| Internet-Draft | IP Traffic Flow Security | September 2022 |
| Hopps | Expires 8 March 2023 | [Page] |
This document describes a mechanism for aggregation and fragmentation of IP packets when they are being encapsulated in ESP payloads. This new payload type can be used for various purposes such as decreasing encapsulation overhead for small IP packets; however, the focus in this document is to enhance IPsec traffic flow security (IP-TFS) by adding Traffic Flow Confidentiality (TFC) to encrypted IP encapsulated traffic. TFC is provided by obscuring the size and frequency of IP traffic using a fixed-sized, constant-send-rate IPsec tunnel. The solution allows for congestion control as well as non-constant send-rate usage.¶
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.¶
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.¶
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."¶
This Internet-Draft will expire on 8 March 2023.¶
Copyright (c) 2022 IETF Trust and the persons identified as the document authors. All rights reserved.¶
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Revised BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Revised BSD License.¶
Traffic Analysis ([RFC4301], [AppCrypt]) is the act of extracting information about data being sent through a network. While directly obscuring the data with encryption [RFC4303], the patterns in the message traffic may expose information due to variations in its shape and timing ([RFC8546], [AppCrypt]). Hiding the size and frequency of traffic is referred to as Traffic Flow Confidentiality (TFC) per [RFC4303].¶
[RFC4303] provides for TFC by allowing padding to be added to encrypted IP packets and allowing for transmission of all-pad packets (indicated using protocol 59). This method has the major limitation that it can significantly under-utilize the available bandwidth.¶
This document defines an aggregation and fragmentation (AGGFRAG) mode for ESP, and its use for IP Traffic Flow Security (IP-TFS). This solution provides for full TFC without the aforementioned bandwidth limitation. This is accomplished by using a constant-send-rate IPsec [RFC4303] tunnel with fixed-sized encapsulating packets; however, these fixed-sized packets can contain partial, whole or multiple IP packets to maximize the bandwidth of the tunnel. A non-constant send-rate is allowed, but the confidentiality properties of its use are outside the scope of this document.¶
For a comparison of the overhead of IP-TFS with the RFC4303 prescribed TFC solution see Appendix C.¶
Additionally, IP-TFS provides for operating fairly within congested networks [RFC2914]. This is important for when the IP-TFS user is not in full control of the domain through which the IP-TFS tunnel path flows.¶
The mechanisms, such as the AGGFRAG mode, defined in this document are generic with the intent of allowing for non-TFS uses, but such uses are outside the scope of this document.¶
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here.¶
This document assumes familiarity with IP security concepts including TFC as described in [RFC4301].¶
As mentioned in Section 1, AGGFRAG mode utilizes an IPsec [RFC4303] tunnel as its transport. For the purpose of IP-TFS, fixed-sized encapsulating packets are sent at a constant rate on the AGGFRAG tunnel.¶
The primary input to the tunnel algorithm is the requested bandwidth to be used by the tunnel. Two values are then required to provide for this bandwidth use, the fixed size of the encapsulating packets, and rate at which to send them.¶
The fixed packet size MAY either be specified manually or be determined through other methods such as the Packetization Layer MTU Discovery (PLMTUD) ([RFC4821], [RFC8899]) or Path MTU discovery (PMTUD) ([RFC1191], [RFC8201]). PMTUD is known to have issues so PLMTUD is considered the more robust option. For PLMTUD, congestion control payloads can be used as in-band probes (see Section 6.1.2 and [RFC8899]).¶
Given the encapsulating packet size and the requested bandwidth to be used, the corresponding packet send rate can be calculated. The packet send rate is the requested bandwidth to be used divided by the size of the encapsulating packet.¶
The egress (receiving) side of the AGGFRAG tunnel MUST allow for and expect the ingress (sending) side of the AGGFRAG tunnel to vary the size and rate of sent encapsulating packets, unless constrained by other policy.¶
As previously mentioned, one issue with the TFC padding solution in [RFC4303] is the large amount of wasted bandwidth as only one IP packet can be sent per encapsulating packet. In order to maximize bandwidth, IP-TFS breaks this one-to-one association by introducing an AGGFRAG mode for ESP.¶
AGGFRAG mode aggregates as well as fragments the inner IP traffic flow into encapsulating IPsec tunnel packets. For IP-TFS, the IPsec encapsulating tunnel packets are a fixed size. Padding is only added to the tunnel packets if there is no data available to be sent at the time of tunnel packet transmission, or if fragmentation has been disabled by the receiver.¶
This is accomplished using a new Encapsulating Security Payload (ESP, [RFC4303]) Next Header field value AGGFRAG_PAYLOAD (Section 6.1).¶
Other non-IP-TFS uses of this AGGFRAG mode have been suggested, such as increased performance through packet aggregation, as well as handling MTU issues using fragmentation. These uses are not defined here, but are also not restricted by this document.¶
The AGGFRAG_PAYLOAD payload content defined in this document consists of a 4 or 24 octet header followed by either a partial datablock, a full datablock, or multiple partial or full datablocks. The following diagram illustrates this payload within the ESP packet. See Section 6.1 for the exact formats of the AGGFRAG_PAYLOAD payload.¶
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Outer Encapsulating Header ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ESP Header... . +---------------------------------------------------------------+ | [AGGFRAG sub-type/flags] : BlockOffset | +---------------------------------------------------------------+ : [Optional Congestion Info] : +---------------------------------------------------------------+ | DataBlocks ... ~ ~ ~ ~ | +---------------------------------------------------------------| . ESP Trailer... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The BlockOffset value is either zero or some offset into or past
the end of the DataBlocks data.¶
If the BlockOffset value is zero it means that the DataBlocks
data begins with a new data block.¶
Conversely, if the BlockOffset value is non-zero it points to the
start of the new data block, and the initial DataBlocks data
belongs to the data block that is still being re-assembled.¶
If the BlockOffset points past the end of the DataBlocks data
then the next data block occurs in a subsequent encapsulating packet.¶
Having the BlockOffset always point at the next available data
