BACKGROUND IMAGE: iSTOCK/GETTY IMAGES
For a true converged infrastructure comparison, start with a detailed understanding of the products and of your IT workloads' particular resource demands.
Converged infrastructure (CI) vendors mandate the hardware and software combination in each product, ensuring the pieces all work together. As a result, the infrastructure is delivered as a fully functioning unit. By offering a small selection of different converged infrastructure units, vendors simplify the testing and ordering processes.
Different CI configurations usually relate to scale. One configuration might work for a few hundred virtual machines (VMs), while the next complements a few thousand. IT shops should find a configuration that accommodates their workload.
Cost is always an issue for IT, but converged infrastructure is a long-term play that pays dividends. Though the initial price is more than the sum of the components separately, the cost of integrating these parts to run VMs is more than the CI premium.
Ready when it arrives
Because the CI vendor works out integration architecturally, the data center can immediately focus on putting the infrastructure to work. The vendor selects each element to perform in conjunction with the others. Every piece of hardware, version of firmware and a lot of settings are fixed to ensure compatibility.
Providers thoroughly test new firmware and configurations, so buyers do not waste time on changes.
Converged infrastructure allows customers to focus on their workloads, not an infrastructure's components. Everything below the VM hardware is defined by the CI product and validated by the vendor. The customer looks after the VM operating system and applications -- without worrying about the physical infrastructure.
Converged infrastructure means:
Storage, networking, computer and management software in a single unit;
Few choices in terms of configuration;
The alternate approach is "best of breed," where the data center's team chooses each component. The servers, network and storage arrays are selected separately, often from different vendors. This puts the onus on the IT organization or an IT integrator to get all the parts working together, and to validate each firmware upgrade and settings change before they roll out to production.
The limits of CI
IT shops need an appliance to run VM workloads, but one size does not fit all. The biggest problem with CI is that it is too fixed. Its rigid structure could lead to waste if resources get saturated in certain areas and not others.
If a unit is designed for server virtualization -- for example, a virtual desktop infrastructure (VDI) – workload will likely use all of the CPU capacity before filling even a small percentage of the disk capacity. Even worse, VDI will make storage performance demands that are very different from the needs of a server platform.
Not every server platform is the same. For example, a media company might use all of the disk capacity and need comparatively little CPU, whereas a medical research organization may utilize every available resource.
To address this, CI vendors offer varying resource balances for different workloads, making a converged infrastructure comparison easier. A dedicated VDI unit is the most common offering, as large-scale VDI is different than server virtualization; the workload-specific units are still a generalized balance of resources. Each customer's workload is slightly different and may not fit well on a fixed configuration.
CI products rarely have space to grow and meet increased load needs, so customers should plan for larger-than-expected growth and buy the unit that accommodates this demand. A client expecting a 7,000 VM rollout may need to choose between a 5,000 VM setup and one that can support 10,000 VM units. The smaller unit will be saturated before the full workload is deployed. The larger unit isn't cost effective, since it will be lightly utilized even at peak.
Once a converged infrastructure unit is full, the user has to find budget to add a whole new unit. Delivery can take weeks or months.
Know your compromise
Who is right for the converged model? Startups are unlikely candidates -- the size of the organization wouldn't justify such a hefty infrastructure. Ideal fits for converged infrastructure are medium and large companies, particularly those preferring to focus on the applications rather than the underlying infrastructure.
CI products can significantly help the IT team that doesn't want to manage device and firmware compatibility conflicts. With CI, an in-house team focuses on the applications and users that make the business run.
There is an inherent comparison between converged and conventional infrastructure: An organization gains operational simplicity but sacrifices configuration flexibility when they go to CI. In the end, when it comes to a converged infrastructure comparison with other options, the question is: What delivers the best value for your particular organization?
The hyper-converged path
Another option is hyper-converged infrastructure. It provides many of the same benefits and addresses some of CI's shortcomings. While the latter combines existing storage, compute and network into a validated configuration, hyper-converged infrastructure combines the storage and compute into one server.
A cluster of multiple servers provides a platform for the workload. Scaling out the cluster requires additional servers -- a much smaller purchase than a new CI unit. Any number of servers can be added to accommodate the changing workload.
The risk with hyper-convergence is that it uses emerging hardware and software platforms and has new operational paradigms compared to converged infrastructure, which uses existing, proven hardware in a new way.
Are converged systems offering better storage options?
Can hyper-convergence solve your data capacity problems?
Are hyper-converged systems a threat to SAN, NAS?