Network Working Group S. Floyd Request for Comments: 4342 ICIR Category: Standards Track E. Kohler UCLA J. Padhye Microsoft Research March 2006 Profile for Datagram Congestion Control Protocol (DCCP) Congestion Control ID 3: TCP-Friendly Rate Control (TFRC) Status of This Memo This document specifies an Internet standards track protocol for the Internet community, and requests discussion and suggestions for improvements. Please refer to the current edition of the "Internet Official Protocol Standards" (STD 1) for the standardization state and status of this protocol. Distribution of this memo is unlimited. Copyright Notice Copyright (C) The Internet Society (2006).
AbstractThis document contains the profile for Congestion Control Identifier 3, TCP-Friendly Rate Control (TFRC), in the Datagram Congestion Control Protocol (DCCP). CCID 3 should be used by senders that want a TCP-friendly sending rate, possibly with Explicit Congestion Notification (ECN), while minimizing abrupt rate changes. 1. Introduction ....................................................2 2. Conventions .....................................................3 3. Usage ...........................................................3 3.1. Relationship with TFRC .....................................4 3.2. Half-Connection Example ....................................4 4. Connection Establishment ........................................5 5. Congestion Control on Data Packets ..............................5 5.1. Response to Idle and Application-Limited Periods ...........7 5.2. Response to Data Dropped and Slow Receiver .................8 5.3. Packet Sizes ...............................................9 6. Acknowledgements ................................................9 6.1. Loss Interval Definition ..................................10 6.1.1. Loss Interval Lengths ..............................12 6.2. Congestion Control on Acknowledgements ....................13
6.3. Acknowledgements of Acknowledgements ......................13 6.4. Determining Quiescence ....................................14 7. Explicit Congestion Notification ...............................14 8. Options and Features ...........................................14 8.1. Window Counter Value ......................................15 8.2. Elapsed Time Options ......................................17 8.3. Receive Rate Option .......................................17 8.4. Send Loss Event Rate Feature ..............................18 8.5. Loss Event Rate Option ....................................18 8.6. Loss Intervals Option .....................................18 8.6.1. Option Details .....................................19 8.6.2. Example ............................................20 9. Verifying Congestion Control Compliance with ECN ...............22 9.1. Verifying the ECN Nonce Echo ..............................22 9.2. Verifying the Reported Loss Intervals and Loss Event Rate ................................................23 10. Implementation Issues .........................................23 10.1. Timestamp Usage ..........................................23 10.2. Determining Loss Events at the Receiver ..................24 10.3. Sending Feedback Packets .................................25 11. Security Considerations .......................................27 12. IANA Considerations ...........................................28 12.1. Reset Codes ..............................................28 12.2. Option Types .............................................28 12.3. Feature Numbers ..........................................28 13. Thanks ........................................................29 A. Appendix: Possible Future Changes to CCID 3 ....................30 Normative References ..............................................31 Informative References ............................................31 Table 1: DCCP CCID 3 Options ......................................14 Table 2: DCCP CCID 3 Feature Numbers ..............................15 RFC4340]. DCCP uses Congestion Control Identifiers, or CCIDs, to specify the congestion control mechanism in use on a half-connection. TFRC is a receiver-based congestion control mechanism that provides a TCP-friendly sending rate while minimizing the abrupt rate changes characteristic of TCP or of TCP-like congestion control [RFC3448]. The sender's allowed sending rate is set in response to the loss
event rate, which is typically reported by the receiver to the sender. See Section 3 for more on application requirements. RFC2119]. All multi-byte numerical quantities in CCID 3, such as arguments to options, are transmitted in network byte order (most significant byte first). A DCCP half-connection consists of the application data sent by one endpoint and the corresponding acknowledgements sent by the other endpoint. The terms "HC-Sender" and "HC-Receiver" denote the endpoints sending application data and acknowledgements, respectively. Since CCIDs apply at the level of half-connections, we abbreviate HC-Sender to "sender" and HC-Receiver to "receiver" in this document. See [RFC4340] for more discussion. For simplicity, we say that senders send DCCP-Data packets and receivers send DCCP-Ack packets. Both of these categories are meant to include DCCP-DataAck packets. The phrases "ECN-marked" and "marked" refer to packets marked ECN Congestion Experienced unless otherwise noted. This document uses a number of variables from [RFC3448], including the following: o X_recv: The receive rate in bytes per second. See [RFC3448], Section 3.2.2. o s: The packet size in bytes. See [RFC3448], Section 3.1. o p: The loss event rate. See [RFC3448], Section 3.1. RFC4341], halves the sending rate in response to each congestion event and thus cannot provide a relatively smooth sending rate.