block allows for recovering the next inner packet in the
presence of outer encapsulating packet loss.¶
An example AGGFRAG mode packet flow can be found in Appendix A.¶
+---------------------------------------------------------------+ | Type | rest of IPv4, IPv6 or pad. +--------
A data block is defined by a 4-bit type code followed by the data
block data. The type values have been carefully chosen to coincide
with the IPv4/IPv6 version field values so that no per-data block
type overhead is required to encapsulate an IP packet. Likewise, the
length of the data block is extracted from the encapsulated IPv4's
Total Length or IPv6's Payload Length fields.¶
Since a data block's type is identified in its first 4-bits, the only
time padding is required is when there is no data to encapsulate. For
this end padding a Pad Data Block is used.¶
In order for a receiver to reassemble fragmented inner packets, the
sender MUST send the inner packet fragments back-to-back in the
logical outer packet stream (i.e., using consecutive ESP sequence
numbers). However, the sender is allowed to insert "all-pad" payloads
(i.e., payloads with a BlockOffset of zero and a single pad
DataBlock) in between the packets carrying the inner packet
fragment payloads. This interleaving of all-pad payloads allows the
sender to always send a tunnel packet, regardless of the
encapsulation computational requirements.¶
When a receiver is reassembling an inner packet, and it receives an "all-pad" payload, it increments the expected sequence number that the next inner packet fragment is expected to arrive in.¶
Given the above, the receiver will need to handle out-of-order arrival of outer ESP packets prior to reassembly processing. ESP already provides for optionally detecting replay attacks. Detecting replay attacks normally utilizes a window method. A similar sequence number based sliding window can be used to correct re-ordering of the outer packet stream. Receiving a larger (newer) sequence number packet advances the window, and received older ESP packets whose sequence numbers the window has passed by are dropped. A good choice for the size of this window depends on the amount of misordering the user is experiencing; however, a value of 3 has been suggested as a default when no more informed choice exists.¶
As the amount of misordering that may be present is hard to predict, the window size SHOULD be configurable by the user. Implementations MAY also dynamically adjust the reordering window based on actual misordering seen in arriving packets.¶
Please note, when IP-TFS sends a continuous stream of packets, there is no requirement for an explicit lost packet timer; however, using a lost packet timer is RECOMMENDED. If an implementation does not use a lost packet timer and only considers an outer packet lost when the reorder window moves by it, the inner traffic can be delayed by up to the reorder window size times the per packet send rate. This delay could be significant for slower send rates or when larger reorder window sizes are in use. As the lost packet timer affects delay of inner packet delivery, an implementation or user could choose to set it proportionate to the tunnel rate.¶
While ESP guarantees an increasing sequence number with subsequently sent packets, it does not actually require the sequence numbers to be generated consecutively (e.g., sending only even numbered sequence numbers would be allowed as long as they are always increasing). Gaps in the sequence numbers will not work for this document so the sequence number stream MUST increase monotonically by 1 for each subsequent packet.¶
When using the AGGFRAG_PAYLOAD in conjunction with replay detection, the window size for both MAY be reduced to the smaller of the two window sizes. This is because packets outside of the smaller window but inside the larger would still be dropped by the mechanism with the smaller window size. However, there is also no requirement to make these values the same. Indeed, in some cases, such as slow tunnels where a very small or zero reorder window size is appropriate, the user may still want a large replay detection window to log replayed packets. Additionally, large replay windows can be implemented with very little overhead compared to large reorder windows.¶
Finally, as sequence numbers are reset when switching SAs (e.g., when re-keying a child SA), senders MUST NOT send initial fragments of an inner packet using one SA and subsequent fragments in a different SA.¶
A note on BlockOffset values, senders MUST encode the BlockOffset
consistent with the immediately preceding non-all-pad payload packet.
Specifically, if the immediately preceding non-all-pad payload packet
ended with a Pad Data Block, this BlockOffset MUST be zero, as Pad
Data Blocks are never fragmented. The BlockOffset MUST be
consistent with the remaining size implied by the native length
encoding of the fragmented inner packet.¶
When the tunnel bandwidth is not being fully utilized, a sender MAY pad-out the current encapsulating packet in order to deliver an inner packet un-fragmented in the following outer packet. The benefit would be to avoid inner packet fragmentation in the presence of a bursty offered load (non-bursty traffic will naturally not fragment). Senders MAY also choose to allow for a minimum fragment size to be configured (e.g., as a percentage of the AGGFRAG_PAYLOAD payload size) to avoid fragmentation at the cost of tunnel bandwidth. The cost with these methods is complexity and added delay of inner traffic. The main advantage to avoiding fragmentation is to minimize inner packet loss in the presence of outer packet loss. When this is worthwhile (e.g., how much loss and what type of loss is required, given different inner traffic shapes and utilization, for this to make sense), and what values to use for the allowable/added delay may be worth researching but is outside the scope of this document.¶
While use of padding to avoid fragmentation does not impact interoperability, used inappropriately it can reduce the effective throughput of a tunnel. Senders implementing either of the above approaches will need to take care to not reduce the effective capacity, and overall utility, of the tunnel through the overuse of padding.¶
To support reporting of congestion control information (described later) using a non-AGGFRAG_PAYLOAD-enabled SA, it is allowed to send an AGGFRAG_PAYLOAD payload with no data blocks (i.e., the ESP payload length is equal to the AGGFRAG_PAYLOAD header length). This special payload is called an empty payload.¶
Currently this situation is only applicable in non-IKEv2 use cases.¶
[RFC4301] provides some direction on when and how to map various values from an inner IP header to the outer encapsulating header, namely the Don't-Fragment (DF) bit ([RFC0791] and [RFC8200]), the Differentiated Services (DS) field [RFC2474] and the Explicit Congestion Notification (ECN) field [RFC3168]. Unlike [RFC4301], AGGFRAG mode may and often will be encapsulating more than one IP packet per ESP packet. To deal with this, these mappings are restricted further.¶
