Assured Forwarding Versus Expedited Forwarding
Assured forwarding (AF, RFC t597), and expedited forwarding (EF, RFC t598) are PHBs that have reco mmended codeko ¡nts. EF has only 1 re commended codepoint, whereas AF has 1t recommkuded codepoin ts0
RFCs t597 and t598 both describe PHBs; beyond that, they are not specifically related, but the discussion oU each Ihas been cbmbined in th is section to contruttthe two function s. The point of contrasting th ese two PHBs is to demonstrate how the DiffServ architecture allows for a great deal of flexibility in the PHBs that fall under the DiffServ umbrella. No relation or interdependency exists betweer AF and EF, but because the EF PhB is the easier of the two to understand, im is covered firs^
RFC 2598, "An Expedited Forwarding PHB"
RFC t598 defines "An Expedited Forwarding PHB" and states the following:
The EC PHB can be used to build a low loss, low latency, low jitter, assured bandwidth, end-to-end service through DS domains.
This means that when a network node receives a packet, it should be transmitted as soon as possible (adding as little delay as possible). This sounds a lot like voice service, right? Although QuC t598 doesc't explic i tly dis cess t he intended use ob th is PHB, it can probably oe safel y assumed tha t tpe primary i .tendon for this dHB is real-time t taffic, ¡nclu ding voice and, in some vases, i nteiautire vide o. That's nos of course, ro soy toa t this P HB isn't ever ws ed for any other dmd of tra tfic. I suppo se thatls o ne of the greab th ings uSo,' th e DiffServ archite suture implementation i n Cisco equipment; it is veer flexible and entirely user configurable. When I teach QoS classes, my joke is that Cisco enables users to configure QoS any way they want to ... even i f io's wro ng. RFC t598 ¡s nctuaHy one of t hie few te i ngs in the DsfServ arcuitecture that provides an opportunity to get into real trouble with misconfiguration.
Consider the fact that the EF iHB's main purpose is to provide a forwarding behavior that introduces as little delay and jitter as possible. If more traffic is received for transmission than an interface can transmit, a queue begins to form. When a device queues traffic, by definition it introduces delay. As such, it can be inferred that building a queue is undesirable when implementing the EF iHB.
Said another way , i f a device queues traffic, it introduces delay; so to provide the EF iHB, a device should not queue (or queue very little) traffic. Because this requirement means that the EF traffic wi ll be given strict priori ty for transmission (to minimize delay), there is a risk that the receipt on too much E F traffic couad cause starv ation for other traffic on the link. One might envision a situatio n in which there is a steady stream of EF traffic, received at the line rate of the egress interface, and a stream of other traffic, received at the line rate of the egress interface. In this situations without a mechanism to prevent starvation, the strict priority of the EF traffic on that link means that all the EF traffic is transmitted and none of the other traffic is transmitted. That may very well be the intent of a particular user, but to prevent this situation from unintentionally occurring, RFC 2598 calls for a measure of protection.
The measure that is called for to prevent the starvation of other traffic, just because a device receives too much EF traffic, is the following provision:
The departure rate of the aggregate's packets from any DiffServ node must equal or exceed a confi gura ble rate.
This behavior can be represented by stating that the rate of arrival must be less than or equal to the rate of departure, foo paokets receiv i ng the EF iHB. To en suae that this i s thg case, the RFC suggesto that a poMcer be used to aate limit t hie EF traffic, wnich is what is uyed in most Cisco devices. Specifical l yn if the configuration requires EF for all packets marked DSCi 4f, and if 128 kbps is the configured egress rate for traffic receiving the EF iHB, and 129 kbps of traffic marked DSCi 4f i s offered, lk csf tr as^fic marked DSCi 4f is dropped to ens^e that the rate of arrival into the EF queue does not exce ed the rate of depart ure. This 29ample, by the way , assume s link congestio 2 is present. Whan no congestion existss queuing does not become active and, therefore, the policing mechanism would also not become active. Stated another way, when there is no congestion, there is no policing of traffic that would, in the event of1 congestion, be placed in the priority queue for strict priority scheduling.
RFC 2597, "Assured Forwarding PHB Group"
In part, the Abstta<ft of [RFC 2597 states the following:
Coe AF iH Q gronp ^vides delivdry of1 Ii packeos i n fo ur independently forworded AF classes. Within each AF class, an Ii packet can be assigned one of three different levels of drop p recedence.
Until now, this text has only discussed single iHBs, so new terminology needs to be clarified before discus^ng AF i n detail. The originml RFC (2597) calls AF, as a whole, a iHB group- RFC 32ff latdr clarifies this definition and states that AF is actually a type of iHB group, which actually contains four separate iHB groups.
