We are currently implementing a multi-cloud strategy, running our front-end on AWS and our heavy-duty database workloads on Azure. However, we are seeing significant latency spikes during cross-cloud data transfers. We’ve tried standard VPNs, but the performance is still laggy. Does anyone have experience setting up dedicated interconnects like AWS Direct Connect or Azure ExpressRoute to solve this? I need a way to ensure sub-10ms latency for our real-time synchronization.
3 answers
In my experience at a fintech firm, relying on a standard internet-based VPN for cross-cloud traffic is a recipe for disaster. We eventually moved to a Megaport-style software-defined network (SDN) to bridge the gap. By using a cloud exchange provider, you can bypass the public internet entirely. This reduced our inter-cloud latency from 80ms down to about 6ms. You also need to ensure your regions are geographically adjacent—for example, use AWS US-East-1 and Azure East US—otherwise, the physical distance will always be a bottleneck regardless of your interconnect.
Are you using a Service Mesh like Istio to manage this traffic, or are you handling the routing manually through your application code?
You should look into 'Private Link' services; they allow you to consume services across clouds as if they were on your local private network.
Jason makes a great point. Private Link combined with a dedicated interconnect is the gold standard for security and speed in enterprise-grade multi-cloud designs.
Kevin, we’re currently handling it manually, but the overhead is becoming a nightmare. I’ve looked into Istio for multi-cluster management, but I’m worried about the sidecar latency adding even more delay to our Azure DB calls. Do you think a "sidecar-less" approach like Ambient Mesh would be more efficient for a hybrid setup like ours where every millisecond counts?