John Hall, Managing Director – Colocation, Proximity Data Centres, discusses the practicalities, pros and cons of housing hybrid cloud environments in colocation data centres.
By combining public and private clouds together, hybrid clouds can optimise available compute, connectivity, bandwidth and storage capabilities which enhances applications responsiveness, user experience and productivity. However, if the hybrid cloud architecture is overly reliant on data centres located hundreds of miles away, it can put applications and services at risk from poor response times due to latency issues while also racking up unnecessary data transit costs.
A more streamlined solution entails hosting private clouds in one or several regional Edge colocation facilities and connecting these to public cloud services hosted by service providers in centralised hyperscale data centres.
Mission critical applications are therefore securely contained within the private Edge cloud environment with only data that is non-time critical sent back to the public cloud – perhaps for further analysis or archiving. This effectively bridges the gap between users/devices in local offices or machines on factory floors, ensuring latency is greatly reduced. At the same time, data transit costs are reigned in with fewer high bandwidth circuits required for backhauling traffic to remote hyperscale data centres.
However, the success of hybrid cloud and multi-cloud implementations are highly dependent on the performance of the network between the various infrastructures. As well as the corporate WAN charges there can be additional costs for connecting to the public cloud using services such as Microsoft ExpressRoute. These solutions offer a high level of security and SLAs but there are costs, depending on the port size required.
Building the business case
Doing the preparation work for enabling hybrid cloud solutions can be challenging – one size does not fit all. Consideration must be given to how your business will manage the hybrid environment and the hardware, network and storage necessary. The hybrid cloud needs to accommodate specific business needs and must seamlessly operate across on- and off-premises platforms.
Be mindful of ‘supplier lock-in’ – ensure that your hybrid cloud implementation costs don’t spiral out of control. Have a clear understanding of data transfer costs between private and public clouds as data egress volumes can increase dramatically depending on the application. Most public cloud provider pricing models are for data egress as well as data storage – the decision on how you engineer the two parts of the solution is vital to keeping costs down.
A typical hybrid application will use the public cloud as a front-end transaction while the private cloud is used for the actual transaction processing. As an example, it is not advisable to run disk-level I/O operations across a cloud boundary as this will increase data egress charges from the public cloud provider.
Therefore, most enterprises will put regulatory reporting, planning and analytics in the data centre or private cloud as well as the database because it reduces data egress charges considerably. If the database is relatively small then it is possible to run it in parallel in the public cloud as well as in the private cloud. But don’t forget that general report generation and analytics can generate more database usage than transactions.
So, it is important to consider all the data that a company uses when you are looking at hybrid cloud implementations. You will be penalised by high level egress costs if you pull lots of data out of the public cloud rather than generate volume into the cloud.
A growing area for hybrid implementations is AI where enterprises are building AI applications for hosting at regional Edge data centres – being closer to the source of the data – and the AI tools built by the public cloud providers for prototyping and achieving scale. This approach helps enterprises start with small projects in the cloud and grow into a larger deployment. Again, a full understanding of the data charges must be understood as part of the process.
As well as the cost of data egress and storage in the cloud, some applications will be highly reliant on low latency. If this is the case, then it is best to host the application in the Edge data centre in a private cloud infrastructure, close to where the data is generated. This will give best performance. Some of the heavier duty processing which is less time-sensitive can be performed in the public cloud.
When allocating applications between an Edge data centre and a hyperscale facility consider how long it will take to migrate all the applications to the new infrastructure and whether any on-premises legacy IT infrastructure require accommodation. Furthermore, it is important to consider what software will be required for managing all environments within a hybrid implementation and whether there will be a direct cloud gateway available for direct connection into public cloud infrastructure.
Finally, carefully evaluate if your IT team has sufficient expertise in hybrid cloud functionalities, management, integration and administration. Equally, does your prospective Edge data centre provider have suitably qualified engineers on site to assist with configuration? The level of on-site engineering competence available is important, especially for configuring and interconnecting complex Edge cloud environments. These may involve public, private and perhaps on-premise legacy applications. With this, the flexibility to carry out pre-production testing in the data centre will be a bonus to ensure everything works prior to launching.
CIOs and IT professionals should therefore take a holistic view when considering how regional colocation data centres can get their data and applications closer to users and customers. Aside from the connectivity options, latency and cost factors, it is wise to check your Edge data centre’s hybrid IT engineering skills.Click below to share this article