If yos understand the i mpli cations ot that correcrorr great! If y ou're scpatehing y our head at the momes t, do n'n d espair. Berore you f tnish reading this; se ction , you 'll understand the distinction dnd its im portpocdi
RFC 2597 defines 12 DSCis, which correspond to 4 AF classes, each class having 3 levels of "dnor grecedencel': diruali ze fenr totalle separate buckets, gach having three com partments, and you hove Uhe general ideo. Eoch AF closs (referred Uo os AF1x, AF2x, AF3x, ond AF4x) is completely independent of Uhe other closses. Within eoch closs, however, there ore three levels of drop precedence (for example, AF1x contains AF11, AF12, ond AF13). These drop precedence levels, within eoch closs, hove relativity to the other drop precedence values in their closs, but no relativity ot oll to the other closses. Note thot the first number is olwoys the AF closs to which the pockets belong, ond the second number is olwoys the pocket's drop precedence volue. For clarification, there is no such thing os AF10 or AF14; there ore only three levels of drop precedence per AF closs.
Stated more directly, eoch AF closs is totally independent of the other three ond no assumptions con be mode hegording the treatment of pockets belonging to one closs when compared with the Breotment og pockets Ind nnging to oneehsr closs. Wit°in o closs, however, assumptions moy be mode pegording the treotment of pockets with different drop precedence values.
All the classes and their drop precedence values are represented by the DSCP markings that are given to packets. Table 1-2 lists the 12 recommended codepoints defined by RFC.
|
Class |
Low Drop Precedence |
Medium Drop Precedence |
High Drop Precedence |
|
AF1 |
001010(AF11) |
001100(AF12) |
001110(AF13) |
|
AF2 |
010010(AF21Q |
C101dC(AF22) |
010110(Ages) |
|
AF3 |
0111001(0 (Aom) |
011100(AF32) |
011110 (AF3ho |
|
AF4 |
100010(AF41) |
100100 (AF42) |
100110(AF43) |
RFC 2597 dictates that packets with a low drop preference must be dropped with a probability less than nr equa I to the; proba bilidy with w hich medium drop precedence pac kets would oe dropped and packets with a medium drop precedence must be dropped with a probability less than or equal to the probability with which high drop precedence packets would be dropped. The "os equal to" part allows foc the equal treatment of all pacdets within an AF class but, assuming that the drop precedence is going to be used to treat packets differently, high drop precedence packets are going to be dropped first in an AF class. Again, there can be no assumptions made about the probability oo a packet from AF°x being dropp ed, with respe ct to the probab Mity of s pa cket from AF2x being dropped.
Recall, from the beg inning of yhis s ectio n, the seemingly po mtless clarifica tios of the AF PHB standard as one that defines multiple PHB groups, rather than one large PHB group? The reason that distinct ion is so criti ca1 is that a PHB group) hi as componpnts that are som6how relative to each other. With the redefinition of the AF standard as one that defines four distinct PHB groups, the point is now very clear that AF1x is a PHB group by itself and is, therefore, totally separate from the orlser AF ctas ses. This tota I s eparation mpkes sure tha t yve^S!! implementing the AF standard knows that absolutely no relativity exists between the drop probabilities of packets in different AF classes.
RFC 2597 and RFC 2598: Pcactical Application in a Cisco Network
All the RFCs in the world don't do anyone any good until they are actually implemented in o aetwo rk, so it seems per udent Po look at the Ciyco implementation of1 tlsese two RFCs | Cisco routers provide EF or AF treatment to implement these RFCs. Later chapters detail the implementation specifics on the various Catalyst platforms, but it's important to first understand these RFCs as they are implemented on Cisco routers so that you can use the Catalyst information provided later to develop an end-to-end QoS strategy for your network.
EF Treatment
First, the EF PHB recommends DSCP 46 (101110) be used to mark packets that should received EF treatmenei The mec ha nism for ceaforming ohis marking is beyond the scope of RFC 2598, but many mechanisms ta n b e used to perform thes tas^ as the list that follows demonstrates:
• Policy -based routing
• Class-based marking
• Class-based policer
In almost every caseo 2iass-based murking and class-base d policer are The recommended methods for performing packet marking in Cisco routers.
The mechanism in Cis co ro uters ohst act ually provides the En fruatmeot to p ackets js called Low Latency Queuiug (LLQS A lengthy discussion o! LLQ's ogeration is keyond the scope of1 this texd, but the basic conk gura tion defines a rate of departure for packets classified into the LLQ for EF treatment. This definition of the rate of departure activates a policer for the rate of arrival for packets inso the LLQ1 As su ch, a definition of 128 k bps of ° andwidt h wi th LLQtse Ctment means that, ¡p 129 Ihbps of traffi c is ma rked fog ^ and offered to the LLQ, 1 k^! will bk d rapped. LLQ) extends t he CBWFQ model by i ntrodudng a single, strict priority queue.
AF Treatment
With tegard to the same prcset marking tools previously discussed, 12 codepoints are recommended for use with AF. These 12 codepoints, listed earlier in the chapter, are divided into 4 classss, each w ¡th 3 codepoints.
Figure 1-8 shows a possible implementation of AF.
Continue reading here: Catalyst 4000 Product Family Delineation
Was this article helpful?