AGGFRAG mode never maps the inner DF bit as it is unrelated to the AGGFRAG tunnel functionality; AGGFRAG mode never needs to IP fragment the inner packets and the inner packets will not affect the fragmentation of the outer encapsulation packets.¶
The ECN value need not be mapped as any congestion related to the constant-send-rate IP-TFS tunnel is unrelated (by design) to the inner traffic flow. The sender MAY still set the ECN value of inner packets based on the normal ECN specification [RFC3168], [RFC4301] and [RFC6040].¶
By default, the DS field SHOULD NOT be copied, although a sender MAY choose to allow for configuration to override this behavior. A sender SHOULD also allow the DS value to be set by configuration.¶
[RFC4301] specifies how to modify the inner packet IPv4 TTL [RFC0791] or IPv6 Hop Limit [RFC8200].¶
[RFC4301] also specifies how to apply policy to authenticated and unauthenticated ICMP error packets (e.g., Destination Unreachable) arriving at or being forwarded through the endpoint. In particular, whether to process, ignore or forward said packets. With one exception this document does not change the handling of these packets, they should be handled as specified in [RFC4301].¶
The one way in which an AGGFRAG tunnel differs in ICMP error packet mechanics is with PMTU. When fragmentation is enabled on the AGGFRAG tunnel, then no ICMP "too-big" errors need to be generated for arriving ingress traffic as the arriving inner packets will be naturally fragmented by the AGGFRAG encapsultation.¶
Otherwise, when fragmentation has been disabled on the AGGFRAG tunnel, then the treatment of arriving inner traffic exactly maps to that of a non-AGGFRAG ESP tunnel. Explicitly, IPv4 with DF set and IPv6 packets which cannot fit in it's own outer packet payload will generate the appropriate ICMP "too-big" error as directed by [RFC4301], and IPv4 packets without DF set will be IP fragmented as directed by [RFC4301].¶
Packets egressing the tunnel continue to be handled as specified in [RFC4301].¶
All other aspects of PMTU and the handling of ICMP "Too Big" messages (i.e., with regards to the outer AGGFRAG/ESP tunnel packet size) also remain unchanged from [RFC4301].¶
Unlike [RFC4301], there is normally no effective MTU (EMTU) on an AGGFRAG tunnel as all IP packet sizes are properly transmitted without requiring IP fragmentation prior to tunnel ingress. That said, a sender MAY allow for explicitly configuring an MTU for the tunnel.¶
If fragmentation has been disabled on the AGGFRAG tunnel, then the tunnel's EMTU and behaviors are the same as normal IPsec tunnels [RFC4301].¶
This document does not specify mixed use of an AGGFRAG_PAYLOAD-enabled SA. A sender MUST only send AGGFRAG_PAYLOAD payloads over an SA configured for AGGFRAG mode.¶
Just as with normal IPsec/ESP SAs, AGGFRAG SAs are unidirectional. Bidirectional IP-TFS functionality is achieved by setting up 2 AGGFRAG SAs, one in either direction.¶
An AGGFRAG tunnel used for IP-TFS can operate in 2 modes, a non-congestion-controlled mode and congestion-controlled mode.¶
In the non-congestion-controlled mode, IP-TFS sends fixed-sized packets over an AGGFRAG tunnel at a constant rate. The packet send rate is constant and is not automatically adjusted regardless of any network congestion (e.g., packet loss).¶
For similar reasons as given in [RFC7510] the non-congestion-controlled mode MUST only be used where the user has full administrative control over any path the tunnel will take, and MUST NOT be used if this is not the case. This is required so the user can guarantee the bandwidth and also be sure as to not be negatively affecting network congestion [RFC2914]. In this case, packet loss should be reported to the administrator (e.g., via syslog, YANG notification, SNMP traps, etc.) so that any failures due to a lack of bandwidth can be corrected. The use of circuit breakers is also RECOMMENDED (Section 2.4.2.1).¶
Users that choose the non-congestion-controlled mode need to understand that this mode will send packets at a constant rate utilizing a constant fixed bandwidth and will not adjust based on congestion. Thus, if they do not guarantee the bandwidth required by the tunnel, the tunnel's operation, as well as the rest of their network, may be negatively impacted.¶
One expected use case for non-congestion-controlled mode is to guarantee the full tunnel bandwidth is available and preferred over other non-tunnel traffic. In fact, a typical site-to-site use case might have all of the user traffic utilizing the IP-TFS tunnel.¶
Non-congestion-controlled mode is also appropriate if ESP over TCP is in use [RFC8229]. However, the use of TCP is considered a highly non-preferred, and a fall-back only solution for IPsec. This is also one of the reasons that TCP was not chosen as the encapsulation for IP-TFS instead of AGGFRAG.¶
With the congestion-controlled mode, IP-TFS adapts to network congestion by lowering the packet send rate to accommodate the congestion, as well as raising the rate when congestion subsides. Since overhead is per packet, by allowing for maximal fixed-size packets and varying the send rate, transport overhead is minimized.¶
The output of the congestion control algorithm will adjust the rate at which the ingress sends packets. While this document does not require a specific congestion control algorithm, best current practice RECOMMENDS that the algorithm conform to [RFC5348]. Congestion control principles are documented in [RFC2914] as well. [RFC4342] provides an example of the [RFC5348] algorithm which matches the requirements of IP-TFS (i.e., designed for fixed-size packets and send rate varied based on congestion).¶
The required inputs for the TCP friendly rate control algorithm described in [RFC5348] are the receiver's loss event rate and the sender's estimated round-trip time (RTT). These values are provided by IP-TFS using the congestion information header fields described in Section 3. In particular, these values are sufficient to implement the algorithm described in [RFC5348].¶
At a minimum, the congestion information MUST be sent, from the receiver and from the sender, at least once per RTT. Prior to establishing an RTT the information SHOULD be sent constantly from the sender and the receiver so that an RTT estimate can be established. Not receiving this information over multiple consecutive RTT intervals should be considered a congestion event that causes the sender to adjust its sending rate lower. For example, [RFC4342] calls this the "no feedback timeout" and it is equal to 4 RTT intervals. When a "no feedback timeout" has occurred [RFC4342] halves the sending rate.¶
An implementation MAY choose to always include the congestion
