Public or private: which cloud is right for you?

Public cloud infrastructure is the starting point for lots of discussions when our customers are looking at deployment options for a new project, but is it the only way to go? Adam Hughes talks to Cerillion’s Chief Technical Architect, Andy Blackman, on the options available to CSPs/DSPs when choosing the right cloud platform for them and some of the factors that might influence that decision.

Hi Andy, thanks for joining us today. There’s lots of talk about public cloud BSS/OSS solutions and we’d like to explore when public cloud infrastructure is a good choice and what other options might be available. The public cloud hyperscalers are all household names now, what’s made them so successful?
In a word – flexibility. One of the great things about public cloud is the ability to ramp up and ramp down on consumption as your usage dictates, and there are lots of situations in which this is beneficial.

Some of the common ones we see are scenarios such as start-ups and greenfield deployments where the growth projections underpinning the business model predict a swift increase in volume starting from zero or a very low entry-level. In this situation there’s often a focus on minimising initial outlay and in a solution that allows for rapid scalability.

Another would be about variable usage, where you might be expecting great fluctuations in usage in an unpredictable manner – the ability to turn up and turn down the consumption of resources can bring cost savings.

A third scenario would be in short term use, where the business activity is only planned to run for a finite and short period, for example a PoC or similar use case that doesn’t have the longer-term use that delivers a lower TCO.
 
Those all sound quite relevant to lots of business models, so why would people ever look at anything else?
Well, we’re always talking about cloud infrastructure, and the benefits to operators of devolving the responsibility for the infrastructure to suppliers are certainly clear. Not everyone wants to be maintaining a lot of kit. Having that taken care of by specialist companies allows operators to focus on their core business, which is delivering connectivity to their customers. But we often see operators whose situation means that the public cloud hyperscaler solutions aren’t always the right ones for them.
 
We’ve just discussed the benefits that public cloud infrastructure offers, where would people see a benefit in something else?
We see more and more of this, and there’s quite a range of scenarios. The public cloud infrastructure is excellent for those unpredictable models, but as with so many things in life flexibility comes at a price, and there’s a reason we see AWS and Azure generating such large profits.

Where an operator’s consumption pattern is more predictable and less volatile, then the need for extreme versatility isn’t there. As a greenfield launch or start-up business the public cloud makes a lot of sense to cope with “hockey stick” growth, but when you’re an established operator with single digit growth that flexibility can start to look very costly.

Local legislation and regulation might also mandate data sovereignty and the public cloud providers aren’t yet in every country. So, if your local authority requires data to be kept in-country then the public cloud might immediately put your data sovereignty at risk.

Another challenge we get asked about is security. A lot of the true public cloud co-located infrastructure options mean you never know with whom you’re sharing your hardware; is it a competitor, a bitcoin miner, a criminal enterprise, or any other less than desirable neighbour. We’ve seen this concern reflected in the private cloud solutions offered by the public cloud providers.

There are also less dramatic but no less important factors, such as operators who still maintain data centres and hosting solutions as that’s part of the products and services that they themselves offer to enterprise and government – why pay for hosting when you already have it yourselves?

One more to add into the mix, and this isn’t always the case as it can be quite geography dependent, is latency. For a lot of applications microsecond latency isn’t a prerequisite, but for others it can be important with highly time sensitive applications such as real-time/online charging. In this case you want your OCS/CCS as physically close to the network as possible.
 
So, for CSPs/DSPs who might be facing some of these concerns, what other options do they have?
They’ve got a few, Adam. Over the last couple of years, we’ve seen more and more operators looking at both fully private cloud and hybrid cloud models.
 
Could you explain a little more about both of those?
Sure – private cloud is essentially what the name implies; dedicated cloud infrastructure that is not shared with anyone else. Service providers want their responsibility for the application to end at the browser, and private cloud delivers that in a way that’s highly secure and focused on the business. It can be right-sized for the business too, so whilst it doesn’t deliver the truly elastic scaling that public cloud does, it can still be flexed up and down and that’s also reflected in a much lower TCO.

And for established operators who have a more predictable consumption profile than start-ups the cost savings can be significant. It can also be located in-country, addressing those data sovereignty issues and because it can be locked down in a way that co-lo public cloud can’t be (truly separate from any other users) it provides that extra layer of physical security that public cloud doesn’t.