As explained in [RFC3448], Section 1, the penalty of having smoother throughput than TCP while competing fairly for bandwidth with TCP is that the TFRC mechanism in CCID 3 responds slower to changes in available bandwidth than do TCP or TCP-like mechanisms. Thus, CCID 3 should only be used for applications with a requirement for smooth throughput. For applications that simply need to transfer as much data as possible in as short a time as possible, we recommend using TCP-like congestion control, such as CCID 2. CCID 3 should also not be used by applications that change their sending rate by varying the packet size, rather than by varying the rate at which packets are sent. A new CCID will be required for these applications. RFC3448]. Conforming CCID 3 implementations MAY track updates to the TCP throughput equation directly, as updates are standardized in the IETF, rather than wait for revisions of this document. However, conforming implementations SHOULD wait for explicit updates to CCID 3 before implementing other changes to TFRC congestion control. RFC3448], Section 3.2. Each DCCP-Data packet has a sequence number and the DCCP header's CCVal field contains the window counter value, which is used by the receiver in determining when multiple losses belong in a single loss event. In the typical case of an ECN-capable half-connection, each DCCP- Data and DCCP-DataAck packet is sent as ECN Capable, with either the ECT(0) or the ECT(1) codepoint set. The use of the ECN Nonce with TFRC is described in Section 9. 2. The receiver sends DCCP-Ack packets acknowledging the data packets at least once per round-trip time, unless the sender is sending at a rate of less than one packet per round-trip time, as indicated by the TFRC specification ([RFC3448], Section 6). Each DCCP-Ack packet uses a sequence number, identifies the most recent packet
received from the sender, and includes feedback about the recent loss intervals experienced by the receiver. 3. The sender continues sending DCCP-Data packets as controlled by the allowed transmit rate. Upon receiving DCCP-Ack packets, the sender updates its allowed transmit rate as specified in [RFC3448], Section 4.3. This update is based on a loss event rate calculated by the sender using the receiver's loss intervals feedback. If it prefers, the sender can also use a loss event rate calculated and reported by the receiver. 4. The sender estimates round-trip times and calculates a nofeedback time, as specified in [RFC3448], Section 4.4. If no feedback is received from the receiver in that time (at least four round-trip times), the sender halves its sending rate. RFC4340]. During or after CCID 3 negotiation, the client and/or server may want to negotiate the values of the Send Ack Vector and Send Loss Event Rate features. RFC3448]. The following discussion summarizes information from [RFC3448], which should be considered normative except where specifically indicated otherwise. Loss Event Rate The basic operation of CCID 3 centers around the calculation of a loss event rate: the number of loss events as a fraction of the number of packets transmitted, weighted over the last several loss intervals. This loss event rate, a round-trip time estimate, and the average packet size are plugged into the TCP throughput equation, as specified in [RFC3448], Section 3.1. The result is a fair transmit rate close to what a modern TCP would achieve in the same conditions. CCID 3 senders are limited to this fair rate. The loss event rate itself is calculated in CCID 3 using recent loss interval lengths reported by the receiver. Loss intervals are precisely defined in Section 6.1. In summary, a loss interval is up to 1 RTT of possibly lost or ECN-marked data packets, followed by an arbitrary number of non-dropped, non-marked data packets. Thus, long loss intervals represent low congestion rates. The CCID 3 Loss
Intervals option is used to report loss interval lengths; see Section 8.6. Other Congestion Control Mechanisms The sender starts in a slow-start phase, roughly doubling its allowed sending rate each round-trip time. The slow-start phase is ended by the receiver's report of a data packet drop or mark, after which the sender uses the loss event rate to calculate its allowed sending rate. [RFC3448], Section 4, specifies an initial sending rate of one packet per round-trip time (RTT) as follows: The sender initializes the allowed sending rate to one packet per second. As soon as a feedback packet is received from the receiver, the sender has a measurement of the round-trip time and then sets the initial allowed sending rate to one packet per RTT. However, while the initial TCP window used to be one segment, [RFC2581] allows an initial TCP window of two segments, and [RFC3390] allows an initial TCP window of three or four segments (up to 4380 bytes). [RFC3390] gives an upper bound on the initial window of min(4*MSS, max(2*MSS, 4380 bytes)). Therefore, in contrast to [RFC3448], the initial CCID 3 sending rate is allowed to be at least two packets per RTT, and at most four packets per RTT, depending on the packet size. The initial rate is only allowed to be three or four packets per RTT when, in terms of segment size, that translates to at most 4380 bytes per RTT. The sender's measurement of the round-trip time uses the Elapsed Time and/or Timestamp Echo option contained in feedback packets, as described in Section 8.2. The Elapsed Time option is required, while the Timestamp Echo option is not. The sender maintains an average round-trip time heavily weighted on the most recent measurements. Each DCCP-Data packet contains a sequence number. Each DCCP-Data packet also contains a window counter value, as described in Section 8.1. The window counter is generally incremented by one every quarter round-trip time. The receiver uses it as a coarse-grained timestamp to determine when a packet loss should be considered part of an existing loss interval and when it must begin a new loss interval. Because TFRC is rate-based instead of window-based, and because feedback packets can be dropped in the network, the sender needs some mechanism for reducing its sending rate in the absence of positive feedback from the receiver. As described in Section 6, the receiver sends feedback packets roughly once per round-trip time. As specified in [RFC3448], Section 4.3, the sender sets a nofeedback
timer to at least four round-trip times, or to twice the interval between data packets, whichever is larger. If the sender hasn't received a feedback packet from the receiver when the nofeedback timer expires, then the sender halves its allowed sending rate. The allowed sending rate is never reduced below one packet per 64 seconds. Note that not all acknowledgements are considered feedback packets, since feedback packets must contain valid Loss Intervals, Elapsed Time, and Receive Rate options. If the sender never receives a feedback packet from the receiver, and as a consequence never gets to set the allowed sending rate to one packet per RTT, then the sending rate is left at its initial rate of one packet per second, with the nofeedback timer expiring after two seconds. The allowed sending rate is halved each time the nofeedback timer expires. Thus, if no feedback is received from the receiver, the allowed sending rate is never above one packet per second and is quickly reduced below one packet per second. The feedback packets from the receiver contain a Receive Rate option specifying the rate at which data packets arrived at the receiver since the last feedback packet. The allowed sending rate can be at most twice the rate received at the receiver in the last round-trip time. This may be less than the nominal fair rate if, for example, the application is sending less than its fair share. RFC3448], Section 4.4, the allowed sending rate is never reduced to fewer than two packets per round- trip time as the result of an idle period. CCID 3 revises this to take into account the larger initial windows allowed by [RFC3390]: the allowed sending rate is never reduced to less than the [RFC3390] initial sending rate as the result of an idle period. If the allowed sending rate is less than the initial sending rate upon entry to the idle period, then it will still be less than the initial sending rate when the idle period is exited. However, if the allowed sending rate is greater than or equal to the initial sending rate upon entry to the idle period, then it should not be reduced below the initial sending rate no matter how long the idle period lasts. The sender's allowed sending rate is limited to at most twice the receive rate reported by the receiver. Thus, after an application- limited period, the sender can at most double its sending rate from one round-trip time to the next, until it reaches the allowed sending rate determined by the loss event rate.