information in its AGGFRAG payload header if sending on an IP-TFS-enabled
SA. Since IP-TFS normally will operate with a large packet
size, the congestion information should represent a small portion of
the available tunnel bandwidth. An implementation choosing to always
send the data MAY also choose to only update the LossEventRate
and RTT header field values it sends every RTT though.¶
When choosing a congestion control algorithm (or a selection of algorithms), note that IP-TFS is not providing for reliable delivery of IP traffic, and so per packet ACKs are not required and are not provided.¶
It is worth noting that the variable send-rate of a congestion-controlled AGGFRAG tunnel, is not private; however, this send-rate is being driven by network congestion, and as long as the encapsulated (inner) traffic flow shape and timing are not directly affecting the (outer) network congestion, the variations in the tunnel rate will not weaken the provided inner traffic flow confidentiality.¶
In additional to congestion control, implementations that support non-congestion control mode SHOULD implement circuit breakers [RFC8084] as a recovery method of last resort. When circuit breakers are enabled an implementation SHOULD also enable congestion control reports so that circuit breakers have information to act on.¶
The pseudowire congestion considerations [RFC7893] are equally applicable to the mechanisms defined in this document, notably the text on inellastic traffic.¶
One example of a simple slow-trip circuit breaker (CB) an implementation may provide would utilize 2 values, the amount of persistent loss rate required to trip the CB, and the required length of time this persistent loss rate must be seen to trip the CB. These 2 value are required configuration from the user. When the CB is tripped the tunnel traffic is disabled, and an appropriate log message or other management type alarm is triggered indicating operate intervention is required.¶
An AGGFRAG-enabled SA receiver has a few tasks to perform.¶
The receiver MAY process incoming AGGFRAG_PAYLOAD payloads as soon as they arrive as much as it can. I.e., if the incoming AGGFRAG_PAYLOAD packet contains complete inner packet(s), the receiver should extract and transmit them immediately. For partial packets, the receiver needs to keep the partial packets in the memory until they fall out from the reordering window, or until the missing parts of the packets are received, in which case it will reassemble and transmit them. If the AGGFRAG_PAYLOAD payload contains multiple packets they SHOULD be sent out in the order they are in the AGGFRAG_PAYLOAD (i.e., keep the original order they were received on the other end). The cost of using this method is that an amplification of out-of-order delivery of inner packets can occur due to inner packet aggregation.¶
Instead of the method described in the previous paragraph, the receiver MAY reorder out-of-order AGGFRAG_PAYLOAD payloads received into in-sequence-order AGGFRAG_PAYLOAD payloads (Section 2.2.3), and only after it has an in-order AGGFRAG_PAYLOAD payload stream would the receiver transmits the inner packets. Using this method will ensure the inner packets are sent in order. The cost of this method is that a lost packet will cause a delay of up to the lost packet timer interval (or the full reorder window if no lost packet timer is used). Additionally, there can be extra burstiness in the output stream. This burstiness can happen when a lost packet is dropped from the re-order window, and the remaining outer packets in the re-order window are immediately processed and sent out back to back.¶
Additionally, if congestion control is enabled, the receiver sends congestion control data (Section 6.1.2) back to the sender as described in Section 2.4.2 and Section 3.¶
Finally, a note on receiving incorrect BlockOffset values. To account
for misbehaving senders, a receiver SHOULD gracefully handle the case
where the BlockOffset of consecutive packets, and/or the inner
packet they share, do not agree. It MAY drop the inner packet, or one
or both of the outer packets.¶
In order to support the congestion-controlled mode, the sender needs to know the loss event rate and to approximate the RTT [RFC5348]. In order to obtain these values, the receiver sends congestion control information on its SA back to the sender. Thus, to support congestion control the receiver MUST have a paired SA back to the sender (this is always the case when the tunnel was created using IKEv2). If the SA back to the sender is a non-AGGFRAG_PAYLOAD enabled SA then an AGGFRAG_PAYLOAD empty payload (i.e., header only) is used to convey the information.¶
In order to calculate a loss event rate compatible with [RFC5348], the
receiver needs to have a round-trip time estimate. Thus the sender
communicates this estimate in the RTT header field. On startup this
value will be zero as no RTT estimate is yet known.¶
In order for the sender to estimate its RTT value, the sender
places a timestamp value in the TVal header field. On first receipt
of this TVal, the receiver records the new TVal value along with
the time it arrived locally. Subsequent receipt of the same TVal
MUST NOT update the recorded time.¶
When the receiver sends its congestion control header it places this latest recorded
TVal in the TEcho header field, along with 2 delay values, Echo
Delay and Transmit Delay. The Echo Delay value is the time delta
from the recorded arrival time of TVal and the current clock in
microseconds. The second value, Transmit Delay, is the receiver's
current transmission delay on the tunnel (i.e., the average time
between sending packets on its half of the AGGFRAG tunnel).¶
When the sender receives back its TVal in the TEcho header field
it calculates 2 RTT estimates. The first is the actual delay found by
subtracting the TEcho value from its current clock and then
subtracting Echo Delay as well. The second RTT estimate is found by
adding the received Transmit Delay header value to the sender's own
transmission delay (i.e., the average time between sending packets on
its half of the AGGFRAG tunnel). The larger of these 2 RTT estimates
SHOULD be used as the RTT value.¶
The two RTT estimates are required to handle different combinations of
faster or slower tunnel packet paths with faster or slower fixed
tunnel rates. Choosing the larger of the two values guarantees that
the RTT is never considered faster than the aggregate transmission
delay based on the IP-TFS send rate (the second estimate), as well
as never being considered faster than the actual RTT along the tunnel
packet path (the first estimate).¶
The receiver also calculates, and communicates in the LossEventRate
header field, the loss event rate for use by the sender. This is
slightly different from [RFC4342] which periodically sends all the loss
interval data back to the sender so that it can do the calculation.