In addition, for mission critical systems it is imperative that when issues arise, and they inevitably do, that the engineering team have access to all levels of the system from network all the way up to the application to analyse and identify the root cause. In a public cloud deployment this is extremely difficult as there are many players in the game and your issue is mixed in with all other customers on the same platform. In comparison private cloud provides the right level of isolation and also access to the systems to rectify any problems in a much more agile and efficient manner.
 
You said that the private cloud approach is cheaper, is that really the case?
Definitely. Public cloud can be cheaper in the short term, no doubt, but a lot of operators are looking at five-year TCOs or sometimes even longer. Once you get beyond year three the accumulated cost of public cloud goes well beyond anything you see in private cloud. The break point is typically at about 36-40 months, but a lot of our customers are signing five-year agreements or even longer.

It’s also a fixed, predictable cost. The true cost of ownership of the public cloud model is, by definition, variable, so it’s almost impossible at the time of signature to say with certainty what that total usage cost is going to be and it’s inevitably higher than the sticker numbers the vendors quote during the sales process. With the private cloud model there’s a certainty of outcome that you just can’t get in the public cloud world.
 
You also mentioned about hybrid models, could you talk a little more about what those are and how they work?
Yes, of course. Hybrid models work as you might expect. With an increasingly micro and hybrid services architecture it’s possible to deploy different applications on to different infrastructure, playing to the strengths of each.

Some applications may be more bursty in their consumption, whilst others may be holding highly confidential data; some might benefit from very low latency but there may be those whose service length is much shorter or are brought up and down as needed. This all plays very well to hybrid cloud, as well as multicloud and distributed cloud models.
 
Could you give us some hybrid cloud examples?
Yes, there’s quite a few good examples. That short lifecycle that I just mentioned, it’s not just PoCs and the like, that of course come immediately to mind, it’s also things like non-production environments that might be needed temporarily during projects for testing, training, migration staging environments, these are all common examples.

Separating out customer-facing apps from internal systems is a good way to deal with production usage that’s quite variable. Integration layers that don’t hold customer data but support a variable traffic load work well in the public cloud, whilst the customer data can be secured within a private cloud.

Another good example is placing mission critical and low latency solutions like an OCS/CCS close to the network within a private cloud, yet having those connected to mass market applications that are only available in the public cloud like generic CRMs. This combination can reap the benefits of both approaches, the important point is to look at the applications individually, the use cases they support, and choose the deployment method that works best in each situation.

There’s another private cloud scenario that’s got a lot of attention recently. CSPs/DSPs often sell their own hosting services, something we touched on above. We’ve been offering a private cloud solution with all the benefits of that “ends at the browser” responsibility, but deployed within the service provider’s own hosting capability, i.e. in their data centres. This allows them to sell more hosting capability to the supplier who then delivers their solutions back to them as a fully managed cloud service, that’s a proper win-win.
 
There would certainly seem to be advantages to looking at all the options, is this something that you’re just thinking about now or are these approaches actually being used?
This is all being used now; I can give you a couple of real world examples if I’m allowed to name names! Cable & Wireless Seychelles is a great example of that private cloud approach where we’re benefitting from their on-island hosting capability. It’s good for them from a utilisation perspective and it keeps all the data in-country, but they benefit from a fully managed solution from the infrastructure up.

The other one is Neos Networks in the UK, who deliver connectivity solutions to B2B and public sector organisations, so for them security is paramount, and the Cerillion private cloud solution (and true end-to-end private cloud) gives them the confidence they need to serve their particular customer base.
 
Andy, you’ve covered a lot of ground there, is there anything you’d like to add before we wrap-up?
It’s horses for courses, Adam. Public cloud is great, but it’s not always the right answer. The first thing we do with our customers is look at the use cases and talk through the options. Public cloud on paper and from the cloud providers sales pitch sounds like the holy grail. However, when you peel back the layers of the onion and apply it to real life mission critical systems the business case often begins to unravel.

The true model that I feel is fit for purpose is the hybrid one in that the fundamental components within the infrastructure are built on private cloud and then augmented by services delivered from the public cloud. This is the picture we have with current customers, and it works very well, gaining the best of both worlds, but without fixating on one path or the other.