RFC4340], with the following further clarifications. o Drop Code 2 ("receive buffer drop"). The allowed sending rate is reduced by one packet per RTT for each packet newly acknowledged as Drop Code 2, except that it is never reduced below one packet per RTT as a result of Drop Code 2. o Adjusting the receive rate X_recv. A CCID 3 sender SHOULD also respond to non-network-congestion events, such as those implied by Data Dropped and Slow Receiver options, by adjusting X_recv, the receive rate reported by the receiver in Receive Rate options (see Section 8.3). The CCID 3 sender's allowed sending rate is limited to at most twice the receive rate reported by the receiver via the "min(..., 2*X_recv)" clause in TFRC's throughput calculations ([RFC3448], Section 4.3). When the sender receives one or more Data Dropped and Slow Receiver options, the sender adjusts X_recv as follows: 1. X_inrecv is equal to the Receive Rate in bytes per second reported by the receiver in the most recent acknowledgement. 2. X_drop is set to the sending rate upper bound implied by Data Dropped and Slow Receiver options. If the sender receives a Slow Receiver option, which requests that the sender not increase its sending rate for roughly a round-trip time [RFC4340], then X_drop should be set to X_inrecv. Similarly, if the sender receives a Data Dropped option indicating, for example, that three packets were dropped with Drop Code 2, then the upper bound on the sending rate will be decreased by at most three packets per RTT, by the sender setting X_drop to max(X_inrecv - 3*s/RTT, min(X_inrecv, s/RTT)). Again, s is the packet size in bytes. 3. X_recv is then set to min(X_inrecv, X_drop/2). As a result, the next round-trip time's sending rate will be limited to at most 2*(X_drop/2) = X_drop. The effects of the Slow Receiver and Data Dropped options on X_recv will mostly vanish by
the round-trip time after that, which is appropriate for this non-network-congestion feedback. This procedure MUST only be used for those Drop Codes not related to corruption (see [RFC4340]). Currently, this is limited to Drop Codes 0, 1, and 2. Section 6.1. Alternately, a CCID 3 implementation MAY use the Maximum Packet Size to derive s. In this case, s is set to the Maximum Segment Size (MSS), the maximum size in bytes for the data segment, not including the default DCCP and IP packet headers. Each packet transmitted then counts as one MSS, regardless of the actual segment size, and the TCP throughput equation can be interpreted as specifying the sending rate in packets per second. CCID 3 implementations MAY check for applications that appear to be manipulating the packet size inappropriately. For example, an application might send small packets for a while, building up a fast rate, then switch to large packets to take advantage of the fast rate. (Preliminary simulations indicate that applications may not be able to increase their overall transfer rates this way, so it is not clear that this manipulation will occur in practice [V03].) RFC3448], Section 6. Each feedback packet contains an Acknowledgement Number, which equals the greatest valid sequence number received so far on this connection. ("Greatest" is, of course, measured in circular sequence space.) Each feedback packet also includes at least the following options:
1. An Elapsed Time and/or Timestamp Echo option specifying the amount of time elapsed since the arrival at the receiver of the packet whose sequence number appears in the Acknowledgement Number field. These options are described in [RFC4340], Section 13. 2. A Receive Rate option, defined in Section 8.3, specifying the rate at which data was received since the last DCCP-Ack was sent. 3. A Loss Intervals option, defined in Section 8.6, specifying the most recent loss intervals experienced by the receiver. (The definition of a loss interval is provided below.) From Loss Intervals, the sender can easily calculate the loss event rate p using the procedure described in [RFC3448], Section 5.4. Acknowledgements not containing at least these three options are not considered feedback packets. The receiver MAY also include other options concerning the loss event rate, including Loss Event Rate, which gives the loss event rate calculated by the receiver (Section 8.5), and DCCP's generic Ack Vector option, which reports the specific sequence numbers of any lost or marked packets ([RFC4340], Section 11.4). Ack Vector is not required by CCID 3's congestion control mechanisms: the Loss Intervals option provides all the information needed to manage the transmit rate and probabilistically verify receiver feedback. However, Ack Vector may be useful for applications that need to determine exactly which packets were lost. The receiver MAY also include other acknowledgement-related options, such as DCCP's Data Dropped option ([RFC4340], Section 11.7). If the HC-Receiver is also sending data packets to the HC-Sender, then it MAY piggyback acknowledgement information on those data packets more frequently than TFRC's specified acknowledgement rate allows. RFC3448], Section 5.2, a loss interval begins with a lost or ECN-marked data packet; continues with at most one round-trip time's worth of packets that may or may not be lost or marked; and completes with an arbitrarily long series of non-dropped, non-marked data packets. For example, here is a single loss interval, assuming that sequence numbers increase as you move right:
Lossy Part <= 1 RTT __________ Lossless Part __________ / \/ \ *----*--*--*------------------------------------- ^ ^ ^ ^ losses or marks Note that a loss interval's lossless part might be empty, as in the first interval below: Lossy Part Lossy Part <= 1 RTT <= 1 RTT _____ Lossless Part _____ / \/ \/ \ *----*--*--***--------*-*--------------------------- ^ ^ ^ ^^^ ^ ^ \_ Int. 1 _/\_____________ Interval 2 _____________/ As in [RFC3448], Section 5.2, the length of the lossy part MUST be less than or equal to 1 RTT. CCID 3 uses window counter values, not receive times, to determine whether multiple packets occurred in the same RTT and thus belong to the same loss event; see Section 10.2. A loss interval whose lossy part lasts for more than 1 RTT, or whose lossless part contains a dropped or marked data packet, is invalid. A missing data packet doesn't begin a new loss interval until NDUPACK packets have been seen after the "hole", where NDUPACK = 3. Thus, up to NDUPACK of the most recent sequence numbers (including the sequence numbers of any holes) might temporarily not be part of any loss interval while the implementation waits to see whether a hole will be filled. See [RFC3448], Section 5.1, and [RFC2581], Section 3.2, for further discussion of NDUPACK. As specified by [RFC3448], Section 5, all loss intervals except the first begin with a lost or marked data packet, and all loss intervals are as long as possible, subject to the validity constraints above. Lost and ECN-marked non-data packets may occur freely in the lossless part of a loss interval. (Non-data packets consist of those packet types that cannot carry application data; namely, DCCP-Ack, DCCP- Close, DCCP-CloseReq, DCCP-Reset, DCCP-Sync, and DCCP-SyncAck.) In the absence of better information, a receiver MUST conservatively assume that every lost packet was a data packet and thus must occur in some lossy part. DCCP's NDP Count option can help the receiver determine whether a particular packet contained data; see [RFC4340], Section 7.7.
