Virtual workload balancing is overlooked well past the eleventh hour in many enterprises. This seems counterintuitive, given that enterprises are becoming highly dependent on IT for success. Capacity planning, or workload forecasting, is all about optimisation. It is the process for determining how capable and prepared the IT infrastructure is to meet future workload demands whilst efficiently managing resources. This process is typically measured within the context of applications that run in rapidly changing, multi layered, virtualised and often cloud based environments. Capacity planning is critical for reducing cost and increasing business productivity. Good capacity management also saves time and helps to protect against unforeseen challenges caused by workload changes – and saves the urgent need to throw expensive resources at solving resultant performance issues and workload bottlenecks.
Today, capacity planning is a mandatory discipline for enterprises. The problem is that many companies either don’t have the right tools, the right staff expertise, or both. Fortunately the technology exists today to allow the majority of capacity planning tasks to be automated, doing the work more efficiently than a traditional expert would be capable of. It is now possible to continuously apply smart predictive analytics to optimise applications that run across countless virtualised systems. Automated predictive analytics, utilising artificial intelligence for IT operations (AIOps), can help an enterprise to gain a competitive advantage by optimally maintaining the crucial balance between cost and performance. Though the deployment of AIOps, organisations can effectively optimise their application availability and workloads. AIOps capabilities powerfully predict capacity needs, proactively balance infrastructure utilisation, while automating anomaly detection, response and resolution.
A company can then efficiently meet service levels and availability requirements, without introducing undue risk.
Scenario-based, or automated? Old school capacity planners are a rare occurrence now
Some enterprises choose to rely on staff expertise and scenario-based capacity planning to include the exploration of worse-case events, as opposed to continuous automated predictive analytics that solve performance models across the entire data centre. Automated predictive analytics produce regular reports that are able to show which of the systems are likely to violate future service levels, as well as when (and why) this will occur. Conversely, scenario-based capacity planning mostly follows a one-at-time modelling. A scenario-based approach works great for fragmented “what-if” analysis, but less well for a continuous overall analysis. It will answer detailed questions such as ‘Which of the strongly cost-controlled infrastructure configurations will still meet service levels?’ Or: ‘How will the entirety of applications respond after adding a new application to a production system?’ Whilst the scenario-based approach allows for a wide array of possible situations, including experimentation with some extreme challenges, automated predictive analytics are what capacity planners will find most useful and less time-consuming in their day-to-day monitoring of applications and systems. The automated process does not require the prohibitive amounts of up-front capacity planning time or staff expertise that the old school capacity planning model does. By relying on automated predictive analytics enterprises can be informed of future problems well in advance of their occurrence, providing sufficient insight to enable organisations to avoid issues entirely.
Application performance and utilisation is optimised by continuous re-balancing of the underlying infrastructure, including VMs, network paths, and storage load distribution. State-of-the art custom AIOps analytics use machine learning to memorise VM workload patterns, then recommend the optimal placement of VMware ESX clusters to proactively avoid memory or CPU contention. The best analytics tools on the market also use machine learning to monitor port utilisation on storage ports, and provide recommendations on rebalancing ports on the storage array. Enterprises deploying such tools can reduce their infrastructure costs by 30-50 per cent after deployment by avoiding unnecessary infrastructure spending.
Looking at the latest market developments, good workload forecasting and AIOps will always be part of an advanced infrastructure performance management (IPM) solution. Infrastructure performance management utilising AIOps capabilities is key when it comes to activities such as system measurement, application monitoring, and general optimisation. An efficient IT infrastructure relies on highest levels of transparency to be able to proactively control or avoid issues.
Infrastructure performance management (IPM) - common beliefs
IT environments are invariably managed by humans so some common beliefs persist: ‘I don’t need any IPM unless I have a real issue, hardware is available at reasonable prices and hence replaces constant IPM solutions – one can always throw a new device at an issue’, ‘IPM and forecasting once a year will do’, ‘IPM and capacity management take up too much time’, ‘Analysis of workloads and general interpretation is too complex and mostly obsolete’.
Why do these attitudes persist when there are so many reported cases of outages in the press, as well as innovative solutions in existence to solve these issues? Many enterprises, especially throughout Europe, are slow to realise that their IT infrastructure lacks transparency. Appropriate measuring tools capable of mitigating performance issues and delivering valuable insights on future deployments in support of growing business requirements, will make an enormous difference in assuring business continuity. Just because a major disaster has not hit yet, or sudden issues can be solved by doing some quick firefighting that mostly relies on the expertise of senior staff, doesn’t mean this lucky streak will go on forever.
The current mismatch between proven IPM solutions utilising AIOps capabilities that deliver proper workload forecasting, and the stubbornness of enterprises unwilling to turn their IT infrastructure into a model of transparency, can only be solved by enabling IT staff with the necessary insight into the end-to-end IT stack. Enterprises that neglect IPM are at risk of learning the value of true transparency the hard way – in the event of disaster replacement or outsource of those responsible is always an option.
Louise Dilley, Regional Services Director EMEA, Virtual Instruments