See Appendix B for a suggested way to
calculate the loss event rate value. Initially this value will be
zero (indicating no loss) until enough data has been collected by the
receiver to update it.¶
In addition to normal packet loss information, AGGFRAG mode supports use
of the ECN bits in the encapsulating IP header [RFC3168] for
identifying congestion. If ECN use is enabled and a packet arrives at
the egress (receiving) side with the Congestion Experienced (CE) value set,
then the receiver considers that packet as being dropped, although it
does not drop it. The receiver MUST set the E bit in any
AGGFRAG_PAYLOAD payload header containing a LossEventRate value
derived from a CE value being considered.¶
[RFC3168] and [RFC4301], updated by [RFC6040] defines behaviors for marking the outer ECN field value based on the ECN field of the inner packet. As AGGFRAG mode may have multiple inner packets present in a single outer packet, and there is no obvious correct way to map these multiple values to the single outer packet ECN field value, the tunnel ingress endpoint SHOULD operate in the "compatibility" mode rather than the "default" mode from RFC6040. In particular this means that the ingress (sending) endpoint of the tunnel always sets the newly constructed outer encapsulating packet header ECN field to Not-ECT [RFC6040].¶
IP-TFS is meant to be deployable with a minimal amount of configuration. All IP-TFS specific configuration should be specified at the unidirectional tunnel ingress (sending) side. It is intended that non-IKEv2 operation is supported, at least, with local static configuration.¶
YANG and MIB documents have been defined for IP-TFS in [I-D.ietf-ipsecme-yang-iptfs] and [I-D.ietf-ipsecme-mib-iptfs].¶
Bandwidth is a local configuration option. For non-congestion-controlled mode, the bandwidth SHOULD be configured. For congestion-controlled mode, the bandwidth can be configured or the congestion control algorithm discovers and uses the maximum bandwidth available. No standardized configuration method is required.¶
The fixed packet size to be used for the tunnel encapsulation packets MAY be configured manually or can be automatically determined using other methods such as PLMTUD ([RFC4821], [RFC8899]) or PMTUD ([RFC1191], [RFC8201]). As PMTUD is known to have issues, PLMTUD is considered the more robust option. No standardized configuration method is required.¶
Congestion control is a local configuration option. No standardized configuration method is required.¶
As mentioned previously AGGFRAG tunnels utilize ESP payloads of type AGGFRAG_PAYLOAD.¶
When using IKEv2, a new "USE_AGGFRAG" Notification Message enables the AGGFRAG_PAYLOAD payload on a child SA pair. The method used is similar to how USE_TRANSPORT_MODE is negotiated, as described in [RFC7296].¶
To request use of the AGGFRAG_PAYLOAD payload on the Child SA pair, the initiator includes the USE_AGGFRAG notification in an SA payload requesting a new Child SA (either during the initial IKE_AUTH or during CREATE_CHILD_SA exchanges). If the request is accepted then the response MUST also include a notification of type USE_AGGFRAG. If the responder declines the request the child SA will be established without AGGFRAG_PAYLOAD payload use enabled. If this is unacceptable to the initiator, the initiator MUST delete the child SA.¶
As the use of the AGGFRAG_PAYLOAD payload is currently only defined for non-transport mode tunnels, the USE_AGGFRAG notification MUST NOT be combined with USE_TRANSPORT notification.¶
The USE_AGGFRAG notification contains a 1 octet payload of flags that specify requirements from the sender of the notification. If any requirement flags are not understood or cannot be supported by the receiver then the receiver SHOULD NOT enable use of AGGFRAG_PAYLOAD (either by not responding with the USE_AGGFRAG notification, or in the case of the initiator, by deleting the child SA if the now established non-AGGFRAG_PAYLOAD using SA is unacceptable).¶
The notification type and payload flag values are defined in Section 6.1.4.¶
The packet and data formats defined below are generic with the intent of allowing for non-IP-TFS uses, but such uses are outside the scope of this document.¶
ESP Next Header value: 144¶
An AGGFRAG payload is identified by the ESP Next Header value AGGFRAG_PAYLOAD which has the value 144, which has been reserved in the IP protocol numbers space. The first octet of the payload indicates the format of the remaining payload data.¶
0 1 2 3 4 5 6 7 +-+-+-+-+-+-+-+-+-+-+- | Sub-type | ... +-+-+-+-+-+-+-+-+-+-+-
This document defines 2 payload sub-types. These payload formats are defined in the following sections.¶
The non-congestion control AGGFRAG_PAYLOAD payload consists of a
4-octet header followed by a variable amount of DataBlocks data as
shown below.¶
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
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Sub-Type (0) | Reserved | BlockOffset |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| DataBlocks ...
+-+-+-+-+-+-+-+-+-+-+-
DataBlocks data before the start of a
new data block. If the start of a new data block
occurs in a subsequent payload the BlockOffset
will point past the end of the DataBlocks data.
In this case all the DataBlocks data belongs to
the current data block being assembled. When the
BlockOffset extends into subsequent payloads it
continues to only count DataBlocks data (i.e.,
it does not count subsequent packets
non-DataBlocks data such as header octets).¶
The congestion control AGGFRAG_PAYLOAD payload consists of a 24
octet header followed by a variable amount of DataBlocks data as
shown below.¶
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
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Sub-type (1) | Reserved |P|E| BlockOffset |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| LossEventRate |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| RTT | Echo Delay ...
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
... Echo Delay | Transmit Delay |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| TVal |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| TEcho |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| DataBlocks ...
+-+-+-+-+-+-+-+-+-+-+-
LossEventRate.¶
1/LossEventRate.¶
0x3FFFFF the value MUST be set to 0x3FFFFF.¶
TVal
value which it is sending back in TEcho. If the delay
is equal to or larger than 0x1FFFFF the value MUST be
set to 0x1FFFFF.¶
0x1FFFFF the
value MUST be set to 0x1FFFFF.¶
TEcho field, along with
an Echo Delay value of how long that echo took.¶
TVal
field. The received TVal is placed in TEcho along with
an Echo Delay value indicating how long it has been since
receiving the TVal value.¶
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
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Type | IPv4, IPv6 or pad...
