subscribe: Daily Newsletter


Discovery moves to hyperconvergence


Kathy Gibson reports from IDC Cloud and Data Centre Roadshow in Rosebank – Discovery is well on its way to hyperconvergence, having made the initial moves towards a software-defined storage infrastructure.

Johan Marais, virtualisation manager at Discovery Holdings, says the company began its virtualisation journey about seven years ago, with the aim of consolidating and saving data centre space.

To start off with, he explains that the virtualisation was based on Discovery’s traditional architecture consisting of fibre channel San and blade servers.

“For five years we managed to scale ad grow as we went along,” Marais says. “It became so easy it was a natural fit for developers, and we saw growth of 30% to 40% per annum.”
This increased the stress on the traditional stack, which still supported silos and made for multiple points of failure.

When the hyperconvergence buzzword became popular, Marais said there was very little customer experience in the local market.

“We had committed to change the way we work and operate, so the ask was quite big.”

Discovery also had a specific list of features that it needed, and few of the vendors at that time could fulfil this list.

The team found that VSAN v1 fitted the bill at that time. “And VMware very quickly ramped up the development of this product, and we started using some of the things we didn’t think we needed.”

What the organisation has learnt is that is necessary to stick to HCL standards, and this might mean changing the way it does things.

In addition, it was a steep learning curve for the IT staff members, learning to manage storage in a virtual environment.

The administrators also need to align themselves with the network team, because the network becomes more critical than ever.

“With the additional responsibility, comes capacity planning, and the virtualisation team had to learn more about how storage works,” Marais adds.

As an organisation, Discovery is about to embark on a data centre migration – and it has committed to migrating the entire stack within four weeks. “We don’t have worry about it because the system was designed to spread the load,” he explains.

The next move is to set up a virtual SAN cluster split across two sites. Each site will be a fault domain, offering site-level protection with zero data loss and near-instantaneous recovery with automated failover.

Discovery has committed to performing infrastructure as code. By having a software-defined storage strategy we have ticked one of the boxes to make that a reality.”