Head In The Cloud

When it comes to choosing the best way to store and access their data, companies need to consider their needs and resources carefully. The first step is understanding howclouds are hosted.

“Cloud services are provided by server farms located in extensive hyperscale data centre facilities – often across two or three such facilities to provide redundancy,” explains Jan Hnizdo, CEO of Teraco Data Environments. “Cloud providers have ‘cloud on-ramps’, which enable end-users to access cloud services directly. These are usually located in neutral data centres with large carrier ecosystems, allowing the cloud provider to reach a large target market.” When choosing a cloud service, Hnizdo says, a key consideration is connectivity to the various cloud operators. “The connectivity needs to be as close to the cloud on-ramp as possible to reduce latency and ensure maximum uptime to the cloud operator.”

Mix’n match

You also need to decide what kind of cloud solution you want: public, private or a hybrid of the two.

“Public cloud is the likes of AWS and Microsoft Azure, with servers and software owned and operated by a third-party service provider,” explains Hnizdo. In this case, the cloud provider also manages the supporting infrastructure such as the data centre.

Private cloud services are dedicated solely to a single user. “This cloud can be physically located in an enterprise’s data centre or a third-party service provider’s facility.”

Finally, there’s the hybrid option. “Most enterprises opt to go this route, using public clouds for specific workloads and choosing to retain others in their private clouds,” Hnizdo says. “This also lands up becoming the most cost-effective solution for most enterprises.” He points out that the hybrid option is also typically the best solution for integrating legacy applications and hardware.

Scalability is another key consideration, says Hnizdo. “Compared to traditional data centre architecture, a private cloud is far more scalable, but doesn’t provide as much scalability and agility as a public cloud service. Should a business require additional resources on a private cloud, this generally is not immediately available.”

On the edge

We’re Wgenerating far more data than ever, and sending all that data to and from centralised servers or clouds hogs bandwidth and leads to latency issues. The solution may lie in edge computing.

“Edge is a distributed computing model that brings computation, data storage, and power closer to the point of action or occurrence of an event,” explains Hamilton Ratshefola, general manager of IBM Southern Africa.

He says that processing data where it is created – at the edge – reduces the amount of data sent to the cloud. “Although it may interact with a centralised cloud, edge computing doesn’t need contact with one, allowing for offline reliability.

For example, an edge computing-enabled internet of things (IoT) device can create, process, store, and act on data, even when the device is not connected to the internet. When a connection is available, relevant data is then shared within a continuous operating environment. This also allows for more immediate application of analytics and AI capabilities.”

Applications

Edge computing has the potential to transform processes and entire industries, says Ratshefola. “For example, sustainable agriculture companies are equipping crops with IoT-enabled sensors and using edge computing to monitor the growth needs and ideal harvest time for individual plants. Automotive companies are essentially making cars into edge devices, equipping them with internal and external sensors that generate data.”

Edge devices could include everything from smart thermostats and VR glasses to industrial robots and smart buildings. The challenge, says Ratshefola, lies in the management and monitoring of all these devices – especially at scale and in dynamic environments. “It also breaks down the neat physical boundaries of the cloud data centre – forcing us to think about issues of security, addressability, management, ownership and compliance.”

More importantly, it multiplies the scaling issues of cloud-based management techniques. Because edge networks, gateways and devices increase the number of compute nodes by several orders of magnitude, it must be possible to deploy, update, monitor and recover the edge compute space without human intervention, says Ratshefola. “The system must have a deep awareness of the nature, location and purpose of different devices with different capabilities and uses, and be able to use that awareness to make informed, policy-driven decisions.”