A regulated financial firm needs to deploy a deep-learning risk-scoring service. Mandatory constraints include that all raw trading and customer data remain inside the company's U.S. data center to satisfy data-sovereignty rules. During quarterly options-expiration events, traffic briefly spikes by a factor of eight, so the solution must allow rapid horizontal scaling. The engineering team also requires container orchestration that supports GPU nodes, rolling updates, and automatic rollback.
Which deployment environment best satisfies all of these requirements without violating the data-residency policy?
Edge deployment that pushes the model onto individual trading terminals on the trading floor
Pure cloud deployment using a fully managed serverless inference service
Cluster deployment in the on-premises data center managed by a local Kubernetes or OpenShift cluster
Hybrid deployment that bursts raw traffic to a public cloud cluster during peak periods
A locally managed cluster provides horizontal autoscaling across many on-premises nodes, supports GPU accelerators, and permits rolling updates and rollbacks through platforms such as Kubernetes or OpenShift. Because the cluster runs entirely in the firm's own data center, no customer or trading data leaves the protected environment, so regulatory and data-sovereignty rules are not breached. A pure public-cloud approach offers elasticity but would move regulated data off-site. A hybrid model still transmits data to a public cloud during bursts, creating compliance risk. Deploying the model to individual trading terminals at the edge keeps data local but cannot elastically scale or be centrally orchestrated to meet sudden 8× load increases.
Ask Bash
Bash is our AI bot, trained to help you pass your exam. AI Generated Content may display inaccurate information, always double-check anything important.
What is container orchestration, and why is it important in this scenario?
Open an interactive chat with Bash
How does horizontal scaling help during traffic spikes?
Open an interactive chat with Bash
What are GPU nodes, and why are they necessary for deep learning?