Organizations have a lot of choices beyond the old one-to-one physical IT infrastructure strategy of servers, storage and network components underlying applications. So what's the right choice for your IT infrastructure strategy?
To support diverse applications, including a growing number of as-a-Service tools, a flexible infrastructure must replace legacy systems.
Options include converged or engineered infrastructures, such as Cisco Unified Computing System (UCS), IBM PureSystems or Dell VRTX. Or IT shops could go for a more cloud-like scale-out model based on commodity servers. Perhaps they should completely forget about the physical platform and go for a public-cloud-based service.
Each infrastructure option has its own strengths -- and its own weaknesses.
Converged systems take away a lot of the technical issues that organizations run into with a pure scale-out approach. Vendors engineer from the outset how the servers, storage and networking equipment within a system work together, simplifying management for the IT team. However, expansion is not always easy. To grow IT capacity often requires over-engineering, implementing another converged system alongside the existing one just to gain the desired headroom. There is also the issue of heterogeneity -- management becomes difficult if the fabric network consists of multiple vendors or if converged systems from more than one vendor are in place. Homogeneity is key.
A private cloud environment may then be a better option. Although private cloud can, and generally should, be implemented on converged systems, the majority of implementations run on standard high-volume servers built into racks and rows with separate storage and network systems. Adding incremental resources is far simpler in this approach than with converged infrastructure pods. New servers, storage and network resources can be plugged in and work with the rest of the platform reasonably simply with proper management software.
The IT team must possess the right skills, encompassing the technical aspects of how such a platform works, and expertise in, for example, how to populate a rack or a row without leaving hot spots or overloaded power sources.
Converged infrastructure shops and private cloud adopters must ensure that their management software does not just focus on one aspect of the platform: The virtual environment has dependencies on the physical, and these must be understood by the software. For example, when a physical disk system fails, it affects any virtual data stores that sit on that system. If the management software understands this set up, it will direct backups or mirrors onto a different physical storage system. Comprehensive management software enables business continuity, minimizing downtime and automating recovery as much as possible. Products should support hot images, data mirroring and network path virtualization across multiple physical network interface cards and connections.
Public cloud services, whether Infrastructure, Platform or Software as a Service (I/P/SaaS), appear to remove all the issues around managing the platform. However, the IT team needs infrastructure visibility to identify any trends that the cloud provider might miss. For example, if storage resources are running low, the business needs to know if end-user response is suffering and what scenarios could occur.
Management makes it all work
In reality, most organizations' IT infrastructure strategies end up including a hybrid mix of the above options. This brings in more issues. Whereas a single converged system can pretty much look after itself, once it needs to interact with a public cloud infrastructure, it requires extra management services.
Whatever platform your organization implements, choose software that can look at the system from end to end. End users always care about system response time. A converged system tends to look inwardly, monitoring internal connection speeds. The converged infrastructure will report that everything is running at hyper-speed, while an end user accessing applications from a hand-held device over a public network and using a mix of functions from the converged system and a public cloud might disagree. The management software must be able to monitor all of these factors in play and identify causes of the problem.
Management software can attempt to resolve issues via several channels: using a less congested network, offloading workload to a different virtual machine or applying more storage. Any change has repercussions: adding network bandwidth means higher IOPS, which could require a different tier of storage or for more virtual cores to be thrown at the server. The adjustments should be carried out in essentially real time to be invisible to end users.
The existing systems management vendors -- CA Technologies, IBM and BMC Software -- are getting there with their products; HP is lagging behind. The data center infrastructure management vendors, including Nlyte Software and Emerson Network Power, are making great strides by adding monitoring and management of the data center facility and its equipment to existing systems management tools. EMC and VMware are making a play for the market through the software-defined data center or SDDC strategy, but they may need to bolster their understanding of the physical side.
About the author:
Clive Longbottom is the co-founder and service director of IT research and analysis firm Quocirca, based in the U.K. Longbottom has more than 15 years of experience in the field. With a background in chemical engineering, he's worked on automation, control of hazardous substances, document management and knowledge management projects.