Cloud bursting takes advantage of the public cloud, which is available as required and offers the option to pay to the proportion in which you use the system. The main benefit is to be able to reduce the workload locally by transitioning back and forth to and from the cloud. The main source of doubts arises from knowing exactly when this is needed, and to execute it somewhat automatically.
The idea is relatively simple, yet to setup and run things smoothly is not. The first thing to look for is the compatibility of the workloads, mainly referring to the adaptability of the application in the cloud.
Shared clouds have a host of standard services and resources, and the application must work well under those conditions. If the workload was designed with running in the cloud in mind, chances are that it’ll work well in the cloud and transitioning back to internal dependencies when needed will be straightforward.
Cloud bursting limitations
Sometimes, even if the workload is designed for the cloud, bursting can bring issues related to networking, and others. Generally speaking, the location in where the data is kept and the protocols used for exchanging it can be picky, and can cause incompatibility to some degree. Other problems are latency, lagging, freezing etc.
Depending on the load, it can take a huge amount of time to upload it to the cloud, and downloading it back can have extra costs.
Cloud bursting has three main characteristics, and each one of them will require different ways to go about.
1. Distributed load balancing
To start with, this option uses load-balancers, and the workload stays in between the data center and the cloud service. The business works with the load locally, allows for resources like virtual machines, data storage in the cloud and sends the workload to them.
To finish it up, administrators monitor the workload in order to keep it within the pre-established limits. Once those limits are reached, the traffic is redirected to the cloud. Once the load is back to normal levels, it is then brought back to local servers, and the cloud services are not needed anymore.
This method implies that both local servers and the cloud are in use, although not always simultaneously, and the load balancing defines when and where each one will be needed most.
To achieve this, the business must be prepared for cloud deployment beforehand, which can impact costs. Also, as a general rule, the cloud service parameters are fixed, so you cannot increase or decrease its capacity. This can be a limiting factor when operating with cloud burst methods.
2. Manual bursting
Another common method is to manually switch from local to cloud services based on reports and notifications from the load balancer. This gives business owners the ability to provision cloud services that have the exact size needed for any given moment, and after the usage is done, to simply remove the cloud service from the equation.
However, having the human component dealing with these situations can be pretty complicated. Mistakes and delays can jeopardize and hinder whole operations when data traffic gets stuck. Also, overseeing and delaying the removal of the cloud services will have an added cost. This is why the manual approach is not recommended.
The manual approach is useful for understanding needs and testing. On the other hand, admins on the cloud prefer to use coded applications to automate the process, which is theory is more reliable.
3. Automated bursting
This is the most recommended method, since it requires little to no interaction with human operators. This method relies on software and other services to understand the needs for resources at any given time and provision the services automatically when the time comes. It also fully deals with the removal of the services when they’re not needed anymore.
Applications that automate the process do so using the cloud service APIs, which allows for customization of the service parameters almost in real time, while eliminating the possibility of human mistakes in the manual methods.
Making cloud bursting work
To be able to successfully deploy and operate cloud bursting systems, the network between the local storage and the cloud must be well established and secure, meaning that a decent data transfer rate with the least latency is desired.
Additional security measures such as using VPNs will contribute to adding security, and services like AZURE and AWS direct connect can provide more bandwidth without traffic slowdowns. Adding these extra services comes at the price of having to test each one of them separately and in conjunction with the whole ecosystem, to ensure everything will work as it should.