The cloud comes in a range of shapes and forms. There is the private cloud, which gives you flexibility, but is costly to maintain and configure and has limited scalability.
Then there is public cloud, which gives you immense amounts of flexibility and scale, but sits outside of the firewall.
Many IT directors feel it’s not secure. Lastly, there’s hybrid cloud, which has all the advantages of the public cloud and the private cloud and none of the disadvantages.
Unfortunately, nobody seems to have cracked the hybrid cloud.
In practice, what you get is a “hybrid cloud” that consists of new applications built in the public cloud, and old legacy apps sitting in the private cloud.
What you need to make hybrid cloud really hybrid is cloud bursting.
A hybrid cloud that incorporates cloud bursting will allow you to take a workload and spin it up on the private cloud, but if the workload needs more resources, it can be seamlessly moved out to the public cloud and easily work with data sources, no matter where they are – in the cloud or on-premise.
Cloud bursting is therefore a great way for businesses to handle peaky demand patterns, such as e-commerce providers with big peaks in sales at Christmas, or news and sports websites with steady demand that spikes when something big happens, for example the World Cup or the Olympics.
Cloud bursting can also be a useful tool for businesses that need to carry out analysis on large datasets, and for traditional applications such as month-end accounting runs where the demand is predictable but requires servers and storage to sit idle most of the time.
However, not all peaky demand applications are suitable for cloud bursting. Cloud bursting works best for applications that don’t depend on a complex application delivery infrastructure or integration with other applications, components and systems that are housed on-premise.
Read more about hybrid clouds
The appeal of cloud bursting is easy to see, but the level of complexity involved negates its benefits – at least for now.
Cloud bursting helps organizations use the public cloud to manage sudden spikes in demand. But what challenges or issues might it introduce.
Cloud bursting barriers
There are many reasons why cloud bursting is most suited to non-complex applications and why many applications remain either public or private, but not a mixture of both. The biggest barrier is compatibility between on-premise and cloud platforms.
Microsoft Azure, Amazon Web Services (AWS), Google and VMware are not interchangeable and probably never will be, so any application that bursts needs to have some translation element to enable the app to talk to different platforms. One of the reasons Azure cloud platform is doing so well in the market is the company’s commitment to making Azure compatible with on-premise systems and the platform’s ability to run Linux and Windows workloads. This makes it easier to create truly hybrid applications that can burst.
The other major issues with cloud bursting are around bandwidth and latency. Application workloads tend to be relatively small and can be moved quickly, but the associated data is not usually so small and any attempt to move it across the internet in one go will take time.
The upshot of this is that your network must be able to cope with, and prioritise, the extra traffic caused by the increased inbound and outbound data, and your cloud provider must be near enough not to cause problems with latency.
Once again, Microsoft’s decision to put two new datacentres for Azure in the UK means it is becoming one of the best bets for cloud bursting. Note that any extra costs for bandwidth from your cloud provider need to be factored into the equation, otherwise the cost benefits of cloud bursting may be considerably less.
Licences are also an issue. If you are relying on commercially licensed software to power your application, then to get cloud bursting working, you need to be able to extend that licence to the cloud and/or change the contract to allow short-term extensions to an existing licence to support additional use.
Also, there is a lack of management tools that can provide a single overview of the multiple environments and, more importantly, decide when it is appropriate to burst, and that can initiate the move.
Lastly, there are the inevitable security issues that are created by moving data to and from the public cloud and delivering an application on a shared environment.
Solving the cloud bursting problem can – and has – been approached in many different ways, including software-defined storage (SDS) such as Avere’s FXT Edge Filers; products that solve the platform incompatibility problems, such as Cloudian’s HyperStore; new cloud bursting tools, such those from US startup Velostrata; management systems, such as those from CA and RightScale that can trigger and manage cloud bursting; and systems that treat cloud bursting as a load balancing problem.
Application delivery controllers
Application delivery controllers (ADCs) are next-generation load balancers that are proving to be fundamental building blocks for advanced application and network platforms. They enable the flexible scaling of resources as demand rises and/or falls and offload work from the servers themselves. They also provide a number of other services that are essential to the effective operation of on-demand applications, including:
- Network traffic compression – to speed up transmission.
- Data caching – to make sure regularly requested data is
- readily available.
- Network connection multiplexing – making effective use of multiple network connections.
- Network traffic shaping – a way of reducing latency by prioritising the transmission of workload packets and ensuring quality of service (QoS).
- Application layer security – the inclusion of web application firewall (WAF) capabilities to protect on-demand applications from outside attack.
- Secure sockets layer (SSL) management – acting as the landing point for encrypted traffic and managing the decryption and rules for ongoing transmission.
- Content switching – routing requests to different web services depending on criteria such as the language settings of a web browser or the type of device the request is coming from.
- Server health monitoring – ensuring servers are functioning as expected and serving up data and results that are fit
- for transmission.
ADC products have been on the market for some time and are available from a wide range of suppliers, including A10 Networks, Brocade, Citrix, F5 Networks, Kemp Technologies and Radware.
To solve the platform incompatibility problem, US business Cloudian has created 100% S3-compatible storage that allows users to seamlessly burst data from an on-premise private cloud, using standard server hardware, into an Amazon S3 public cloud storage bucket and back using the standard Amazon S3 API on-premise, private and in the cloud.
To help maintain data and workloads across the different clouds (public, private) and on-premise, CA has produced a unified infrastructure management service, which monitors a number of on-premise services (Nutanix, OpenStack, VMware) and public clouds, including AWS. It then displays the data on a single desktop, showing cost and performance and automatically triggering cloud bursting to additional cloud-based resources based on predefined constraints.
Avere’s FXT Edge Filers concentrate on storage, particularly network-attached storage (NAS). The system allows users to burst data and computation to Amazon’s EC2, while cloud compute and on-premise servers can access data in private, on-premise and public cloud with low latency and security similar to on-premise.
US startup Velostrata has a new approach to cloud bursting. It decouples the compute part from the storage, promising no changes to the apps. The compute portion of the app is streamed to the cloud and is up and running in 20 minutes, with the data migrated in the background and the data interconnect between on-premise and cloud optimised to keep latency to a minimum.
As yet, there is no single product that allows bursting of all the main computing building blocks – network, storage and compute. That requires software-defined infrastructure (SDI), where the application runs on the most appropriate systems for the current workload, the data is stored where it is needed, and the network directs traffic to the right place with the right amount of bandwidth. Unfortunately, SDI just isn’t there yet, and until it is, IT directors will have to rely on cloud bursting as a stopgap.
Marcus Austin is a service director at Quocirca.