The End-of-Row (EoR) or Middle-of-Row (MoR) architecture consolidates network switching for a row of racks into a single, large chassis-based switch. A primary disadvantage is the cabling infrastructure; every server requires a dedicated, long cable run to the central switch. This results in extensive, complex, and difficult-to-manage cabling bundles that can impede airflow and increase costs.

Additionally, the EoR/MoR model is often considered less scalable than alternatives like Top-of-Rack (ToR). It follows a "scale-up" model, requiring a significant upfront investment in a large chassis. Once the port capacity of this chassis is reached, expansion becomes a major, disruptive, and expensive project, unlike the incremental "scale-out" approach of ToR designs.

A. More equipment: This is incorrect. The EoR/MoR design consolidates switching into fewer, larger devices compared to the Top-of-Rack (ToR) model, which uses many smaller switches.

B. Higher latency for rack-to-rack traffic within the row: This is incorrect. EoR/MoR provides lower latency for this specific traffic path, as communication between servers in different racks within the same row is a single hop through the central switch.

1. Aruba Data Center Network Design for Aruba CX Switching

Version 1.0 (March 2021)

Page 11

Table 1 "Comparison of ToR

MoR

and EoR designs": This table explicitly details the disadvantages of EoR/MoR. For "Cabling

" it states

"Requires structured cabling system. Can result in a large amount of cabling." This directly supports option C. For "Scalability

" it notes

"Requires large upfront investment. Can result in unused ports

" which supports the "less scalable" aspect of option D.

2. HPE FlexFabric Cookbook

Version 2.11 (December 2016)

Page 21

Section "End-of-Row": The document describes EoR

stating

"The main disadvantage of this model is the large amount of cabling required between the server racks and the EoR switches." This validates option C. It also implies scalability issues by contrasting it with the flexibility of other models.