Fallacy 6-8: Administration, Cost, and Homogeneity
You'll learn how organizational boundaries, operational expenses, and platform diversity create hidden complexity in distributed systems.
The Remaining Fallacies
After topology changes (Fallacy 5), three final misconceptions round out the classic fallacies of distributed computing. While less technical, they're just as consequential:
Fallacy 6: There Is One Administrator
The assumption: Someone has unified control over the entire distributed system.
Reality: Different teams, departments, companies, or cloud providers manage different parts. When your service depends on a third-party API, you can't simply "restart it" during an outage. Coordination across administrative boundaries requires formal processes, SLAs, and escalation paths—not Slack messages.
Fallacy 7: Transport Cost Is Zero
The assumption: Moving data across the network is free.
Reality: Network infrastructure costs money—bandwidth fees, cloud egress charges, hardware, and operational overhead. Serialization/deserialization consumes CPU cycles. A chatty microservice design might work locally but become prohibitively expensive at scale, especially across regions or cloud providers.
Fallacy 8: The Network Is Homogeneous
The assumption: All network segments use compatible technologies and protocols.
Reality: You'll encounter varied hardware (routers, switches, NICs), protocols (HTTP/1.1, HTTP/2, gRPC), data formats (JSON, Protocol Buffers, XML), and performance characteristics. A service optimized for low-latency datacenter networking might struggle over WAN connections or mobile networks.
Practical Implications
These fallacies mean you must design for:
- Multiple stakeholders with different priorities and response times
- Cost-aware architectures that minimize unnecessary data movement
- Protocol negotiation and format flexibility across heterogeneous environments
Key Takeaway: Distributed systems span organizational, financial, and technological boundaries—effective design requires acknowledging multiple administrators, real transport costs, and platform heterogeneity from the start.