Cisco Cisco IOS Software Release 12.0(13)S7

This document explains the basics of the buffer recarving functionality, buffer carves that are triggered 
upon changing MTUs, the behavioral change to carve-level 1 functionality, the addition of the new 
command 

 to enable or disable packet buffer recarve optimization, and behavioral 

change to the command “no bfrp carve-level”.

Each line card (LC) on a Cisco 12000 series Internet router has two types of memory:

Route or processor memory (dynamic RAM [DRAM]): This memory enables mainly the onboard 
processor to run Cisco IOS software and store network routing tables (Forwarding Information Base 
[FIB] adjacency).

Packet memory (synchronous dynamic RAM [SDRAM]): Line card packet memory temporarily 
stores data packets awaiting switching decisions by the line card processor.

This document discusses mainly the buffer carving algorithm, the addition of a new CLI command to 
control buffer optimization, a behavioral change to carve-level 1 functionality and a behavioral change 
to the command no bfrp carve-level. Buffer carving mechanism is related with the packet memory 
(SDRAM), which is divided into two banks: ToFab and FrFab (towards the Fabric and from the Fabric). 
The ToFab memory is used for packets that come in one of the interfaces on the LC and make their way 
to the fabric, whereas the FrFab memory is used for packets that are going out of an interface on the LC 
from the fabric.

The Tofab and Frfab queues are the most important concepts to understand in order to efficiently 
troubleshoot ignored packets in the Cisco 12000 Series Internet Router.

ToFab (toward the Fabric) and "Rx" (received by the router) are two different names for the same thing, 
as are "FrFab" (from the Fabric) and "Tx" (transmitted by the router). For example, the ToFab Buffer 
Management ASIC (BMA) is also referred to as the RxBMA. This document uses the ToFab or FrFab 
convention, but you may see the Rx or TX nomenclature used elsewhere.

Cisco IOS software implements a buffer-carving algorithm that divides SDRAM into various buffer 
sizes. The Cisco 12000 series Internet router connects to several different media types, with different 
MTUs. A mechanism was needed to make the most efficient use of packet memory buffers. The buffer 
carving algorithm was devised after studying the makeup of Internet traffic and considering the various 
media types that the Cisco 12000 series Internet router supports.