Workloads are the workhorses of your mission-critical applications—they’re what get the job done. The environments that these workloads run in, and the way we manage them, have evolved significantly. Back in the old days, applications were built as monoliths. They were self-contained systems that included not just the capabilities you needed them for but also all the supporting functions—everything from the user interface to data storage and access—that enable those primary capabilities. In other words, all of the application’s workloads were tightly knit within that program.

Monolithic applications still exist, but modern apps are much more likely to be built using a microservices architecture where components that run a specific process as a service are assembled into applications. Because these components are independently developed and run, there are a lot of advantages to this approach. Applications are simpler and faster to deploy, are easier to scale, increase fault isolation, support more agile development, and much more. Of course, with these benefits come some trade-offs. Services are running in multiple applications and environments, which creates management complexity.

All of this matters because it fundamentally changes how IT infrastructures are managed. In practice, you no longer have a small collection of applications, each of which runs independently in a single environment, but a massive collection of services that all run interdependently in a variety of environments.

Effective infrastructure management requires smart workload management

IT infrastructures will continue to evolve with new technologies, architectures, and capabilities emerging and business requirements ever-changing. The idea is overwhelming. The important thing is to move beyond what’s outmoded to make the absolute most of what’s available right now. In 2023, this means taking four key approaches to managing your workloads.

Assess fitness for target environments

Each environment in a contemporary hybrid/multi-cloud infrastructure—on premises, public cloud, private cloud—has its own unique characteristics. A given workload may not be suitable for a particular destination. For example:

  •   A particular public cloud service provider (CSP) may not be able to meet a workload’s network demand based on historical on-premises values.
  •   A CSP may not be able to meet a workload’s performance SLAs—or may not be able to meet them at the desired cost level.
  •   On-premises components could include waste, such as zombie VMs, that if swept up into a lift-and-shift migration, will start costing you hard opex dollars in the cloud.
  •   Workloads that process sensitive data may need to be kept on premises to meet regulatory requirements.
  •   Workloads that process massive data volumes could balloon data costs if you move them to the public cloud—and it could be prohibitively expensive to migrate that data back out later.

This is just the tip of the iceberg. Because migrating an unfit-for-the-environment workload wastes time and money and is disruptive, these are not lessons you want to learn the hard way. Any-to-any workload movement planning requires automating the assessment process and getting recommendations based on the specific target environment.

Think about workload placement operationally and strategically

It’s not enough just to ensure the workload is fit for the environment. The way it’s deployed can have an impact on cost and performance. To make smarter decisions about workload placement, you should consider three key factors. The first is workload behavior during peak hours over a representative time period. It’s critical that you do not aggregate or sample this data to derive average utilization, as this can be highly deceptive. You need to understand when it reaches peak CPU utilization during the time series and how long it stays there. A workload that runs near 100% utilization almost 100% of the time is very different from one that operates at around 10% half the time, going up to around 80% for the other half and peaking at 100% for only a brief period.

The second consideration is the workload’s objective and your risk tolerance within that context. Does it process mission-critical transactions? In this case, you probably want to invest a little more to bring the latency risk as near to zero as possible. On the other hand, if the workload performs application backups, starting at midnight and finishing at 2 am, you might want to consider accepting a lower level of performance if it delivers meaningful savings.

Finally, you want to assess available capacity in your infrastructure—not just what’s currently free. For example, if your mission-critical transactional workload is only busy from 6 am to 6 pm, Monday through Friday, that expensive gear sits relatively idle for 12 hours each weekday and 24 hours over the weekend. So it may, in fact, make sense to use it for the backup process. Even though that backup doesn’t need or warrant the expensive processing power, this co-location could end up saving you money.

Keep workloads rightsized on an ongoing basis

Your business, and therefore your workloads, are constantly changing. Workloads that are fully optimized at initial placement can drift, altering the performance/cost/risk balance. As requirements evolve, existing workloads are updated or new ones added, which can change how they run in their current environment. Even in the absence of new or different business requirements, increasing demand or growing data volumes, for example, could cause workloads to outgrow their current configurations. Even if workloads are currently optimized based on the configuration options that were available at the time of initial placement, there could be new, better options, particularly in public cloud environments. CSPs are constantly offering new solution, configuration, and pricing options. If a new selection that’s a better fit for your workload becomes available—e.g., the same performance but at a lower cost—you want to know about it. To keep your workloads rightsized on an ongoing basis, you need to be able to do three things:

  •   Know what’s happening in your environment in real time to detect and alert to significant shifts in performance, capacity, or cost as they happen.
  •   Identify misallocated resources and unnecessary spend to eliminate waste.
  •   Quickly identify and move to the most stable and affordable configurations for your resources and requirements.
Plan ahead for workload needs

Of course, not all change is unplanned. Proactive planning is crucial to minimize the unexpected—and the last-minute scrambling and potential disruption it can cause. Forecasting is both an art and a science, but when it comes to anticipating capacity needs across your global hybrid infrastructure, you can load up on the science side of that equation by incorporating several capabilities into your capacity forecasting process.

First, you need to understand your workloads’ historical patterns. While it’s true that, to co-opt the old financial saw, past performance is no guarantee of future results, look-back data can be valuable. The key is to use it as one data point, but not the only one you use for future planning.

When you add scenario-based projections to the baseline derived from historical data, you can better understand the impact of decisions in an uncertain future. What-if analysis can help you understand the implications of new or changing requirements on capacity and budget, enabling you to make more informed decisions.

And finally, you should leverage the predictive power of AI to identify trends, forecast future growth, anticipate budget-exceeding spend, and more.

Set your workloads—and your business—up for success

IT infrastructures in 2023 are complex, heterogenous environments with a lot of moving parts. A smart workload-centric approach that takes advantage of current technologies and capabilities ensures that they perform effectively and cost-efficiently to meet the needs of your business.