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-
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
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| 0x4 | IHL | TypeOfService | TotalLength |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Rest of the inner packet ...
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-
These values are the actual values within the encapsulated IPv4 header. In other words, the start of this data block is the start of the encapsulated IP packet.¶
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
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| 0x6 | TrafficClass | FlowLabel |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| PayloadLength | Rest of the inner packet ...
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-
These values are the actual values within the encapsulated IPv6 header. In other words, the start of this data block is the start of the encapsulated IP packet.¶
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
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| 0x0 | Padding ...
+-+-+-+-+-+-+-+-+-+-+-
As discussed in Section 5.1, a notification message USE_AGGFRAG is used to negotiate use of the ESP AGGFRAG_PAYLOAD Next Header value.¶
The USE_AGGFRAG Notification Message State Type is 16442¶
The notification payload contains 1 octet of requirement flags. There are currently 2 requirement flags defined. This may be revised by later specifications.¶
+-+-+-+-+-+-+-+-+ |0|0|0|0|0|0|C|D| +-+-+-+-+-+-+-+-+
Data Block). This value only
applies to what the sender is capable of receiving; the sender MAY
still send packet fragments unless similarly restricted by the
receiver in its USE_AGGFRAG notification.¶
Per the INT area directors direction, this document requests IANA allocate an IP protocol number from "Protocol Numbers - Assigned Internet Protocol Numbers" registry¶
This document requests IANA create a registry called "AGGFRAG_PAYLOAD Sub-Type Registry" under a new category named "ESP AGGFRAG_PAYLOAD Parameters". The registration policy for this registry is "Expert Review" ([RFC8126] and [RFC7120]).¶
This initial content for this registry is as follows:¶
Sub-Type Name Reference
--------------------------------------------------------
0 Non-Congestion Control Format This document
1 Congestion Control Format This document
3-255 Reserved
¶
This document requests a status type USE_AGGFRAG be allocated from the "IKEv2 Notify Message Types - Status Types" registry.¶
[ RFC Ed.: please remove this entire section as well as the reference to RFC7942 prior to publication. ]¶
[Section added during IESG review to help with evaluation]¶
This section records the status of known implementations of the protocol defined by this specification at the time of posting of this Internet-Draft, and is based on a proposal described in [RFC7942]. The description of implementations in this section is intended to assist the IETF in its decision processes in progressing drafts to RFCs. Please note that the listing of any individual implementation here does not imply endorsement by the IETF. This is not intended as, and must not be construed to be, a catalog of available implementations or their features. Readers are advised to note that other implementations may exist.¶
According to RFC 7942, "this will allow reviewers and working groups to assign due consideration to documents that have the benefit of running code, which may serve as evidence of valuable experimentation and feedback that have made the implemented protocols more mature. It is up to the individual working groups to use this information as they see fit".¶
Currently the author and contributors are aware of 1 full and completed implementation and 1 underway implementation of IP-TFS as defined in this document. These 2 are described below.¶
The entire IP-TFS protocol including congestion control mode has been implemented in VPP (Vector Packet Processor), and published to github with an Open Source (Apache 2) License. VPP is a highly efficient forwarding plane implemented in user-space utlizing direct control and polling of physical devices to provide high speed low-latency forwarding in Linux. By pinning packet processing threads directly to CPU cores for their exclusive use a high degree of control is given to the protocol designer.¶
The IKEv2 additions were implemented in Strongswan and are licensed using the GNU public license used by the Strongswan project.¶
Finally, an extensive automation suite was also created and is included with the open source implementation, which tests the functionality as well as the performance of the implementation, and most importantly verifies, through precise timing tracing and time-stamping, the decoupling of the users offered load from the tunnel packets (i.e., the Traffic Flow Security).¶
The verification process utilized the TREX packet generator for high bandwidth testing as well as other tools such as iperf. The test hardware included large servers with 10GE, 40GE and 100GE network interfaces, as well as small SoC (system on a chip) network appliances, and also cloud deployments.¶
Tested IP-TFS tunnel rates ranged from 10M all the way to 10GE on the small network appliance, for the large servers multiple 10GE tunnel rates were tested as well.¶
Offered loads included partial, full and oversubscribed bandwidths from various flow types consisting of small packets, large packets, random sized packets, sequential sized packets, and multiple IMIX variations sized flows. Timing analysis was done with variable rate traffic, impulse traffic and random bursty traffic.¶
The quality of the reference implementation should be considered production level as it underwent extensive testing and verification.¶
The organization responsible for this implementation is LabN Consulting, L.L.C.¶
URLs to the implementation follow.¶
