduce the potential for future disruption by evaluating the CI vendor’s product roadmap against anticipated computing needs. If the case for CI is simply too weak, organizations should stay with a more traditional data center model.
How can converged infrastructure help you save money in the long-term?