Every industry wants productivity enhancing tools and reportable metrics, and of course the data center industry is no different. The premise of data center infrastructure management (DCIM) is the proverbial “single pane of glass,” which promises to alleviate the challenges related to the traditionally cultural and management issues of IT vs. facilities. Even the term “infrastructure” is at issue. If you ask different people in various domains of the mission critical environment what "infrastructure upgrade” means, you will get some diverse answers. To put things into perspective, facilities personnel are concerned if the pipes carrying the chilled water need to be upsized to meet the rising heat load. In contrast, the IT network staff may be wondering if the existing copper network cabling will continue to meet the rising bandwidth requirements or if they need to upgrade to fiber.
Nonetheless, there is a real need for a system that can collect and aggregate information from facilities and IT systems and then be able to analyze it and deliver a cohesive picture in a manner that is meaningful and correlates to some actionable items for everyone. Originally, most DCIM products were fairly basic and were more of a facility side power usage effectiveness (PUE) dashboard. DCIM programs have now gone through several generations and in their current iteration have become a far more sophisticated tool set.
DCIM is depicted as the ultimate data repository, able to deliver an integrated view of all the assets and status of the physical facility infrastructure (space, power, cooling, and network cabling), as well as IT systems (servers, storage, networking, and even applications). Nonetheless, it begs the question, is the view through one looking glass into the data center really holistic and will the view it provides be meaningful for everyone?
So how do CxO-level executives interpret and evaluate the usefulness of the promised deliverables, when the two traditionally opposing facilities management & IT factions within their organization cannot even agree on what DCIM means, much less who should specify, purchase, install, and operate DCIM systems? Moreover, from the CFO’s office, the underlying question is “Where is the ROI?” to which the answers to date have not been very clear cut. Considering the costs of some of these products, as well as the related expenditures and efforts to implement them, it is important to exam the direct business benefits and the less tangible, yet important, operational paybacks that DCIM product vendors are claiming to deliver.
Before beginning the acquisition process for a DCIM system, any organization should begin by creating and sorting out a list of their own data center’s primary requirements, as well as their key pain points. It is extremely important to engage the primary stakeholders: facilities, operations, and IT teams and make them part of the meeting to determine the DCIM requirements. Ultimately all should play an active part on the product evaluation committee. This is a time consuming process in itself and should be done well before calling the legions of sales reps for the proverbial dog and pony show. That is not to say that during the multitude of vendor presentations they will not discover a useful or “must have” feature or function that they had not initially envisioned, and then consider adding it to the final requirements list.
While the facility side aspects of DCIM are fairly mature (although they are constantly being refined), the IT capabilities continue to develop. Some of the more advanced DCIM offerings are currently capable of also looking directly into the IT hardware to obtain real-time power draw, power supply capacity and status, air intake and internal operating temperature, fan speed, CPU utilization, etc., as well as even controlling maximum power draw (“power capping”). This can eliminate the need to deploy facility side hardware such as cabinet by cabinet air temperature sensors, and intelligent cabinet level PDUs (although branch circuit monitoring to each cabinet at the electrical distribution panel is still advisable to avoid circuit overloads and tripped breakers).
Some DCIM packages can even access data from the IT operating systems, virtualization software, and even the applications, allowing them to correlate to their resource/power utilization (some are agentless, while others require installing software). This allows them to deliver a highly granular, integrated, and more meaningful insight of the total data center operational performance. It should be noted that IT may already have some of this information already available via a variety of IT management software packages (e.g., OEM hardware management tools, asset management tools, software consoles, and network management). However, integrating these more advanced IT capabilities into a centralized DCIM console may offer some additional strategic benefits for those organizations willing to invest the effort and capital cost. This type of view can help optimize and correlate IT hardware and even application performance in relation to its energy usage. It is this level of analysis that is a unique benefit to help deliver improved IT system performance, better resource capacity planning and provisioning, as well as overall energy efficiency optimization and system availability.
However, is this or the next generations of DCIM products extensible beyond the walls of the facility to also manage cloud based resources?
Like all new “solutions” DCIM has gone through several iterations in both technology and scope over the past five years. More recently DCIM offerings by some vendors have begun to refer to its current development stage as DCIM 2.0.
As DCIM matured, it has encompassed more IT centric functionality. In the long term this IT aspect will become a potentially more significant toolset as the data center becomes a highly virtualized resource available in many forms, rather than just bound to a particular physical data center facility. Moving forward, DCIM capabilities will increasingly need to focus more on the monitoring and managing of virtualized assets and the software defined data center (SDDC) as well as external computing resources. This means that DCIM systems will not just need to gather and integrate facility-based information (assuming that the organization still operates its own data center), such as IT hardware, virtualization software, and application information, it will also have to be able to query multiple external cloud platforms (IaaS, PaaS, SaaS, etc.) across diverse cloud service providers (which currently all have proprietary consoles and tools).
Therefore, a DCIM system must provide an integrated global view of any and all external virtual and physical IT assets (even if they are geographically distributed, which is becoming more common). It should also be able to provide a cost basis model (e.g., localized cost of power, site PUE, IT resource efficiency, cloud service costs), capable of analyzing and optimizing the entire operational cost model. This would help optimize resource and operational workload allocation, as well as a basis for departmental bill-backs or cost sharing. Nevertheless, while the eventual results are worthwhile, like any project there are implementation challenges and costs to consider.
