Holiday season peak loads : High Stakes, no margin for errorSudhanwa Rajurkar
For any business with an online business model, peak periods coincide with more business, and hugely increased loads.
For a low-margin business like retail, these peak period opportunities are critical to book profits for an entire year. The stakes are very high and there is no room for error.
Sudden surges in traffic, will bring the service platform to its knees if not handled properly. Delayed customer responses lead to lost sales and affect the brand negatively.
In addition, unwanted surges like DOS attacks can have similar crippling effects leading to loss of sales opportunities and a tarnished brand.
A smooth and seamless digital customer experience is crucial to retain customers and make them come back again. You certainly don’t want to be one of those negatively affected by such wanted or unwanted surges. It requires careful planning and implementation to be on the right side.
One of the CoreView customers, faced a similar challenge, during their growth period. Customer Acquisition was happening at a feverish pace, adding extra load, every day. Every user action accessing the product would mean several API requests to their customer platform.
Festive and Holiday seasons would mean sudden traffic surges. Systems’ ability to handle peak-time traffic and planned/unplanned surges were supercritical. With 10 million customers, even 1% unhappy customers can mean a hundred thousand customer calls to handle!!
CoreView had ingrained these requirements right up front in the entire solution from architecture-design-development-testing-Infrastructure.
‘Add More Infrastructure’ can be a simple strategy to scale up; but this allows inefficiencies in individual modules to be overlooked, dramatically increasing the deployment costs.
CoreView architects came up with an architecture that scaled vertically, before scaling horizontally. Every module was profiled and tuned for maximum performance as a collective system. Throttling Knobs, switches, alerts & notifications were provided for fine-grain control over vertical performance.
This enabled us to do a better capacity planning of the infra for current as well projected load.
Extreme Automation was used to simulate projected loads, sustained load conditions, sudden surges. This helped us to bake the system even beyond projected loads for the coming few years. Automated Monitoring, Capturing, and Analyzing of application counters to verify resource utilization (CPU, Memory, IO, Caches, handles, etc.), enabled optimizing individual modules using performance profiling.
Extreme automation, benchmarking, and controlled rollouts allowed us to validate and tweak our understanding of average loads, seasonal loads, peak loads, and This, in turn, allowed us to firm up our capacity planning w.r.t infrastructure.
The production systems are deployed, scaled out for average and peak loads. The system operates at ~ 50% of its benchmarked capacity. Spare capacity made is available on-demand for extreme traffic surges.
This combination of purposefully architected, crafted, deployed data services enabled us to absorb all wanted, unwanted, sudden surges. In the last five years, while in production, the platform has successfully handled an average load of eight billion requests per day, along with all seasonal, traffic surges without a flutter in the customer experience. The platform has also sustained a few DOS attacks without a lot of damage.
At CoreView, we thrive on such challenges. We have proven time and again, that we can successfully handle challenges like Traffic Surges during peak seasons.
Do you have a challenge, that you can throw at us?