Multipath TCP Working Group M. Amend Internet-Draft Deutsche Telekom Intended status: Experimental J. Kang Expires: May 7, 2020 Huawei November 4, 2019 Multipath TCP Extension for Robust Session Establishment draft-amend-mptcp-robe-01 Abstract Multipath TCP extends the plain, single-path limited, TCP towards the capability of multipath transmission. This greatly improves the reliability and performance of TCP communication. For backwards compatibility reasons the Multipath TCP was designed to setup successfully an initial path first, after which subsequent paths can be added for multipath transmission. For that reason the Multipath TCP has the same limitations as the plain TCP during connection setup, in case the selected path is not functional. This document proposes a set of implementations and possible combinations thereof, that provide a more Robust Establishment (RobE) of MPTCP sessions. It includes RobE_TIMER, RobE_SIM, RobE_eSIM and RobE_IPS. RobE_TIMER is designed to stay close to MPTCP in that standard functionality is used wherever possible. Resiliency against network outages is achieved by modifying the SYN retransmission timer: If one path is defective, another path is used. RobE_SIM and RobE_eSIM provides the ability to simultaneously use multiple paths for connection setup. They ensure connectivity if at least one functional path out of a bunch of paths is given and offers beside that the opportunity to significantly improve loading times of Internet services. RobE_IPS provides a heuristic to select properly an initial path for connection establishment with a remote host based on empirical data derived from previous connection information. In practice, these independent solutions can be complementary used. This document also presents the design and protocol procedure for those combinations in addition to the respective stand-alone solutions. Amend & Kang Expires May 7, 2020 [Page 1] Internet-Draft MPTCP RobE November 2019 Status of This Memo 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 May 7, 2020. Copyright Notice Copyright (c) 2019 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 Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License. Table of Contents 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 1.1. Terminology . . . . . . . . . . . . . . . . . . . . . . . 6 2. Implementation without MPTCP protocol adaptation . . . . . . 7 2.1. Re-transmission Timer(RobE_TIMER) . . . . . . . . . . . . 7 2.2. Simultaneous Initial Paths Simple Version (RobE_SIM) . . 8 2.3. Heuristic Initial Path Selection (RobE_IPS) . . . . . . . 9 2.3.1. Architecture . . . . . . . . . . . . . . . . . . . . 9 2.3.2. Typical Scenarios . . . . . . . . . . . . . . . . . . 10 2.3.3. Path decision information . . . . . . . . . . . . . . 13 2.3.4. Initial Path Selection use local RTT information . . 14 2.4. Combination of RobE_SIM and RobE_IPS . . . . . . . . . . 14 2.5. Combination of RobE_TIMER and RobE_IPS . . . . . . . . . 15 3. Implementation with Bi-directional MPTCP Support . . . . . . 16 3.1. Simultaneous Initial Paths Extended Version (RobE_eSIM) . 16 Amend & Kang Expires May 7, 2020 [Page 2] Internet-Draft MPTCP RobE November 2019 3.1.1. RobE_eSIM implicit Negotiation and Procedure . . . . 17 3.1.2. RobE_eSIM explicit Negotiation and Procedure . . . . 18 3.1.3. Protocol Adaptation . . . . . . . . . . . . . . . . . 19 3.1.4. Fallback Mechanisms . . . . . . . . . . . . . . . . . 20 3.1.5. Comparison Robe_SIM and RobE_eSIM . . . . . . . . . . 22 3.1.6. Security Consideration . . . . . . . . . . . . . . . 23 3.2. Heuristic Initial Path Selection with remote RTT Measurement . . . . . . . . . . . . . . . . . . . . . . . 23 3.2.1. Description . . . . . . . . . . . . . . . . . . . . . 23 3.2.2. Protocol Adaptation . . . . . . . . . . . . . . . . . 24 3.2.3. Fallback Mechanism . . . . . . . . . . . . . . . . . 25 3.2.4. Security Consideration . . . . . . . . . . . . . . . 25 4. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 25 5. Informative References . . . . . . . . . . . . . . . . . . . 26 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 26 1. Introduction Multipath TCP Robust Session Establishment (MPTCP RobE) is a set of extensions to regular MPTCP [RFC6824] and its next upcoming version [I-D.ietf-mptcp-rfc6824bis], which releases single path limitations during the initial connection setup. Several scenarios require and benefit from a reliable and in time connection setup which is not covered by [RFC6824] and [I-D.ietf-mptcp-rfc6824bis] so far. MPTCP was designed to be compliant with the TCP standard [RFC0793] and introduced therefore the concept of an initial TCP flow while adding subsequent flows after successful multipath negotiation on the initial path. While fulfilling its purpose, MPTCP is however fully dependent on the transmission characteristics of the communication link selected for initiating MPTCP. Figure 1 shows the traditional way of MPTCP handshaking with a MP_CAPABLE exchanged first, followed when successful negotiated by additional flows engaging MP_JOIN. [RFC6824] and the upcoming next