High availability and disaster recovery
For regulatory compliance, banks and other financial institutions often run production environments in self-built data centers (IDCs) that use Apsara Stack. This requirement places high demands on the stability and high availability (HA) of the underlying cloud-native platform, Kubernetes. Key challenges include achieving an active-active setup across two data centers and preventing communication latency with remote data centers from causing widespread service timeouts. These timeouts can affect the security of end-user funds and accounts. When a disaster occurs, it is critical to limit the impact, reduce the number of affected users, and shorten the recovery time.
Change control and assurance
For large-scale distributed systems in finance, a variety of release policies are available to support different scenarios. These policies include grouped releases, beta releases, and canary releases. These policies help smoothly transition traditional architectures, meet financial technology risk requirements, and implement container services for large-scale operations and maintenance (O&M) in finance.
You can trace release tickets to record changes to containerized applications and manage their versions.
You can release multiple applications in batches and set release dependencies between them.
You can control releases with policies that support grouping, batches, canary releases, pauses, and rollbacks.
You can scale containers elastically on demand, based on resource usage, or on a schedule.
Unlimited scalability
The finance industry is increasingly embedding financial services into daily life. This shift creates many high-traffic scenarios.
A distributed architecture makes it possible to split and scale out services to handle traffic surges when hot spots occur. However, a distributed application layer is not enough if there are physical resource shortages or data layer bottlenecks. A complete system is needed to manage all layers, from the access layer to the data layer. This system allows the entire stack to be treated as a single unit that can be scaled out and in quickly.