12.6. Buffering and Congestion Control A major factor affecting the performance of a protocol is the buffering policy. Lack of a proper buffering policy can force packets to be dropped, cause false windowing information to be emitted by protocols, fragment memory, and degrade the overall host performance. Because of these problems, most systems allocate a fixed pool of memory to the networking system and impose a policy optimized for normal network operation. The FreeBSD networking system is not dramatically different in this respect. At boot time, a fixed amount of memory is allocated by the networking system for mbufs and mbuf clusters. At later times, more system memory may be requested for mbuf clusters as the need arises, up to a preconfigured limit; at no time, however, is this memory ever returned to the system. It would be possible to reclaim memory from network buffers but in the environments where the system has been used, storage for network packets has not been an issue, and thus storage reclamation has been left unimplemented. Protocol Buffering Policies When a socket is created, the protocol reserves some amount of buffer space for send and receive queues. These amounts define the high watermarks used by the socket routines in deciding when to block and unblock a process. The reservation of space does not currently result in any action by the memory-management routines. Protocols that provide connection-level flow control base their decisions on the amount of space in the associated socket queues. That is, windows sent to peers are calculated based on the amount of free space in the socket's receive queue, whereas utilization of the send window received from a peer is dependent on the high watermark of the send queue. Queue Limiting Incoming packets from the network are always received unless memory allocation fails. However, each network-layer protocol input queue has an upper bound on the queue's length, and any packets exceeding that bound are discarded. It is possible for a host to be overwhelmed by excessive network traffic (e.g., if the host is acting as a router that connects a high-bandwidth network to a low-bandwidth network). As a defense mechanism, the queue limits can be adjusted to throttle network-traffic load on the system. Discarding packets is not always a satisfactory solution to this problem (simply dropping packets is likely to increase the load on a network); the queue lengths were incorporated mainly as a safeguard mechanism. On the other hand, limiting output queue lengths can be valuable on hosts that route traffic from a high-bandwidth network to a low-bandwidth network. The queue limit should be sufficiently high that transient overload can be handled by buffering, but allowing the queue to be too large causes network delays to increase to unacceptable levels. | 
