Is Hosted UC Really the Way to Go?

Is the hosted UC cloud really the way to go? In a two-word answer, “It depends.”

I’m writing this post based on my personal experience with hosted UC solutions and presenting a slightly different perspective beyond the hype of the UC cloud. Personally, I think the cloud industry players are overselling the OpEx cloud model, as compared with the level of maturity expected in a premises-based model.

 

Some technologies that we've been discussing forever are now going mainstream. These include video conferencing, omnichannel contact centers, mobility, UC APIs, and of course the UCaaS and CCaaS hosted solutions.

From my perspective, we’re still in the hype cycle of the UC and contact center cloud. Yes, many applications, such as Salesforce and Microsoft 365, have already moved to the cloud successfully. These applications don’t require the real-time characteristics of UC and the contact center. Solid quality-of-service (QoS) -- meaning minimal jitter, delay/latency, and packet loss -- is a necessary requirement for a great UC and contact center experience. In addition, the traditional five nines (99.999%) uptime model (meaning five minutes of outage annually, or essentially always on) is also a requirement. Several CIOs that we work for have wanted to remove the telecom headache that plagues them, as UC technologies are the most sensitized to various network dips and possible network anomalies.

According to Zeus Kerravala, of ZK Research, UCaaS sales already are far greater than premises sales, but the on-premises installed base is so much larger than the cloud base. But Zeus said he expects the installed base to be more than 50% cloud seats vs. on-prem seats by 2023. Whether this shift will be driven by the hype of the last few years or a real need for enterprises to move to the cloud is a matter of debate.

No matter how many applications move toward a cloud solution, call quality and uptime will always vary by provider.

To date in my experience with clients, cloud reliability isn’t quite on par with a premises-based model. In the world of digital telephony, enterprises historically took a five-nines reliability model for granted -- it was a communications rite of passage and requirement. VoIP then lowered that bar slightly to four nines, with the introduction of network dependencies and network redundancy required to get to that five-nines model. Over time, VoIP has moved to a five-nines model, and therefore the expectations for five nines continues.

I haven’t seen cloud telephony meet the reliability of a five-nines model or deliver on QoS consistent with a premises-based system. Case in point, downdetector.com, which tracks cloud uptimes, indicated that in first-quarter 2018 outages for a national UCaaS provider equated to more than 10 hours per year. That’s below three nines (99.9%) availability, the equivalent of a single circuit with no redundancy. In my experience, a three-nines model is the baseline for circuit delivery and not for real-time communications applications reliability.

Another case in point, our firm uses one of the national cloud providers, and in the last 12 months outages have exceeded 15 hours (99.8%) -- and the problems weren’t on our network. This again is well below three-nines availability -- far below the required reliability models enterprises are used to at five nines.

I’ve seen such poor performance lead to demotions or firings of those responsible for telecom and UC. At the very least, companies start to question their authority.

Some UCaaS providers now offer contract agreements that deliver on a five-nines model, as long as dual SD-WAN or MPLS connections go back to their data centers. These contractual penalties are just that, contractual, and don’t displace the design requirements for a five-nines model. While these providers are putting some skin in the game, generally speaking these penalties don’t have much teeth in them for non-performance.

Some UCaaS providers deliver on shared-resource platforms, others on virtual private platforms, and others still on a private-cloud hybrid approach, with redundancy within the data center only… geo-redundancy optional. With any cloud solution, our consulting practice recommends geo-redundancy; data centers should be separated by at least 500 to 1,000 miles.

And now the idea of customer experience, patient experience, and member experience is front and center; they’re the key buzz terms we're now socializing and hearing about from every organization with which we work. The need to get reliability and QoS comparable to that of a premises-based model is even more important than ever.

Yes, the UC cloud is here, but at what financial cost? Based on my work with clients, I’ve found that over seven years, the total cost of ownership for cloud compared to a new premises-based solution will be anywhere from 40% to 50% higher. And that variance can be as high as 100% in the case of upgrades and incentives offered by premises-based vendors.