Experts at 451 Research point out that the value of data centre infrastructure management data multiplies when integrated, shared and analysed at scale. DCIM-based cloud services or DMaaS solutions are emerging but could artificial intelligence hold the key to unlocking even greater potential? Louise Frampton reports.
New Cloud services for data centre infrastructure management are transforming traditional on-premises approaches. By applying Internet of Things technology, DCIM-based Cloud services can deliver the benefits of DCIM without the deployment costs and, over time, “promise value beyond what is possible on-premises”.
This is according to experts from technology research and advisory company 451 Research and the Uptime Institute, who recently provided industry insight as part of a webinar on ‘the evolution of DCIM’.
Mark Harris, senior vice-president of marketing for the Uptime Institute, comments: “When expectations are set properly, DCIM can be very successful. Data centres have become more complicated over time and more difficult to manage in ‘someone’s head’; we are moving away from best case ‘folklore’ and ‘tribal knowledge’ towards a requirement to have proper systems, processes and policies in place that can handle this complexity.”
Quoting Joe Kava, Google’s vice-president of data centre operations, Rhonda Ascierto from 451 Research points out: “It is impossible for mere mortals to visualise how best to optimise the data centre in real time. However, it’s fairly trivial for computers.”
She adds that the best-run, most-efficient data centres – from hyperscale to micro data centre, Cloud and colo to enteprise – are increasingly data driven.
“People have been saying to me for a few years now that you only need a few simple monitoring points to understand your data centre and to get the most of it. But I haven’t seen that bear out; I don’t think it’s true. Data centres are going to get more and more complex, particularly as workloads change,” comments Ascierto. “I’ve worked with and talked to a lot of people who have implemented DCIM with varying degrees of success and it is clear that DCIM is as much – if not more – a management approach as it is a set of tools.”
The most successful DCIM projects, according to Ascierto, are realised when the business requirement has been carefully considered at the outset. She points out that there is a lot of data obtained within data centres but this tends to be siloed, inconsistent and in different formats so it is hard to compare and integrate into a single dashboard. DCIM overcomes these challenges by aggregating and normalising this data. The traditional DCIM-only approach comprises a number of core components: monitoring; asset management; and analytics, predictive modelling or forecasting.
“The benefits include increased efficiency and reduced costs, and lowering risk of IT disruption. Increased efficiency is about driving up utilisation, for example by identifying stranded power/cooling and avoiding over provisioning. Forecasting is another significant driver,” says Ascierto.
451 Research has conducted its own studies on DCIM and return on investment (ROI). Most of the savings are realised through reduced operator time and resources; and being able to perform on-demand audits.
However, a benefit that 451 Research is hearing more about is the ability to provide power usage data for tech refreshes. Other important benefits include improved fault detection, root cause analysis of failures, real-time visibility of redundancy, the ability to analyse ‘what if’ scenarios to avoid errors, and forecast/anticipate capacity issues that may impact power use and hot spots.
“DCIM is the foundation layer of what is termed the ‘data driven data centre’,” says Ascierto. She explains that the more forward-thinking data centres that are further down the road with DCIM, use the data centre service optimisation approach (DCSO), which integrates DCIM with IT systems and business systems.
These are analysed together, which allows organisations to better match the supply of physical resources at the facilities layer with demand from IT.
“Once this integration has been performed, this paves the way for automation. This is usually the fourth or fifth step of the deployment. So why do this? It all comes down to lowering risk of IT disruption,” says Ascierto.
She explains that the benefits of the DCSO approach include the ability to map the workloads to the assets, including the supporting critical infrastructure; ensuring the physical resource is available and running at the performance that the application requires; real-time load shifting, including pre-emptively when the DCIM flags up an issue – so it can move load away from the area; while traceable work orders also mean you can reduce errors on the floor. It also enables increased efficiency and agility, so that you can load shed, load cap and perform precision load placement.
“The data that you integrate will depend on your business need. Colocation providers, for example, are integrating DCIM with CRM data, in relation to power consumption, so they can bill for use of assets,” says Ascierto.
A major, new trend in the future, according to Ascierto, could be the use of DCIM to support participation in demand response schemes. “There are calls for data centres to use generators more and/or to use less power at peak times. And of course you can trade on that,” she explains.
She highlights the example of a large enterprise data centre in the US allowed to run its generator for 100 hours per year. “The truth is, it doesn’t need to run it at all, except for testing, so the mission critical data centre has signed up for demand response. The utility can remotely switch on the generators when the data centre operator gives them the green light to do so.
“They performed a cost analysis and they broke even. They didn’t make money but what was valuable to them was the business insight; when the utility asked them to switch to demand response, it put them on high alert… they knew when the power was most likely to have a problem. It went well for them but, of course, it does carry risks, so they are in the process of putting in a DCIM system and integrating it with data about their workloads through DCSO. This means when the utility asks them to participate in demand response, they can look at what workloads are running and where. If certain workloads are running, they can decide not to participate in demand response. In addition, if they are on high alert from the utility, they can shift workloads. They are not making a lot of money out of this and do not expect to, but their business reason for doing this is about reducing risk,” Ascierto comments.
The benefits of integrating DCIM data are the ability to better match supply with demand from IT, the provision of end-to-end visibility of what is going on in the facility at any given time, as well as greater insight into the criticality of the workload. It also offers the ability to enable improved operational and new business approaches, but it does require ‘orchestration’, notes Ascierto.
Ascierto points out that a major new development is the emergence of DCIM-based Cloud services or DMaaS. Monitored data about equipment is encrypted and sent to a service provider’s Cloud. Initially developed as mobile systems, full feature systems are now evolving. While some of the data remains completely private, a proportion of the data is anonymised and pooled with data from hundreds of thousands of other facilities; this can then be analysed and shared.
“The reason this is a big deal is that it enables big data analytics,” explains Ascierto. The statistical analysis of large pools of data means that outcomes can be better predicted, which is a major difference.
In the future, the goal will be for DCIM to offer industry benchmarking through DMaaS, which will enable data centres to gain an insight into how their performance compares with others. The aim will also be to provide benefits such as facility risk mitigation, budget impact analysis and recommended actions – not just in terms of infrastructure and equipment but also taking into account data on weather and power. Remote troubleshooting will also be offered in the long-term, along with outcome based services (for example, guaranteeing a certain cost per kW or PUE.)
“There are some things that DCIM does locally that you may want to keep local – for resiliency reasons, in case you lose the network; or you may be legally obliged to keep data on site,” Ascierto acknowledges.
“Furthermore, if you are trying to catch a failure in real time, such as an electrical fault, do you want to go to the Cloud and back?”
Together, DCIM and DMaaSS could provide a powerful combination but there is even greater potential on the horizon: artificial intelligence is another important trend that could integrate with DMaaS services.
As Ascierto points out, machine learning in Google’s data centres successfully cut power for cooling by 40%. This problem solving was achieved without involvement of any human being.
“What if this became available to everyone? This is the big hope and promise of DMaaS,” Ascierto concludes.
To view the webinar in full, visit: https://uptimeinstitute.com