Episode Details

Multi‑cloud network tax: in this episode of M365.fm, Mirko Peters explains why your “cloud‑agnostic” architecture feels brilliant on slides but brutal on invoices—especially once Azure, AWS, and GCP start charging you for every cross‑border packet like three different toll roads billing the same car. He opens with the religion of multi‑cloud: boards demanding vendor neutrality, architects drawing logo‑diagrams full of arrows, and nobody admitting that every extra provider multiplies IAM complexity, monitoring tools, incident dashboards, and, worst of all, egress fees.

Mirko shows how this tax hides in your diagrams. An analytics pipeline that ingests in Azure, trains models in AWS, and archives in GCP looks sophisticated enough for investor decks, but underneath it sits a mesh of ExpressRoute, Direct Connect, and Cloud Interconnect circuits stitched through carrier‑neutral PoPs. Each hop adds latency and cost as your data leaves one sovereign network, pays egress charges, traverses colocation fiber, and re‑enters another cloud that happily advertises “free ingress” while the other two quietly invoice you.

He then walks through the actual handshake when clouds talk. Azure VNets, AWS VPCs, and GCP VPCs are separate countries with different currencies and customs—VNets, Direct Connect gateways, virtual WAN hubs, SD‑WAN overlays, transit centers. To “just connect” them, you layer site‑to‑site VPNs, private interconnects, redundant circuits, and complex DNS forwarding, turning each cross‑cloud request into a miniature import‑export operation measured in milliseconds and line items. The result is a networking matryoshka doll where every new hub, gateway, and monitoring agent adds both failure vectors and billable surfaces.

The episode does not argue against multi‑cloud entirely; it argues against doing it everywhere. Mirko explains where multi‑cloud truly earns its keep—targeted use of a second provider for specific strengths, regulatory separation, or negotiating power on a narrow set of workloads—versus where it becomes superstition dressed up as strategy. He gives you language to distinguish redundancy within one cloud (cheap, inside‑backbone high availability) from cross‑cloud replication (expensive, latency‑heavy, and often redundant in name only).

Throughout, you get practical steps to stop paying the multi‑cloud network tax blindly. Mirko suggests tracing a real packet’s journey between clouds, mapping each hop to hard costs (ports, circuits, egress) and latency, then using that map to simplify: collapsing unnecessary interconnects, centralizing DNS, consolidating observability, and moving some workloads fully into a single provider where the physics—and the pricing—favor you. The episode arms you with arguments for CFOs and architecture boards who need to hear that “best of breed” without cost discipline is just best of bleed.

WHAT YOU WILL LEARN

• Where the hidden multi‑cloud network tax shows up in your diagrams, latency, and invoices.
• How ExpressRoute, Direct Connect, and Cloud Interconnect actually move packets between clouds.
• Why cross‑cloud redundancy is far more expensive than intra‑cloud high availability.
• When multi‑cloud is justified (and when it is just expensive superstition).

  Listen Now