Dynamics of Network Congestion and Tail Drops

As traffic increases on a router's output interface, the queues on that interface start to fill with packets. As mentioned in Chapter 4, the interface fills because the bandwidth of the output link cannot keep up with the amount of traffic scheduled to go out that link. If the traffic continues to pass through the router at a rate that exceeds the speed of the output link, it is possible for the queue system to fill to its maximum capacity. When this happens, the router has no choice but to...

Verifying WRED Configuration

To verify that WRED is active on an interface, issue the show interface command RTA sh int S1 0 Seriall 0 is up, line protocol is up Hardware is M4T MTU 1500 bytes, BW 1544 Kbit, DLY 20000 jsec, reliability 255 255, txload 80 255, rxload 6 255 Encapsulation FRAME RELAY, crc 16, loopback not set Keepalive set (10 sec) LMI enq sent 23, LMI stat recvd 23, LMI upd recvd 0, DTE LMI up LMI enq recvd 0, LMI stat sent 0, LMI upd sent 0 LMI DLCI 1023 LMI type is CISCO frame relay DTE Broadcast queue 0...

Class Based WFQ

Class-based WFQ (CBWFQ) is a follow-on to regular WFQ (refer to Chapter 4) and supports the concept of user-defined traffic classes. Instead of queuing on a per-conversation (per-flow) basis, with CBWFQ you define groups or classes of traffic and then control the QoS for each class. While CBWFQ is not automatic like WFQ. it gives you greater control over traffic queuing and bandwidth allocation. Additionally, CBWFQ incorporates the WRED mechanism covered earlier, in the section Configuring...

Configuring CAR

To apply a rate policy on an interface, you configure by using the rate-limit command in interface configuration mode. The following example limits all inbound FTP traffic on HssiO 0 0 to 240 kbps of bandwidth with 32 KB of burst. Inbound FTP traffic that exceeds the average rate (240 kbps) and the excess burst size (32 KB) is dropped. rate-limit input access-group 101 240000 32O00 32000 conform-action transmit exceed-action drop access-list 101 permit tcp any any eq ftp-data access-list 101...

The RSVP Signaling Process

Prior to any RSVP signaling, the sender and receiver might have identified each other through their application and agreed to start a session. A conferencing application, for example, might offer a directory service that Client A can use to check if Client B is online and available for a call. The steps before RSVP signaling might look something like this When Client A calls Client B, Client A simply selects Client B's name in the directory list and clicks a button to initiale the call. Client...

RSVP and Weighted Fair Queuing

Just as RSVP is a signaling protocol that does not transport data for a session, RSVP itself is not responsible for queuing and dispatching packets for a session after the reservation is made. RSVP in a Cisco router depends on weighted fair queuing (WFQ), as discussed in Chapter 4, to carry out the queuing and dispatching of packets at the link layer (Layer 2) that ultimately delivers the QoS for a session. NOTE You can think of RSVP as the decision-maker for reservations and WFQ as the...

Committed Access Rate

Committed Access Rate (CAR) is used to control bandwidth coming into or going out of an interface. This bandwidth control is called rate limiting and is also known as policing. For example, you might choose to limit the bandwidth coming from a particular source or application. Traffic exceeding the threshold limit you specify can be dropped or reclassified (using IP precedence), based on the policy you define. CAR can also be used to classify packets into IP precedence levels without...

RED and IP Precedence Weighted RED

As a mechanism for QoS, RED drops low-precedence packets before high-precedence packets. This means high-priority applications are less likely to experience a packet drop (and TCP slow start) than low-priority applications. RED achieves QoS by maintaining thresholds at various queue depths one threshold for each precedence level. As a queue fills and exceeds a threshold, the precedence level associated with that threshold is eligible for random drops. A high-precedence level correlates to a...

Configuring IOS as a Proxy for Path and Resv Messages

Cisco ICS can send Path and Resv messages on behalf of clients. This proxy capability is useful for testing RSVP when you don't have any RSVP-enabled clients available. Figure 5-8 depicts such a scenario. Figure 5-8 Router A Can Proxy Path Messages for a Non-RSVP Client (Client A) Client A (sender) 192.168.20.1 non-RSVP Client A (sender) 192.168.20.1 non-RSVP Client B (receiver) 192.168.10.2 non-RSVP Client B (receiver) 192.168.10.2 non-RSVP Client A does nol support RSVP, so it cannot generate...

Effects of Global Synchronization and TCP Slow Start

Global synchronization combined with TCP slow starts can lead to some very undesirable results. When many TCP flows have one or more packets dropped in a tail drop, they all go through TCP slow start at the same time. This means these flows all slow down at the same time, and traffic on the network as a whole drops abruptly. The network gets significantly quieter than it was before the tail drop (again, this assumes most of the traffic is TCP). Next, because the flows are synchronized, they all...

How RED Works

RED randomly drops packets based on the number of packets queued on an interface As a queue reaches its maximum capacity, RED drops packets more aggressively to avoid a tail drop. RED throttles back Mows and takes advantage of TCP slow start. Rather than tail-dropping all packets when the queue is full. RED manages queue depth by randomly dropping some packets as the queue fills (past a certain threshold). As packets drop, the applications associated with those dropped packets slow down and go...

Configuring RSVP

Configuring RSVP is simple so simple, in fact, that the temptation is to enable RSVP without fully understanding how it works. This is not a good idea, which is why this chapter covers the mechanics before it discusses the configuration. RSVP is enabled on router interfaces, not on the router as a whole. This gives you the flexibility to enable RSVP on some interfaces but leave other interfaces alone. Figure 5-7 WFQ on Outbound Interfaces Ensures Data Is Delivered at the Desired QoS To...

Configuring CBWFQ

Configuring CBWTQ comprises three basic steps Step 1 Separate your traffic into classes with class maps. Step 2 Define the QoS for each class using policy maps. Step 3 Apply the policy map to an interface. The following sections demonstrate each of these steps using an example. Separate Your Traffic Into Classes with Class Maps The first step in CBWFQ is to separate your traffic into different classes so you can later apply QoS to those classes. Class maps define the names of your classes and...