RFC3448] defines the TFRC congestion control mechanism in terms of a one-way transfer of data, with data packets going from the sender to the receiver and feedback packets going from the receiver back to the sender. However, CCID 3 applies in a context of two half- connections, with DCCP-Data and DCCP-DataAck packets from one half- connection sharing sequence number space with DCCP-Ack packets from the other half-connection. For the purposes of CCID 3 congestion control, loss interval lengths should include data packets and should exclude the acknowledgement packets from the reverse half-connection. However, it is also useful to report the total number of packets in each loss interval (for example, to facilitate ECN Nonce verification). CCID 3's Loss Intervals option thus reports three lengths for each loss interval, the lengths of the lossy and lossless parts defined above and a separate data length. First, the lossy and lossless lengths are measured in sequence numbers. Together, they sum to the interval's sequence length, which is the total number of packets the sender transmitted during the interval. This is easily calculated in DCCP as the greatest packet sequence number in the interval minus the greatest packet sequence number in the preceding interval (or, if there is no preceding interval, then the predecessor to the half- connection's initial sequence number). The interval's data length, however, is the number used in TFRC's loss event rate calculation, as defined in [RFC3448], Section 5, and is calculated as follows. For all loss intervals except the first, the data length equals the sequence length minus the number of non-data packets the sender transmitted during the loss interval, except that the minimum data length is one packet. In the absence of better information, an endpoint MUST conservatively assume that the loss interval contained only data packets, in which case the data length equals the sequence length. To achieve greater precision, the sender can calculate the exact number of non-data packets in an interval by remembering which sent packets contained data; the receiver can account for received non-data packets by not including them in the data length, and for packets that were not received, it may be able to discriminate between lost data packets and lost non-data packets using DCCP's NDP Count option. The first loss interval's data length is undefined until the first loss event. [RFC3448], Section 6.3.1 specifies how the first loss interval's data length is calculated once the first loss event has occurred; this calculation uses X_recv, the most recent receive rate, as input. Until this first loss event, the loss event rate is zero,
as is the data length reported for the interval in the Loss Intervals option. The first loss interval's data length might be less than, equal to, or even greater than its sequence length. Any other loss interval's data length must be less than or equal to its sequence length. A sender MAY use the loss event rate or loss interval data lengths as reported by the receiver, or it MAY recalculate loss event rate and/or loss interval data lengths based on receiver feedback and additional information. For example, assume the network drops a DCCP-Ack packet with sequence number 50. The receiver might then report a loss interval beginning at sequence number 50. If the sender determined that this loss interval actually contained no lost or ECN-marked data packets, then it might coalesce the loss interval with the previous loss interval, resulting in a larger allowed transmit rate. RFC3448], Section 6). The sending rate for acknowledgements is relatively low -- roughly once per round-trip time -- so there is no need for explicit congestion control on acknowledgements.
Loss Intervals option is bounded in size (and receiver state), so acks-of-acks are not required. RFC4340], Section 11.1, for general information on quiescence. Let T equal the greater of 0.2 seconds and two round-trip times. (A CCID 3 receiver has a rough measure of the round-trip time so that it can pace its acknowledgements.) The receiver detects that the sender has gone quiescent after T seconds have passed without receiving any additional data from the sender. RFC3168]. In the typical case of an ECN-capable half-connection (where the receiver's ECN Incapable feature is set to zero), the sender will use the ECN Nonce for its data packets, as specified in [RFC4340], Section 12.2. Information about the ECN Nonce MUST be returned by the receiver using the Loss Intervals option, and any Ack Vector options MUST include the ECN Nonce Sum. The sender MAY maintain a table with the ECN nonce sum for each packet and use this information to probabilistically verify the ECN nonce sums returned in Loss Intervals or Ack Vector options. Section 9 describes this further. Option DCCP- Section Type Length Meaning Data? Reference ----- ------ ------- ----- --------- 128-191 Reserved 192 6 Loss Event Rate N 8.5 193 variable Loss Intervals N 8.6 194 6 Receive Rate N 8.3 195-255 Reserved Table 1: DCCP CCID 3 Options
The "DCCP-Data?" column indicates that all currently defined CCID 3- specific options MUST be ignored when they occur on DCCP-Data packets. The following CCID-specific feature is also defined. Rec'n Initial Section Number Meaning Rule Value Req'd Reference ------ ------- ----- ----- ----- --------- 128-191 Reserved 192 Send Loss Event Rate SP 0 N 8.4 193-255 Reserved Table 2: DCCP CCID 3 Feature Numbers The column meanings are described in [RFC4340], Table 4. "Rec'n Rule" defines the feature's reconciliation rule, where "SP" means server-priority. "Req'd" specifies whether every CCID 3 implementation MUST understand a feature; Send Loss Event Rate is optional, in that it behaves like an extension ([RFC4340], Section 15). RFC3448], Section 3.2.1. Window counters use circular arithmetic modulo 16 for all operations, including comparisons; see [RFC4340], Section 3.1, for more information on circular arithmetic. For reference, the DCCP generic header is as follows. (The diagram is repeated from [RFC4340], Section 5.1, which also shows the generic header with a 24-bit Sequence Number field.) 0 1 2 3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Source Port | Dest Port | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Data Offset | CCVal | CsCov | Checksum | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ | Res | Type |1| Reserved | Sequence Number (high bits) . +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ . Sequence Number (low bits) | +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
