Reliability and scalability

Reskyt designs and operates the platform with ecommerce traffic peaks in mind, such as Black Friday or sale campaigns, where load is concentrated into a few hours.

Where the load sits in a Reskyt app

The split is not intuitive, and it is worth understanding before sizing anything.

Most browsing traffic goes to your ecommerce. The app displays your store: every product page, every search and every checkout step are requests to your platform, not to Reskyt.

Reskyt concentrates load at specific moments. App startup, configuration loading and, above all, bulk notification sends.

A push campaign generates two chained peaks: the send itself and, shortly after, the users entering your store at the same time.

Caution

Before a large campaign, it is worth sizing the ecommerce side as well, since that is what receives the resulting traffic, and giving advance notice to whoever operates that platform.

How the platform is built

The infrastructure is deployed on AWS, with the primary region in eu-west-1 (Ireland), and is managed through Infrastructure as Code.

Redundancy across zones. The architecture uses redundancy across Availability Zones to reduce the dependency on a single zone and improve fault tolerance.

Managed services. Reskyt relies on AWS managed services where applicable, including ECS for running containers, RDS for databases and S3 for storage.

Caching and CDN. There are caching mechanisms, and a CDN is used (Amazon CloudFront or Cloudflare depending on the architecture in each case) to serve content closer to the user and reduce load on origin services.

Asynchronous processing. Jobs that do not need an immediate response are handled through queues. Notification sending is the clearest case: the push API accepts the request and delivers asynchronously.

Monitoring. The platform is monitored with Amazon CloudWatch and New Relic. See monitoring.

What happens if the platform degrades

The Reskyt services involved at runtime are notification sending and app configuration. The architecture uses caching, CDN and decoupling between components to improve the resilience of the service.

The specific impact of an incident depends on the affected component. The commitments applicable to each customer depend on the corresponding contract.

In the opposite direction, the availability of your store conditions that of the app: if your ecommerce does not respond, the app has no content to display. See Reskyt and your ecommerce.

See also continuity and recovery.