By Paul Mercina
What is the cost of downtime at the edge? Statistics for server and network outages—usually expressed in the thousands of dollars per minute — have been bandied about for some time. Edge computing, however, is new enough that little in the way of reliable, industry-transferrable information is yet available.
To judge from the use cases, the price will be high. From oil pipeline shutoffs to traffic management in smart cities, mission critical decisions will be made at the edge. And consumer expectations of financial services applications, entertainment options, and retail experiences will continue to skyrocket, sending customers and revenues to companies capable of outperforming the rest.
Thus failures at the edge will run the gamut from the potentially life-threatening to the brand-undermining. But edge delivery models are in their infancy, and enterprises are just beginning to explore the intricacies of the corresponding maintenance challenge.
Although some IT leaders hope that the distributed nature of edge computing will contain the impacts of any outage, it is unlikely that uptime—even narrowly local availability—will become any less vital in this new era. The question is how to prepare for the unknowns in order to drive innovation, cost-efficiency, and performance.
A Familiar Roadmap to a New Destination
Much can be extrapolated from experience. The maintenance struggles of the average telecom managing truck roll after truck roll, or the daily frustrations of network administrators overseeing increasingly diverse, hybrid infrastructures, provide a peek at the complexities waiting at the edge.
As in past evolutionary stages, the enterprise will not automatically gain slack in the budget or staffing to reinvest in solving new problems. To the contrary, simply keeping the lights on will become more difficult as the number of sites to support multiplies from a handful to the thousands and beyond. Yet digital transformation will still be the overarching goal.
It is reasonable to expect that edge computing will undergo a similar trajectory as previous technologies. Without adequate knowledge of proven, affordable support strategies tailored to edge environments, many organizations will initially pursue a reactive maintenance approach. They’ll run equipment until it drops, repair or replace it as best and as fast as possible, and then, hopefully, learn from the experience.
This will gradually lead to more preventive maintenance, in which time-to-failure and other information guides proactive efforts to reduce downtime. Lifecycle management initiatives and regular field technician checkups, for example, could help enhance reliability, albeit at a significant cost. Fortunately, with big data analytics, preventive maintenance can be more effectively targeted based on the myriad factors contributing to failures and outages.
The ultimate goal will, of course, be predictive maintenance, a truly forward stance toward enterprise infrastructure. Sadly, for many multinationals, let alone SMBs across India, predictive maintenance is still an aspiration within their largely centralized infrastructures. Network and data center managers already suffer severe alert fatigue, so an explosion of “blinking lights” coming from edge systems could easily overwhelm response capabilities.
The solutions will not be as simple as ruggedized equipment designed to withstand the elements. Applying the field technician-driven support strategies of years past would be too expensive and result in downtime that is too long. And even many current remote management and backup/restore approaches will fail against the low-bandwidth, intermittent-connection realities of some edge deployments.
It is relatively easy to define what edge maintenance cannot be but far more difficult to determine what it must include. The general outlines are, however, beginning to emerge from the fog.
Elements of the Edge Maintenance Solution
Superior edge performance will rapidly become a differentiator, so Indian companies—both established and new market entrants—will be seeking to accelerate the transformation from reactive to preventive to predictive maintenance, or even skip stages along the way. This will require a view of IT maintenance that transcends immediate break/fix concerns and takes into account a comprehensive view of infrastructure management.
Four key elements will most certainly make an appearance in an effective edge maintenance solution:
- Discovery tools to understand what is installed on edge networks and provide oversight and adjustment as organic and acquisition-based expansion occurs. Tracking the shifting network topography and implementing standardized server configuration changes, among other tasks, will require sophisticated applications.
- Remote monitoring and, increasingly, remote management and repair to reduce technician call-outs. The automated, machine learning-driven options applied to today’s data centers will grow to accommodate ever more complex infrastructure.
- An affordable support model with the geographic reach and specialized expertise to respond to all edge deployments. For many companies, this will mean some form of third party support delivery to share field engineering, spare parts, and replacement hardware resources, which will be called on rarely but urgently.
- Optimization systems to fine-tune performance across a network of highly distributed systems.
There are many parts of this equation yet to be worked out, and there will not be a cookie-cutter edge maintenance solution for all industries. What edge means—and what it demands—in precision agriculture or aerospace will differ from in banking or healthcare. Hardware, operating systems, and protocol choices will diversify to such an extent that today’s multivendor environments will, in retrospect, appear laughably simple.
The heterogeneous nature of edge computing will, therefore, require flexible IT maintenance options to fit the use cases. All of this variability notwithstanding, it is clear that securing a competitive edge at the edge will require a deeper look at discovery, monitoring, support, and optimization.
(The author is Director of Product Management at Park Place Technologies and the views expressed in the article are his own)