The CCVal field has enough space to express 4 round-trip times at quarter-RTT granularity. The sender MUST avoid wrapping CCVal on adjacent packets, as might happen, for example, if two data-carrying packets were sent 4 round-trip times apart with no packets intervening. Therefore, the sender SHOULD use the following algorithm for setting CCVal. The algorithm uses three variables: "last_WC" holds the last window counter value sent, "last_WC_time" is the time at which the first packet with window counter value "last_WC" was sent, and "RTT" is the current round-trip time estimate. last_WC is initialized to zero, and last_WC_time to the time of the first packet sent. Before sending a new packet, proceed like this: Let quarter_RTTs = floor((current_time - last_WC_time) / (RTT/4)). If quarter_RTTs > 0, then: Set last_WC := (last_WC + min(quarter_RTTs, 5)) mod 16. Set last_WC_time := current_time. Set the packet header's CCVal field to last_WC. When this algorithm is used, adjacent data-carrying packets' CCVal counters never differ by more than five, modulo 16. The window counter value may also change as feedback packets arrive. In particular, after receiving an acknowledgement for a packet sent with window counter WC, the sender SHOULD increase its window counter, if necessary, so that subsequent packets have window counter value at least (WC + 4) mod 16. The CCVal counters are used by the receiver to determine whether multiple losses belong to a single loss event, to determine the interval to use for calculating the receive rate, and to determine when to send feedback packets. None of these procedures require the receiver to maintain an explicit estimate of the round-trip time. However, implementors who wish to keep such an RTT estimate may do so using CCVal. Let T(I) be the arrival time of the earliest valid received packet with CCVal = I. (Of course, when the window counter value wraps around to the same value mod 16, we must recalculate T(I).) Let D = 2, 3, or 4 and say that T(K) and T(K+D) both exist (packets were received with window counters K and K+D). Then the value (T(K+D) - T(K)) * 4/D MAY serve as an estimate of the round- trip time. Values of D = 4 SHOULD be preferred for RTT estimation. Concretely, say that the following packets arrived: Time: T1 T2 T3 T4 T5 T6 T7 T8 T9 ------*---*---*-*----*------------*---*----*--*----> CCVal: K-1 K-1 K K K+1 K+3 K+4 K+3 K+4
Then T7 - T3, the difference between the receive times of the first packet received with window counter K+4 and the first packet received with window counter K, is a reasonable round-trip time estimate. Because of the necessary constraint that measurements only come from packet pairs whose CCVals differ by at most 4, this procedure does not work when the inter-packet sending times are significantly greater than the RTT, resulting in packet pairs whose CCVals differ by 5. Explicit RTT measurement techniques, such as Timestamp and Timestamp Echo, should be used in that case. RFC4340]. Section 8.1.) The receive rate then equals the number of data bytes received in the most recent t seconds, divided by t. Receive Rate options MUST NOT be sent on DCCP-Data packets, and any Receive Rate options on received DCCP-Data packets MUST be ignored.
RFC3448], Section 5, for a normative calculation of loss event rate. Before any losses have occurred, when the loss event rate is zero, the Loss Event Rate option value is set to "11111111111111111111111111111111" in binary (or, equivalently, to 2^32 - 1). The loss event rate calculation uses loss interval data lengths, as defined in Section 6.1.1. Loss Event Rate options MUST NOT be sent on DCCP-Data packets, and any Loss Event Rate options on received DCCP-Data packets MUST be ignored.
Each 9-byte Loss Interval contains three fields, as follows: ____________________ Loss Interval _____________________ / \ +--------...-------+--------...--------+--------...--------+ | Lossless Length |E| Loss Length | Data Length | +--------...-------+--------...--------+--------...--------+ 3 bytes 3 bytes 3 bytes The receiver reports its observed loss intervals using a Loss Intervals option. Section 6.1 defines loss intervals. This option MUST be sent by the data receiver on all required acknowledgements. The option reports up to 28 loss intervals seen by the receiver, although TFRC currently uses at most the latest 9 of these. This lets the sender calculate a loss event rate and probabilistically verify the receiver's ECN Nonce Echo. The Loss Intervals option serves several purposes. o The sender can use the Loss Intervals option to calculate the loss event rate. o Loss Intervals information is easily checked for consistency against previous Loss Intervals options, and against any Loss Event Rate calculated by the receiver. o The sender can probabilistically verify the ECN Nonce Echo for each Loss Interval, reducing the likelihood of misbehavior. Loss Intervals options MUST NOT be sent on DCCP-Data packets, and any Loss Intervals options on received DCCP-Data packets MUST be ignored.
The NINTERVAL parameter is equal to "n" as defined in [RFC3448], Section 5.4. Loss interval sequence numbers are delta encoded starting from the Acknowledgement Number. Therefore, Loss Intervals options MUST NOT be sent on packets without an Acknowledgement Number, and any Loss Intervals options received on such packets MUST be ignored. The first byte of option data is Skip Length, which indicates the number of packets up to and including the Acknowledgement Number that are not part of any Loss Interval. As discussed above, Skip Length must be less than or equal to NDUPACK = 3. In a packet containing multiple Loss Intervals options, the Skip Lengths of the second and subsequent options MUST equal zero; such options with nonzero Skip Lengths MUST be ignored. Loss Interval structures follow Skip Length. Each Loss Interval consists of a Lossless Length, a Loss Length, an ECN Nonce Echo (E), and a Data Length. Lossless Length, a 24-bit number, specifies the number of packets in the loss interval's lossless part. Note again that this part may contain lost or marked non-data packets. Loss Length, a 23-bit number, specifies the number of packets in the loss interval's lossy part. The sum of the Lossless Length and the Loss Length equals the loss interval's sequence length. Receivers SHOULD report the minimum valid Loss Length for each loss interval, making the first and last sequence numbers in each lossy part correspond to lost or marked data packets. The ECN Nonce Echo, stored in the high-order bit of the 3-byte field containing Loss Length, equals the one-bit sum (exclusive-or, or parity) of data packet nonces received over the loss interval's lossless part (which is Lossless Length packets long). If Lossless Length is 0, the receiver is ECN Incapable, or the Lossless Length contained no data packets, then the ECN Nonce Echo MUST be reported as 0. Note that any ECN nonces on received non-data packets MUST NOT contribute to the ECN Nonce Echo. Finally, Data Length, a 24-bit number, specifies the loss interval's data length, as defined in Section 6.1.1.
