Fault tolerance: the questions you should ask your data centre

Unlocking the answers

When considering where to store your data, most businesses will not question the resilience and speed of the network for transferring data from the office to the data centre.

Without the appropriate network the potential loss of data is huge if there is a disaster. So what should you be asking your data centre before you are locked in?

TechRadar Pro spoke to David Barker, technical director at 4D Data Centres, to discover the answers.

TechRadar Pro: How much attention should data centres pay to fault tolerance?

David Barker: This is a key question for any data centre. You need to ensure that your supplier does not have any single point of failure and that it can at least withstand one fault without bringing the network down.

A second consideration to explore with the potential provider is that after the first fault, does it have a strategy in place to actively work on the fault while affecting as few clients as possible? And in the event of a second failure how does it ensure its system can 'fail gracefully'?

TRP: Where does Software-defined Networking (SDN) come into it?

DB: SDN eliminates the complex and static nature of legacy-distributed network architectures through the use of a standards-based software abstraction between the network control plane and underlying data forwarding plane.

By ensuring the control plane and management layers are separated from the actual hardware they aren't affected by any single fault. The data centre engineers should be able to analyse all aspects of the network (from capacity on links to current configurations), while being aware of your critical files.

This way the network can fail in a way that protects the connectivity to your data so back up continues and latest versions are saved and stored.

Check that your potential provider has SDN in place and that it has the relevant staff trained on the system to ensure effective monitoring and maintenance.

TRP: What about convergence?

DB: Convergence is another feature of a well-designed network. Legacy infrastructure generally has two networks: one for application traffic between servers and users, and one over which applications access files on storage devices.

Both of these use different technologies, with each optimised for its role. But this comes at a very high cost in dedicated components, incompatible interconnection technologies and replicating management for each of the networks.

A converged network can perform both roles of a legacy network but uses a common technology (lossless Ethernet) that brings many economies of scale and is far easier to manage.

In a converged network, a port is a port; you, or the software with SDN, decide how you want to use it at the time you plug a server or storage device into it. This system has enough intelligence to separate the two kinds of traffic to ensure that each gets the appropriate priority and bandwidth.

Ask questions about whether the data centre network is converged and how this can be customised to ensure your critical files are always prioritised, while files of less importance stored when there is enough bandwidth to accommodate the backup and transfer of the files.

While technology for data storage is fairly well developed, the underlying network that it relies on has been slow to catch-up. Make sure you ask the right questions about your provider's network to ensure it can cope with the changing demands of your data.

ABOUT THE AUTHOR

Editor, TechRadar Pro

Désiré (Twitter, Google+) has been musing and writing about technology since 1997. Following an eight-year stint at ITProPortal.com where he discovered the joys of global techfests, developing an uncanny attraction for anything silicon, Désiré now heads up TechRadar Pro.