One of the main impediments of DCIM adoption for existing facilities is related to retrofitting energy monitoring sensors. Older sites in particular are most likely to gain the maximum benefit from installing granular energy monitoring, particularly for the cooling system’s individual components (CRAC, CRAHs, chillers, pumps, fans, etc.). This information can then be used, along with aisle and rack level environmental sensors, to optimize the cooling system’s overall efficiency. However, energy monitoring generally requires the installation of current transformers and potential transformers in the electrical panels. In some cases this may require that panels be de-energized due to safety requirements, resulting in equipment shutdowns. Obviously, this is a significant issue for lower tier facilities (without redundant power paths), and in some cases even for those with redundant systems, where any type of equipment shutdown requires management approvals.
Regarding monitoring IT power usage, existing sites, that in many cases, IT has addressed this on its own by installing so called “intelligent” cabinet level PDUs (power strips). These generally are connected to an IT Ethernet network (existing production or independent) and typically use TCP/IP to communicate to the vendor’s own management software package. Most DCIM vendors are able to address third party or competing vendor cabinet-type PDUs out of necessity, in order to be competitive and functional in the multi-vendor environment common in most data centers.
In new designs, pre-installing energy monitoring sensors in the main electrical distribution panels is a simple low cost decision and has almost become a de-facto standard in most new facilities. The major cost driven decision is primarily related to how granular the sub-panel breaker monitoring becomes for the cooling systems. However, branch level monitoring has become common, especially on the distribution panels servicing the IT cabinets. It is simpler and less costly to have panel-level energy monitoring on every branch circuit feeding the IT cabinets. Many of the colocation providers already use branch circuit monitoring to charge their customers for monthly energy usage, as well as for ensuring that they do not exceed the circuit ratings (especially for A-B redundant circuits).
The total cost of a DCIM implementation project can be segmented into several key areas: the basic price of the DCIM software, the cost of the energy monitoring hardware and environmental sensors, and the related cost of the hardware installation and wiring required. There are also substantial consulting costs related to systems integration with existing hardware or BMS systems. Moreover, the level of granularity generally related to the number of devices (managed assets) to be monitored has a direct relationship to the final overall cost. When soliciting proposals, be sure to have a clear understanding of each vendor’s licensing terms and costs (per device, node, rack, site, or any other basis). In addition, review the costs for any added modules or functions, as well as annual support.
When considering which system vendor to select, operator training and ease of use can make or break the DCIM project by enhancing or diminishing its effectiveness and value. When evaluating the product offering also consider the additional cost, time, and resources for the training (is there formalized vendor training or just learning on your own system). The scope of training may also provide insight into necessary ongoing personnel resources required to manage the DCIM system once it is operational. Caveat, if the product requires more resources to maintain itself, rather than those tasks it purports to optimize or automate — is this solution a worthwhile investment?
THE BOTTOM LINE
DCIM may have started primarily as a BMS/BIS facility-type platform, with data center specific features (such as a PUE dashboard), however the nature of entire data center centric computing infrastructure and enterprise IT architecture itself has changed radically since DCIM’s inception. The days of the traditional enterprise data centers are diminishing as more organizations are moving toward outsourcing data center ownership and operations by moving to colocation providers and cloud services.
In effect, the ideal enterprise class DCIM platform must be able to adapt and evolve at the same rate as the constantly developing new IT architecture and external assets, as well as environments it will need to manage, be it any combination of physical or virtual resources. The next generation of DCIM must therefore be able to seamlessly interface and integrate with internal IT systems and physical data center infrastructure and IT hardware (and ultimately the applications), as well as any external service providers.
Ultimately, the question of should a DCIM be purchased now in its present level of development or wait until it further matures, will depend on the nature and culture of the organization. From a practical perspective, for those enterprises that are more traditional or conservative and will continue to own (or control) their own dedicated IT resources, the current crop of DCIM products already offer a wide breath of useful tools.
For those more leading edge organizations that are less bound to a physical facility and IT hardware ownership, consider testing and piloting the DCIM platforms that are focused on managing virtual and cloud-based services. When doing vendor evaluations do not just assess their current product’s capabilities, but investigate the next version product deliverables, as well as their overall roadmap to see its direction, and if this matches your organization long- term projected requirements for DCIM as it evolves
One of the ultimate goals of optimizing a data center’s overall efficiency is to be able to measure the utilization (and throughput) of various applications and relate that to its energy consumption. That information can then be used by IT to make more intelligent, holistic decisions about IT performance and energy efficiency and related TCO. It can also be used to compare it with cloud-based services and determine capacity planning, as well as short- and long- term strategic and financial decisions that can help justify the ROI question.
To what degree, as well as who will fund and be primarily responsible for implementing, operating, and managing the DCIM system, can be an impediment to success in itself. Especially if the old politics of IT vs. facilities come into play during the original investigation of the requirements gathering phase, as each faction pronounces what they expect the system to do, as well as how it can impact their own systems, or perhaps even their jobs.
Nonetheless, unless you clearly define the tasks you explicitly need DCIM to address in it its initial implementation, as well as the expectations for futures phases using expanded capabilities, you will wind up with endless feature creep and the possibility of a failed or less than successful project. So while some of the DCIM vendors promise to be all things to all customers, like the proverbial Swiss army knife with 20 or more blades, look for the major functions that you actually need, not just fancy dashboard graphics or mobile apps that just look enticing.