Sequence Numbers: 0 10 20 30 40 44 | | | | | | ----------*--------***-*--------*----------*- Assuming that packet 43 was lost, not marked, this sequence might be divided into loss intervals as follows: 0 10 20 30 40 44 | | | | | | ----------*--------***-*--------*----------*- \________/\_______/\___________/\_________/ L0 L1 L2 L3 A Loss Intervals option sent on a packet with Acknowledgement Number 44 to acknowledge this set of loss intervals might contain the bytes 193,39,2, 0,0,10, 128,0,1, 0,0,10, 0,0,8, 0,0,5, 0,0,10, 0,0,8, 0,0,1, 0,0,8, 0,0,10, 128,0,0, 0,0,15. This option is interpreted as follows. 193 The Loss Intervals option number. 39 The length of the option, including option type and length bytes. This option contains information about (39 - 3)/9 = 4 loss intervals. 2 The Skip Length is 2 packets. Thus, the most recent loss interval, L3, ends immediately before sequence number 44 - 2 + 1 = 43. 0,0,10, 128,0,1, 0,0,10 These bytes define L3. L3 consists of a 10-packet lossless part (0,0,10), preceded by a 1-packet lossy part. Continuing to subtract, the lossless part begins with sequence number 43 - 10 = 33, and the lossy part begins with sequence number 33 - 1 = 32. The ECN Nonce Echo for the lossless part (namely, packets 33 through 42, inclusive) equals 1. The interval's data length is 10, so the receiver believes that the interval contained exactly one non-data packet. 0,0,8, 0,0,5, 0,0,10 This defines L2, whose lossless part begins with sequence number 32 - 8 = 24; whose lossy part begins with sequence number 24 - 5 = 19; whose ECN Nonce Echo (for packets [24,31]) equals 0; and whose data length is 10.
0,0,8, 0,0,1, 0,0,8 L1's lossless part begins with sequence number 11, its lossy part begins with sequence number 10, its ECN Nonce Echo (for packets [11,18]) equals 0, and its data length is 8. 0,0,10, 128,0,0, 0,0,15 L0's lossless part begins with sequence number 0, it has no lossy part, its ECN Nonce Echo (for packets [0,9]) equals 1, and its data length is 15. (This must be the first loss interval in the connection; otherwise, a data length greater than the sequence length would be invalid.) RFC3540], the nonce sum for sequence number S is the one-bit sum (exclusive-or, or parity) of data packet nonces over the sequence number range [I,S], where I is the initial sequence number. Let NonceSum(S) represent this nonce sum for sequence number S, and define NonceSum(I - 1) as 0. Note that NonceSum does not account for the nonces of non-data packets such as DCCP-Ack. Then the Nonce Echo for an interval of packets with sequence numbers X to Y, inclusive, should equal the following one-bit sum: NonceSum(X - 1) + NonceSum(Y) Since an ECN Nonce Echo is returned for the lossless part of each Loss Interval, a misbehaving receiver -- meaning a receiver that reports a lost or marked data packet as "received non-marked", to avoid rate reductions -- has only a 50% chance of guessing the correct Nonce Echo for each loss interval. To verify the ECN Nonce Echo included with an Ack Vector option, the sender maintains a table with the ECN nonce value sent for each packet. The Ack Vector option explicitly says which packets were
received non-marked; the sender just adds up the nonces for those packets using a one-bit sum and compares the result to the Nonce Echo encoded in the Ack Vector's option type. Again, a misbehaving receiver has only a 50% chance of guessing an Ack Vector's correct Nonce Echo. Alternatively, an Ack Vector's ECN Nonce Echo may also be calculated from a table of ECN nonce sums, rather than from ECN nonces. If the Ack Vector contains many long runs of non-marked, non-dropped packets, the nonce sum-based calculation will probably be faster than a straightforward nonce-based calculation. Note that Ack Vector's ECN Nonce Echo is measured over both data packets and non-data packets, while the Loss Intervals option reports ECN Nonce Echoes for data packets only. Thus, different nonce sum tables are required to verify the two options.
