Overview
Similar to the 'Memory Wall,' the I/O wall refers to the massive latency difference between the CPU/RAM and storage devices (SSDs, HDDs) or network interfaces.
Latency Scale
If a CPU cycle were 1 second, a RAM access would take a few minutes, but an SSD access would take days, and a network request could take months.
Mitigation
- NVMe: Reducing protocol overhead for flash storage.
- Direct Memory Access (DMA): Allowing I/O devices to bypass the CPU.
- Asynchronous I/O: Allowing the CPU to do other work while waiting for I/O to complete.
Significance
In data-intensive applications (Big Data, AI), the I/O wall is often the single biggest factor limiting overall system throughput.