The implementation was last updated April, 2021.¶
A second open source implementation has begun by LabN Consulting L.L.C., within the Linux IPsec xfrm stack. Development has also been coordinated with the Linux IPsec community, and was being worked by the same during the most recent IETF 114 hackathon.¶
Currently the quality is alpha level with aggregation-only complete and fragmentation support underway with congestion control to follow.¶
This implementation is licensed under the GNU public license and can be found at the following URLs¶
This document describes an aggregation and fragmentation mechanism to efficiently implement TFC for IP traffic. This approach is expected to reduce the efficacy of traffic analysis on IPsec communication. Other than the additional security afforded by using this mechanism, IP-TFS utilizes the security protocols [RFC4303] and [RFC7296] and so their security considerations apply to IP-TFS as well.¶
As noted in Section 3.1, the ECN bits are not protected by IPsec and thus may constitute a covert channel. For this reason, ECN use SHOULD NOT be enabled by default.¶
As noted previously in Section 2.4.2, for TFC to be maintained, the encapsulated traffic flow should not be affecting network congestion in a predictable way, and if it would be, then non-congestion-controlled mode use should be considered instead.¶
Below, an example inner IP packet flow within the encapsulating tunnel packet stream is shown. Notice how encapsulated IP packets can start and end anywhere, and more than one or less than 1 may occur in a single encapsulating packet.¶
Offset: 0 Offset: 100 Offset: 2000 Offset: 600 [ ESP1 (1404) ][ ESP2 (1404) ][ ESP3 (1404) ][ ESP4 (1404) ] [--750--][--750--][60][-240-][--3000----------------------][pad]
Each outer encapsulating ESPupayload space is a fixed-size of 1404 octets the first 4 octets of which contains the AGGFRAG header. The encapsulated IP packet flow (lengths include IP header and payload) is as follows: a 750-octet packet, a 750-octet packet, a 60-octet packet, a 240-octet packet, a 3000-octet packet.¶
The BlockOffset values in the 4 AGGFRAG payload headers for this
packet flow would thus be: 0, 100, 2000, 600 respectively. The first
encapsulating packet (ESP1) has a zero BlockOffset which points at the
IP data block immediately following the AGGFRAG header. The following
packet's (ESP2) BlockOffset points inward 100 octets to the start of the
60-octet data block. The third encapsulating packet (ESP3) contains the
middle portion of the 3000-octet data block so the offset points past
its end and into the fourth encapsulating packet. The fourth packet's
(ESP4) offset is 600, pointing at the padding which follows the
completion of the continued 3000-octet packet.¶
The current best practice indicates that congestion control SHOULD be done in a TCP-friendly way. A TCP-friendly congestion control algorithm is described in [RFC5348]. For this IP-TFS use case (as with [RFC4342]), the (fixed) packet size is used as the segment size for the algorithm. The main formula in the algorithm for the send rate is then as follows:¶
1
X = -----------------------------------------------
R * (sqrt(2*p/3) + 12*sqrt(3*p/8)*p*(1+32*p^2))
¶
Where X is the send rate in packets per second, R is the
round trip time estimate and p is the loss event rate (the inverse
of which is provided by the receiver).¶
In addition, the algorithm in [RFC5348] also uses an X_recv value (the
receiver's receive rate). For IP-TFS one MAY set this value according to
the sender's current tunnel send-rate (X).¶
The IP-TFS receiver, having the RTT estimate from the sender can use the
same method as described in [RFC5348] and [RFC4342] to collect the loss
intervals and calculate the loss event rate value using the weighted
average as indicated. The receiver communicates the inverse of this
value back to the sender in the AGGFRAG_PAYLOAD payload header field
LossEventRate.¶
The IP-TFS sender now has both the R and p values and can calculate
the correct sending rate. If following [RFC5348], the sender should also
use the slow start mechanism described therein when the IP-TFS SA is
first established.¶
For comparing overhead, the overhead of ESP for both normal and AGGFRAG tunnel packets must be calculated, and so an algorithm for encryption and authentication must be chosen. For the data below AES-GCM-256 was selected. This leads to an IP+ESP overhead of 54.¶
54 = 20 (IP) + 8 (ESPH) + 2 (ESPF) + 8 (IV) + 16 (ICV)¶
Additionally, for IP-TFS, non-congestion control AGGFRAG_PAYLOAD headers were chosen which adds 4 octets for a total overhead of 58.¶
For comparison, the overhead of an AGGFRAG payload is 58 octets per outer packet. Therefore, the octet overhead per inner packet is 58 divided by the number of outer packets required (fractions allowed). The overhead as a percentage of inner packet size is a constant based on the Outer MTU size.¶
OH = 58 / Outer Payload Size / Inner Packet Size OH % of Inner Packet Size = 100 * OH / Inner Packet Size OH % of Inner Packet Size = 5800 / Outer Payload Size¶
Type IP-TFS IP-TFS IP-TFS
MTU 576 1500 9000
PSize 518 1442 8942
-------------------------------
40 11.20% 4.02% 0.65%
576 11.20% 4.02% 0.65%
1500 11.20% 4.02% 0.65%
9000 11.20% 4.02% 0.65%
The overhead per inner packet for constant-send-rate padded ESP (i.e., traditional IPsec TFC) is 36 octets plus any padding, unless fragmentation is required.¶
When fragmentation of the inner packet is required to fit in the outer IPsec packet, overhead is the number of outer packets required to carry the fragmented inner packet times both the inner IP overhead (20) and the outer packet overhead (54) minus the initial inner IP overhead plus any required tail padding in the last encapsulation packet. The required tail padding is the number of required packets times the difference of the Outer Payload Size and the IP Overhead minus the Inner Payload Size. So:¶
