Last week, I explored some of the key issues and core benefits that are prompting enterprises to move to more flexible and cost-effective composable infrastructures. As I pointed out in Part 1 of this blog, composable infrastructure technologies from vendors like TidalScale are designed to address many of the most pressing issues in today’s data centers, such as the rapid growth of data, the challenges of accommodating unpredictable workloads with traditional servers and rack systems, and the inherent inefficiency and outright waste that comes from provisioning servers that cannot address the needs of new-generation applications and those that are dedicated to running just one application. In this part, l will review the role of software-defined resources in ensuring that composable data centers are a realistic and cost-effective end goal for enterprise digital transformation.
How essential are software-defined resources to the composable data center?
A self-imposed incentive by most industries to transform themselves digitally is fueling the demand for a new infrastructure architecture. Lines of businesses are mandating IT organizations adopt a software-defined and service-centric approach to speed up IT provisioning, optimize application performance, and increase IT efficiency. The endgame will be to run the business with processes and operations that are entirely digital. The degree to which today's organizations can be competitive and ultimately profitable are now inextricably linked to decisions made within the datacenter, putting IT organizations at the center of this transformation. The fact that a company's IT will now directly impact business strategy and revenue generation, combined with the ever-quickening pace of the competitive environment, places increased demands on the IT environment.
Since software-defined storage and networking technologies are already are in place in many data centers, how crucial is the introduction of software-defined servers to achieving a truly composable IT environment?
Composability is an implementation of software-defined infrastructure. It is therefore based on many of the principles of SDI – which includes compute, storage and networking (SDI is a bigger paradigm and encompasses everything in the datacenter, including facilities). The introduction of software-defined servers (compute) nicely complements existing software-defined storage and networking deployments as it serves as the layer upon which the latter services can be delivered as "composable workloads" inside containers or virtual machines. In other words, taking a workload-focused view of storage and networking, on software-defined compute enables IT to realize the truly vision of a datacenter-wide software-defined infrastructure paradigm.
What’s necessary for composable infrastructure solutions to be successful?
From a buyer perspective, the rollout of composable infrastructure solutions will be accelerated by an explicit acknowledgment that no matter what the software layer, physical server configurations are rigid. Thus, they inevitably limit options while adding operational constraints:
- The finite amount of real estate in a rack server limits the configuration layout, having to account for processor, memory, onboard storage, interconnect, power distribution units (PDUs), fans, heat sinks, and so forth. The restrictions mean standard servers cannot be optimally configured for their designated workload. This results in underutilized resources and wasted overhead. In a standard server, it is often the case that compute-dense applications do not adequately use memory and disk slots, and in contrast, with storage-intensive applications, the CPU and memory are overprovisioned.
- Cloud service providers and large enterprises that strive for maximum efficiency in their infrastructure rely on customized servers — they adopt specialized servers for each category of workloads. However, managing and maintaining multiple configurations increases the complexity of the infrastructure. Ultimately, there is a need to consider new hardware designs and to demand a common platform that is flexible for different applications.
This makes the role of an operator incredibly complex and error-prone. Given the rapid pace of change in IT, adopting composable software on industry-standard hardware (in the interim, and disaggregated hardware in the longer-term) is a viable infrastructure solution for datacenters of tomorrow. This would include software-defined server solutions.
IDC expects that barring any breakthrough in the mass production of technologies like silicon photonics, (software-defined) composability will progress at a faster rate than (hardware-based) disaggregation — the former being a software-based implementation that can tie together traditional and next-gen hardware in the initial releases, and as the hardware disaggregation progresses, it can incorporate newer architectures into its fold. Given these limitations, in the short term, such infrastructures will be limited to computing platform vendors that are already in the server business and have such technologies already on their product road map. In the long term, composability will be more of an ISV play and disaggregation will be more of a hardware platform (i.e., systems/solutions) play. Of course, the two must be in lockstep with each other for the entire effort to be successful.