compute the round trip time. Alternatively, the sender MAY include Timestamp options on some of its data packets. The receiver will respond with Timestamp Echo options including Elapsed Times, allowing the sender to calculate round-trip times without storing sent packets' timestamps at all. RFC3448], Section 3.2.1 specifies that each data packet contains a timestamp and gives as an alternative implementation a "timestamp" that is incremented every quarter of an RTT, as is the window counter in CCID 3. However, [RFC3448], Section 5.2 on "Translation from Loss History to Loss Events" is written in terms of timestamps, not in terms of window counters. In this section, we give a procedure for the translation from loss history to loss events that is explicitly in terms of window counters. To determine whether two lost packets with sequence numbers X and Y belong to different loss events, the receiver proceeds as follows. Assume Y > X in circular sequence space. o Let X_prev be the greatest valid sequence number received with X_prev < X. o Let Y_prev be the greatest valid sequence number received with Y_prev < Y. o Given a sequence number N, let C(N) be the window counter value associated with that packet. o Packets X and Y belong to different loss events if there exists a packet with sequence number S so that X_prev < S <= Y_prev, and the distance from C(X_prev) to C(S) is greater than 4. (The distance is the number D so that C(X_prev) + D = C(S) (mod WCTRMAX), where WCTRMAX is the maximum value for the window counter -- in our case, 16.) That is, the receiver only considers losses X and Y as separate loss events if there exists some packet S received between X and Y, with the distance from C(X_prev) to C(S) greater than 4. This complex calculation is necessary in order to handle the case where
window counter space wrapped completely between X and Y. When that space does not wrap, the receiver can simply check whether the distance from C(X_prev) to C(Y_prev) is greater than 4; if so, then X and Y belong to separate loss events. Window counters can help the receiver disambiguate multiple losses after a sudden decrease in the actual round-trip time. When the sender receives an acknowledgement acknowledging a data packet with window counter i, the sender increases its window counter, if necessary, so that subsequent data packets are sent with window counter values of at least i+4. This can help minimize errors where the receiver incorrectly interprets multiple loss events as a single loss event. We note that if all of the packets between X and Y are lost in the network, then X_prev and Y_prev are equal, and the series of consecutive losses is treated by the receiver as a single loss event. However, the sender will receive no DCCP-Ack packets during a period of consecutive losses, and the sender will reduce its sending rate accordingly. As an alternative to the window counter, the sender could have sent its estimate of the round-trip time to the receiver directly in a round-trip time option; the receiver would use the sender's round- trip time estimate to infer when multiple lost or marked packets belong in the same loss event. In some respects, a round-trip time option would give a more precise encoding of the sender's round-trip time estimate than does the window counter. However, the window counter conveys information about the relative *sending* times for packets, while the receiver could only use the round-trip time option to distinguish between the relative *receive* times (in the absence of timestamps). That is, the window counter will give more robust performance when there is a large variation in delay for packets sent within a window of data. Slightly more speculatively, a round-trip time option might possibly be used more easily by middleboxes attempting to verify that a flow used conforming end-to-end congestion control. RFC3448], Sections 6.1 and 6.2 specify that the TFRC receiver must send a feedback packet when a newly calculated loss event rate p is greater than its previous value. CCID 3 follows this rule. In addition, [RFC3448], Section 6.2, specifies that the receiver use a feedback timer to decide when to send additional feedback packets. If the feedback timer expires and data packets have been received since the previous feedback was sent, then the receiver sends a
feedback packet. When the feedback timer expires, the receiver resets the timer to expire after R_m seconds, where R_m is the most recent estimate of the round-trip time received from the sender. CCID 3 receivers, however, generally use window counter values instead of a feedback timer to determine when to send additional feedback packets. This section describes how. Whenever the receiver sends a feedback message, the receiver sets a local variable last_counter to the greatest received value of the window counter since the last feedback message was sent, if any data packets have been received since the last feedback message was sent. If the receiver receives a data packet with a window counter value greater than or equal to last_counter + 4, then the receiver sends a new feedback packet. ("Greater" and "greatest" are measured in circular window counter space.) This procedure ensures that when the sender is sending at a rate less than one packet per round-trip time, the receiver sends a feedback packet after each data packet. Similarly, this procedure ensures that when the sender is sending several packets per round-trip time, the receiver will send a feedback packet each time that a data packet arrives with a window counter at least four greater than the window counter when the last feedback packet was sent. Thus, the feedback timer is not necessary when the window counter is used. However, the feedback timer still could be useful in some rare cases to prevent the sender from unnecessarily halving its sending rate. In particular, one could construct scenarios where the use of the feedback timer at the receiver would prevent the unnecessary expiration of the nofeedback timer at the sender. Consider the case below, in which a feedback packet is sent when a data packet arrives with a window counter of K. Window Counters: K K+1 K+2 K+3 K+4 K+5 K+6 ... K+15 K+16 K+17 ... | | | | | | | | | | Data | | | | | | | | | | Packets | | | | | | | | | | Received: - - --- - ... - - -- - -- -- - | | | | | | | | | | | | Events: 1: 2: 3: 4: 5: 6: "A" "B" Timer "B" sent sent received 1: Feedback message A is sent. 2: A feedback message would have been sent if feedback timers had been used.
3: Feedback message B is sent. 4: Sender's nofeedback timer expires. 5: Feedback message B is received at the sender. 6: Sender's nofeedback timer would have expired if feedback timers had been used, and the feedback message at 2 had been sent. The receiver receives data after the feedback packet has been sent but has received no data packets with a window counter between K+4 and K+14. A data packet with a window counter of K+4 or larger would have triggered sending a new feedback packet, but no feedback packet is sent until time 3. The TFRC protocol specifies that after a feedback packet is received, the sender sets a nofeedback timer to at least four times the round- trip time estimate. If the sender doesn't receive any feedback packets before the nofeedback timer expires, then the sender halves its sending rate. In the figure, the sender receives feedback message A (time 1) and then sets the nofeedback timer to expire roughly four round-trip times later (time 4). The sender starts sending again just before the nofeedback timer expires but doesn't receive the resulting feedback message until after its expiration, resulting in an unnecessary halving of the sending rate. If the connection had used feedback timers, the receiver would have sent a feedback message when the feedback timer expired at time 2, and the halving of the sending rate would have been avoided. For implementors who wish to implement a feedback timer for the data receiver, we suggest estimating the round-trip time from the most recent data packet, as described in Section 8.1. We note that this procedure does not work when the inter-packet sending times are greater than the RTT. RFC4340], and security considerations for TFRC have been discussed in [RFC3448], Section 9. The security considerations for TFRC include the need to protect against spoofed feedback and the need to protect the congestion control mechanisms against incorrect information from the receiver. In this document, we have extensively discussed the mechanisms the sender can use to verify the information sent by the receiver. When ECN is used, the receiver returns ECN Nonce information to the sender. When ECN is not used, then, as Section 9 shows, the sender could still use various techniques that might catch the receiver in
an error in reporting congestion, but this is not as robust or non- intrusive as the verification provided by the ECN Nonce. RFC4340]. CCID 3 also introduces three sets of numbers whose values should be allocated by IANA; namely, CCID 3-specific Reset Codes, option types, and feature numbers. These ranges will prevent any future CCID 3- specific allocations from polluting DCCP's corresponding global namespaces; see [RFC4340], Section 10.3. However, we note that this document makes no particular allocations from the Reset Code range, except for experimental and testing use [RFC3692]. We refer to the Standards Action policy outlined in [RFC2434]. Section 8. This document allocates option types 192-194, and option types 184-190 and 248-254 are permanently reserved for experimental and testing use. The remaining option types -- 128-183, 191, 195-247, and 255 -- are currently reserved and should be allocated with the Standards Action policy, which requires IESG review and approval and standards-track IETF RFC publication.
