Cloud versus on-premises

“Any organisation is looking for the capability that an application delivers. So we have customers which have migrated their entire data warehouses and gained power and versatility, scale efficiencies and better cost economics. In fact, with these new platforms scale is simply now out of consideration. Your prototype can become the production system because the only change is the amount of data. It has also become massively simple — here’s the data, this is the query. The work is done behind the scenes, as it were, and that is a huge benefit in removing friction from the deployment of systems in the organisation.”

Server architecture
There are many workloads, Matt McNeill acknowledges, which are still based on server architecture but can be fully virtualised in the cloud. “In a sense I’m changing where I do my computing but fundamentally not how. What we are calling the ‘third generation’ of cloud computing is changing the ‘how’, with a corresponding need to re-architect many of the applications. On the other hand, there are applications that are natively third generation. Machine learning is a good area to look at in this context. Because the data is now wieldy, the machine learning algorithms are less of a technical challenge.”

He emphasises that this new generation of computing is inherently open. “I think that ‘lock in’ game is a disappearing approach to the market that we will no longer see in the future of cloud. Most resources will be community assets that you pay to use. But the effects across every sector will be of disruptive innovation using this new generation or level of cloud technology. In part that has to do with the deployment of capital. Historically, big enterprises kept their market lead by investing where the new or the small could not afford to. The downside was the three to five-year depreciation horizon that resulted. Today, cloud-native disruption is not capital intensive and so the challengers are inherently more agile and competitive.”

 

advertisement

 

First cloud
Innovation is something that Lorcan Cunningham understands. As CTO of Savenet Solutions, he happily points out that they launched a shared online environment for disaster recovery to save costs a matter of some months before Amazon launched its Web Services in 2007. Both were essentially what we later came to call ‘cloud’.

“Our service was a later version of what was previously called an Application Service Provider and more recently SaaS, built on VMware. In our early days we went after back-up and DR and archiving as a hybrid solution, part on-premises and copied in the cloud. The aim was to save the client’s money with solutions that were cost-effective. We still believe that on-premises should be your core applications and servers that you are looking after directly. Then the more peripheral stuff can go to the cloud and reduce your capital investment requirements.”

But it definitely depends on the organisation, Cunningham says. “We recently talked to a €200 million company that had just taken its email back in-house from Office365. For them, it’s a critical application and they had an outage followed by an indifferent experience when they looked for service help. But for many other clients, email is not that critical and the occasional break is not too significant. So the lesson applies to all core business applications.”

Data protection strategy
Similarly, he adds, any move to a cloud service means that it is essential to have your data protection strategy thought through and in place, plus your exit strategy. “You may be simply changing service provider at the end of a contract or invoking your termination rights because you are unhappy. But you have to be certain that the changeover will work seamlessly — before you sign up in the first place. We have clients, for example, that are using Azure or other public cloud services but replicating back into our cloud, so combining their data protection and exit strategies. In other cases, we are backing-up client data to their own on-premises systems.”

“In our experience, core applications like ERP, specialist legacy applications and anything that is affected by latency — or just a network break — are still on-premises for most organisations. We have quite a few pharmaceutical clients, for example, and their site systems are critical and on-premises. Email and some other functions, on the other hand, are beginning to head to the cloud. In some ways, and not just in smaller organisations, it depends on where they are in the hardware refresh cycle. That is a natural time to review systems and consider alternatives, especially where there might be cost savings.”