Skip to main content

Distributed cloud offers the best of both worlds

(Image credit: Shutterstock / Blackboard)

The next big thing in cloud computing offers numerous advantages to the enterprise IT user, says Neo4j’s Jim Webber.

Cloud users benefit from shifting the responsibility of running hardware and software infrastructure to providers. They can leverage the economics of cloud elasticity and benefit from a fast pace of innovation from the cloud vendors.

Yet for a variety of reasons such as data security, cost, and infrastructure, contemporary CIOs in enterprises have tended to use a combination of their own data centers, private cloud, and (multiple) public cloud provision. While the debates of the cost-effectiveness of a multi-cloud strategy continue, the notion of ‘distributed cloud’ has recently entered the lexicon, stretching the discussion along an interesting new axis.

Distributed cloud is a mix of on-premise and cloud computing, but where the abstraction layer for the on-premise part of the system matches that of the cloud. So the abstraction your data center provides looks very similar to your cloud vendor’s abstraction to a software developer.

This is an interesting notion: that some workloads that can be more cost-effectively run on your own servers can be run that way, and other cloud-friendly workloads run on the cloud. But because both workloads target the same APIs, there is a degree of flexibility over time in exactly where workloads will be run. In a sense, you can move the dial between your own infrastructure and the cloud to suit your current needs. 

For the cloud providers like Google, Amazon AWS, Microsoft, Alibaba, etc there is an interesting investment at play. By releasing (some of) the code that runs their clouds, they are improving the quality of non-cloud data centers so that they run more like efficient cloud data centers. At first glance, this is an own-goal. But dig deeper and you’ll see this is a gateway drug to give CIOs the freedom to move systems with relative ease into the cloud and rationalize their cloud versus on-premise spend.

The ISV advantage

From an ISV point of view, you could be forgiven for ignoring the relevance of distributed cloud. But that would be naive, because not all cloud services can be neatly packaged and moved from the cloud providers into your data center, regardless of API uniformity. For example, your data center does not have the specialized hardware like GPS atomic clocks or FPGAs to support advanced transaction processing and data analysis - that remains in the cloud, even if you want it closer to your other systems for security, compliance, or latency reasons.

It’s clear that cloud providers have some amazing innovations to offer. But they are not the sole source of useful information technology. Most ISVs (including the one I work for) have offerings for your on-premise needs as well as existing in the marketplaces of all the biggest public clouds.

This means you can choose an ISV system such as a database and have complete freedom to deploy to on-prem or cloud-provisioned hardware, or indeed simply to consume the existing SaaS version of that software from your chosen cloud. Want to move from on-prem to cloud or back? Simple. Want to move clouds? Also straightforward.

This means you can use the application on a distributed cloud but not be tied into any particular public cloud provider. You can decide between going with an on-premise-only solution or a fully managed cloud solution.

Consider how, for example, you might find great value in Amazon Neptune, but the organization says for cost or uptime reasons it needs to move to Google. In this situation, an Amazon Neptune proprietary solution leaves the CIO exposed, as AWS Neptune is not an option on Google. But a distributed cloud-aware ISV can easily run on either Google or any other public cloud. In this new distributed cloud world, not being coupled to any cloud provider is a definite liberation.

Running ISV apps in a distributed cloud also allows us to manage latency. Picture a core business application partly run by Google in California, 5,000 miles away from your London data center. This can be a mission-critical issue if the data is latency-sensitive. A financial trader can make a local call in London that takes an incredibly small fraction of a second to transmit to a server. Still, if all of the data has to travel to the US and come back, that 10,000-mile round trip takes a surprisingly large number of milliseconds. If I'm running that part of my system on the local on-prem Google cloud, that same transaction can be completed in sub-milliseconds, which is a huge win.

Even in a more conventional business context, every millisecond matters now, however. The speed it takes to render a web page or a screen on an app is critical. It takes half a second to a second before users start to get twitchy. There’s an acceptance criteria of approximately 200 milliseconds, and probably a lot less in the case of Gen Z. As a result, you don't want any app to wait while the database query travels from London to California and back again. The query needs to be answered as near to the user as possible.

There is a similar distributed cloud business advantage in the case of addressing regulatory requirements such as GDPR. In the public cloud, you never know where your data resides at any given moment. 

If you have your own data center, you know the data is physically there. One of the benefits of a distributive cloud is that it gives you the ability to host some data in the cloud and some data physically in your servers. The data is still run on the same cloud operating system, but the pieces of data are physically located in the building next door. When a GDPR inspector comes and wants to know where customer X’s data is, you can say it's in this building, on that computer, and in that rack.

Distributed cloud customers 

The fact that data can have a physical location is a great message for systems that process sensitive information. If there is any concern that the data would be open to hacking if stored in the public cloud, the four walls and 24x7 security offered by a secure data center might be reassuring.

Multiple distributed cloud use cases are opening up, from healthcare, to media to financial services. Any type of customer likely to have a large set of proprietary platforms would benefit.

All in all, the move to distributed cloud will give the market the clear computing and compliance benefits associated with proximity while also accessing the flexibility and cost benefits of a distributed arrangement.

It also offers the ideal on-ramp to public cloud computing. It would make any business run its data center so that it would be straightforward to package up and pass across to a public cloud provider when it makes business sense. 

Buying into a distributed cloud is a sound way of keeping your options open about where you should move your workloads in the future, while working with an ISV gives you extra flexibility and benefit. And now is definitely the time to start seeing how.

Jim Webber, Chief scientist, Neo4j

The author is Chief Scientist at Neo4j, the world’s leading graph database company, and co-author of Graph Database for Dummies (Wiley).