Inner Payload Size = IP Packet Size - IP Overhead
Outer Payload Size = MTU - IPsec Overhead
Inner Payload Size
NF0 = ----------------------------------
Outer Payload Size - IP Overhead
NF = CEILING(NF0)
OH = NF * (IP Overhead + IPsec Overhead)
- IP Overhead
+ NF * (Outer Payload Size - IP Overhead)
- Inner Payload Size
OH = NF * (IPsec Overhead + Outer Payload Size)
- (IP Overhead + Inner Payload Size)
OH = NF * (IPsec Overhead + Outer Payload Size)
- Inner Packet Size
¶
The following tables collect the overhead values for some common L3 MTU sizes in order to compare them. The first table is the number of octets of overhead for a given L3 MTU sized packet. The second table is the percentage of overhead in the same MTU sized packet.¶
Type ESP+Pad ESP+Pad ESP+Pad IP-TFS IP-TFS IP-TFS
L3 MTU 576 1500 9000 576 1500 9000
PSize 522 1446 8946 518 1442 8942
-----------------------------------------------------------
40 482 1406 8906 4.5 1.6 0.3
128 394 1318 8818 14.3 5.1 0.8
256 266 1190 8690 28.7 10.3 1.7
518 4 928 8428 58.0 20.8 3.4
576 576 870 8370 64.5 23.2 3.7
1442 286 4 7504 161.5 58.0 9.4
1500 228 1500 7446 168.0 60.3 9.7
8942 1426 1558 4 1001.2 359.7 58.0
9000 1368 1500 9000 1007.7 362.0 58.4
Type ESP+Pad ESP+Pad ESP+Pad IP-TFS IP-TFS IP-TFS
MTU 576 1500 9000 576 1500 9000
PSize 522 1446 8946 518 1442 8942
-----------------------------------------------------------
40 1205.0% 3515.0% 22265.0% 11.20% 4.02% 0.65%
128 307.8% 1029.7% 6889.1% 11.20% 4.02% 0.65%
256 103.9% 464.8% 3394.5% 11.20% 4.02% 0.65%
518 0.8% 179.2% 1627.0% 11.20% 4.02% 0.65%
576 100.0% 151.0% 1453.1% 11.20% 4.02% 0.65%
1442 19.8% 0.3% 520.4% 11.20% 4.02% 0.65%
1500 15.2% 100.0% 496.4% 11.20% 4.02% 0.65%
8942 15.9% 17.4% 0.0% 11.20% 4.02% 0.65%
9000 15.2% 16.7% 100.0% 11.20% 4.02% 0.65%
Another way to compare the two solutions is to look at the amount of available bandwidth each solution provides. The following sections consider and compare the percentage of available bandwidth. For the sake of providing a well-understood baseline normal (unencrypted) Ethernet as well as normal ESP values are included.¶
In order to calculate the available bandwidth the per packet overhead is calculated first. The total overhead of Ethernet is 14+4 octets of header and CRC plus an additional 20 octets of framing (preamble, start, and inter-packet gap), for a total of 38 octets. Additionally, the minimum payload is 46 octets.¶
Size E + P E + P E + P IPTFS IPTFS IPTFS Enet ESP
MTU 590 1514 9014 590 1514 9014 any any
OH 92 92 92 96 96 96 38 74
------------------------------------------------------------
40 614 1538 9038 47 42 40 84 114
128 614 1538 9038 151 136 129 166 202
256 614 1538 9038 303 273 258 294 330
518 614 1538 9038 614 552 523 574 610
576 1228 1538 9038 682 614 582 614 650
1442 1842 1538 9038 1709 1538 1457 1498 1534
1500 1842 3076 9038 1777 1599 1516 1538 1574
8942 11052 10766 9038 10599 9537 9038 8998 9034
9000 11052 10766 18076 10667 9599 9096 9038 9074
Size E + P E + P E + P IPTFS IPTFS IPTFS Enet ESP
MTU 590 1514 9014 590 1514 9014 any any
OH 92 92 92 96 96 96 38 74
--------------------------------------------------------------
40 2.0M 0.8M 0.1M 26.4M 29.3M 30.9M 14.9M 11.0M
128 2.0M 0.8M 0.1M 8.2M 9.2M 9.7M 7.5M 6.2M
256 2.0M 0.8M 0.1M 4.1M 4.6M 4.8M 4.3M 3.8M
518 2.0M 0.8M 0.1M 2.0M 2.3M 2.4M 2.2M 2.1M
576 1.0M 0.8M 0.1M 1.8M 2.0M 2.1M 2.0M 1.9M
1442 678K 812K 138K 731K 812K 857K 844K 824K
1500 678K 406K 138K 703K 781K 824K 812K 794K
8942 113K 116K 138K 117K 131K 138K 139K 138K
9000 113K 116K 69K 117K 130K 137K 138K 137K
Size E + P E + P E + P IPTFS IPTFS IPTFS Enet ESP
590 1514 9014 590 1514 9014 any any
92 92 92 96 96 96 38 74
----------------------------------------------------------------------
40 6.51% 2.60% 0.44% 84.36% 93.76% 98.94% 47.62% 35.09%
128 20.85% 8.32% 1.42% 84.36% 93.76% 98.94% 77.11% 63.37%
256 41.69% 16.64% 2.83% 84.36% 93.76% 98.94% 87.07% 77.58%
518 84.36% 33.68% 5.73% 84.36% 93.76% 98.94% 93.17% 87.50%
576 46.91% 37.45% 6.37% 84.36% 93.76% 98.94% 93.81% 88.62%
1442 78.28% 93.76% 15.95% 84.36% 93.76% 98.94% 97.43% 95.12%
1500 81.43% 48.76% 16.60% 84.36% 93.76% 98.94% 97.53% 95.30%
8942 80.91% 83.06% 98.94% 84.36% 93.76% 98.94% 99.58% 99.18%
9000 81.43% 83.60% 49.79% 84.36% 93.76% 98.94% 99.58% 99.18%
A sometimes unexpected result of using an AGGFRAG tunnel (or any packet aggregating tunnel) is that, for small- to medium-sized packets, the available bandwidth is actually greater than native Ethernet. This is due to the reduction in Ethernet framing overhead. This increased bandwidth is paid for with an increase in latency. This latency is the time to send the unrelated octets in the outer tunnel frame. The following table illustrates the latency for some common values on a 10G Ethernet link. The table also includes latency introduced by padding if using ESP with padding.¶
ESP+Pad ESP+Pad IP-TFS IP-TFS
1500 9000 1500 9000
------------------------------------------
40 1.12 us 7.12 us 1.17 us 7.17 us
128 1.05 us 7.05 us 1.10 us 7.10 us
256 0.95 us 6.95 us 1.00 us 7.00 us
518 0.74 us 6.74 us 0.79 us 6.79 us
576 0.70 us 6.70 us 0.74 us 6.74 us
1442 0.00 us 6.00 us 0.05 us 6.05 us
1500 1.20 us 5.96 us 0.00 us 6.00 us
Notice that the latency values are very similar between the two solutions; however, whereas IP-TFS provides for constant high bandwidth, in some cases even exceeding native Ethernet, ESP with padding often greatly reduces available bandwidth.¶
We would like to thank Don Fedyk for help in reviewing and editing this work. We would also like to thank Michael Richardson, Sean Turner, Valery Smyslov and Tero Kivinen for reviews and many suggestions for improvements, as well as Joseph Touch for the transport area review and suggested improvements.¶
The following people made significant contributions to this document.¶
Lou Berger LabN Consulting, L.L.C. Email: lberger@labn.net¶