initially populated using the values in Table 2, in Section 8. This document allocates feature number 192, and feature numbers 184-190 and 248-254 are permanently reserved for experimental and testing use. The remaining feature numbers -- 128-183, 191, 193-247, and 255 -- are currently reserved and should be allocated with the Standards Action policy, which requires IESG review and approval and standards-track IETF RFC publication.
RFC3448] does not match the desires of possible users of DCCP. These include the following: 1. The initial sending rate of at most four packets per RTT, as specified in [RFC3390]. 2. The receiver's sending of an acknowledgement for every data packet received, when the receiver receives at a rate less than one packet per round-trip time. 3. The sender's limitation of at most doubling the sending rate from one round-trip time to the next (or, more specifically, of limiting the sending rate to at most twice the reported receive rate over the previous round-trip time). 4. The limitation of halving the allowed sending rate after an idle period of four round-trip times (possibly down to the initial sending rate of two to four packets per round-trip time). 5. The response function used in [RFC3448], Section 3.1, which does not closely match the behavior of TCP in environments with high packet drop rates [RFC3714]. One suggestion for higher initial sending rates is an initial sending rate of up to eight small packets per RTT, when the total packet size, including headers, is at most 4380 bytes. Because the packets would be rate-paced out over a round-trip time, instead of sent back-to-back as they would be in TCP, an initial sending rate of eight small packets per RTT with TFRC-based congestion control would be considerably milder than the impact of an initial window of eight small packets sent back-to-back in TCP. As Section 5.1 describes, the initial sending rate also serves as a lower bound for reductions of the allowed sending rate during an idle period. We note that with CCID 3, the sender is in slow-start in the beginning and responds promptly to the report of a packet loss or mark. However, in the absence of feedback from the receiver, the sender can maintain its old sending rate for up to four round-trip times. One possibility would be that for an initial window of eight small packets, the initial nofeedback timer would be set to two round-trip times instead of four, so that the sending rate would be reduced after two round-trips without feedback.
Research and engineering will be needed to investigate the pros and cons of modifying these limitations in order to allow larger initial sending rates, to send fewer acknowledgements when the data sending rate is low, to allow more abrupt changes in the sending rate, or to allow a higher sending rate after an idle period. [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, March 1997. [RFC2434] Narten, T. and H. Alvestrand, "Guidelines for Writing an IANA Considerations Section in RFCs", BCP 26, RFC 2434, October 1998. [RFC2581] Allman, M., Paxson, V., and W. Stevens, "TCP Congestion Control", RFC 2581, April 1999. [RFC3168] Ramakrishnan, K., Floyd, S., and D. Black, "The Addition of Explicit Congestion Notification (ECN) to IP", RFC 3168, September 2001. [RFC3390] Allman, M., Floyd, S., and C. Partridge, "Increasing TCP's Initial Window", RFC 3390, October 2002. [RFC3448] Handley, M., Floyd, S., Padhye, J., and J. Widmer, "TCP Friendly Rate Control (TFRC): Protocol Specification", RFC 3448, January 2003. [RFC3692] Narten, T., "Assigning Experimental and Testing Numbers Considered Useful", BCP 82, RFC 3692, January 2004. [RFC4340] Kohler, E., Handley, M., and S. Floyd, "Datagram Congestion Control Protocol (DCCP)", RFC 4340, March 2006. [RFC3540] Spring, N., Wetherall, D., and D. Ely, "Robust Explicit Congestion Notification (ECN) Signaling with Nonces", RFC 3540, June 2003.
[RFC3714] Floyd, S. and J. Kempf, "IAB Concerns Regarding Congestion Control for Voice Traffic in the Internet", RFC 3714, March 2004. [RFC4341] Floyd, S. and E. Kohler, "Profile for Datagram Congestion Control Protocol (DCCP) Congestion Control ID 2: TCP-like Congestion Control", RFC 4341, March 2006. [V03] Arun Venkataramani, August 2003. Citation for acknowledgement purposes only.
Full Copyright Statement Copyright (C) The Internet Society (2006). This document is subject to the rights, licenses and restrictions contained in BCP 78, and except as set forth therein, the authors retain all their rights. This document and the information contained herein are provided on an "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY AND THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. Intellectual Property The IETF takes no position regarding the validity or scope of any Intellectual Property Rights or other rights that might be claimed to pertain to the implementation or use of the technology described in this document or the extent to which any license under such rights might or might not be available; nor does it represent that it has made any independent effort to identify any such rights. Information on the procedures with respect to rights in RFC documents can be found in BCP 78 and BCP 79. Copies of IPR disclosures made to the IETF Secretariat and any assurances of licenses to be made available, or the result of an attempt made to obtain a general license or permission for the use of such proprietary rights by implementers or users of this specification can be obtained from the IETF on-line IPR repository at http://www.ietf.org/ipr. The IETF invites any interested party to bring to its attention any copyrights, patents or patent applications, or other proprietary rights that may cover technology that may be required to implement this standard. Please address the information to the IETF at ietf- firstname.lastname@example.org. Acknowledgement Funding for the RFC Editor function is provided by the IETF Administrative Support Activity (IASA).