MPTCP [I-D.ietf-mptcp-rfc6824bis] differ in that a Key-A is sent with the first MP_CAPABLE or not. Amend & Kang Expires May 7, 2020 [Page 3] Internet-Draft MPTCP RobE November 2019 Host A Host B ------------------------ ---------- Address A1 Address A2 Address B1 ---------- ---------- ---------- | | | | SYN + MP_CAPABLE(Key-A[*]) | |--------------------------------------------->| |<---------------------------------------------| | SYN/ACK + MP_CAPABLE(Key-B) | | | | | ACK + MP_CAPABLE(Key-A, Key-B) | |--------------------------------------------->| | | | | | SYN + MP_JOIN(Token-B, R-A) | | |------------------------------->| | |<-------------------------------| | | SYN/ACK + MP_JOIN(HMAC-B, R-B) | | | | | | ACK + MP_JOIN(HMAC-A) | | |------------------------------->| | |<-------------------------------| | | ACK | [*] Key-A in the first MP-capable is related to RFC6824 only and does not exist in RFC6824bis. Figure 1: MPTCP connection setup Multipath TCP itself enables hosts to exchange packets belonging to a single connection over several paths. Implemented in mobile phones (UEs), these paths are usually assigned to different network interfaces within the UE and corresponds to different networks such as cellular and WiFi. The path or network interface for initiating the initial subflow setup is most often provided by the operation system of the UE. For example, if a cellular connection and WiFi are present in a mobile phone, WiFi is usually the interface offered to initiate the MPTCP session. This design falls short in situations where the default path does not provide the best performance compared to other available paths. In a worst case the default path is not even capable of setting up the initial flow letting any other functional path unused. For example, if the WiFi signal is weak, broken or cannot forward traffic to the destination, the establishment of the subflow will be delayed or impossible. This in turn, leads to a longer startup delay or no communication at all for services using MPTCP even if other functional paths are available. Even in scenarios where all paths Amend & Kang Expires May 7, 2020 [Page 4] Internet-Draft MPTCP RobE November 2019 are functional but services would benefit from a setup over the path with the lowest latency, MPTCP has no mean to support this demand. It can be concluded, that sequential path establishment relying with an initial path establishment over an external given default route will result in experience reduction when using MPTCP. So this document proposes solutions to overcome the aforementioned limitations and provides a more robust connection setup compared to traditional MPTCP. RobE_SIM and RobE_eSIM aims to overcome the limitations of [RFC6824] and [I-D.ietf-mptcp-rfc6824bis], using one initial flow and introduces the concept of potential initial flows triggered simultaneously. Potential initial flows gives the freedom to use more than one path to request multipath capability and select the initial flow at a later point. RobE_SIM is a break-before-make mechanism, guaranteeing at least the robust connection establishment, however the RobE_eSIM reuses every potential initial flow request to combine it with less overhead and accelerated multipath availability, leveraging a new MPTCP option MP_JOIN_CAP. From a standardization perspective, the RobE_SIM is fully compliant with [RFC6824] and [I-D.ietf-mptcp-rfc6824bis] and is herein more of a descriptive and procedural nature. The RobE_eSIM requires a new MPTCP option but with the potential to significantly improve the MPTCP experience. For the limitation of the default initial path, RobE_IPS makes no changes to standard MPTCP procedure and improves the performance of connection establishment by introducing an initial path selection strategy and required algorithms. The input for strategy and algorithms is the transmission status information which represents the transmission performance of each available path or network interface. The transmission status information is characterized by at least one of the parameters: signal strength, throughput, round- trip time (RTT) and link success rate. In this way, a path with better transmission performance can be learned and determined and the respective network interface can be used for connection establishment. The most simple approach for a robust MPTCP session establishment is RobE_TIMER, iterating the process of initial path establishment over all available paths, if the previous try has failed. Triggering a new try on a next path is depending on an expiration timer, preferably re-use TCP's in-built expiration timer. Table 1 summarizes the impact of RobE_TIMER, RobE_SIM, RobE_eSIM and RobE_IPS compared to [RFC6824] and [I-D.ietf-mptcp-rfc6824bis]. Amend & Kang Expires May 7, 2020 [Page 5] Internet-Draft MPTCP RobE November 2019 +---------+--------+---------------+--------+-------------+---------+ | Scenari | MPTCP | RobE_TIMER | RobE_S | RobE_eSIM | RobE_IP | | o | | | IM | | S | +---------+--------+---------------+--------+-------------+---------+ | IP | Delaye | In the scope | No | No impact | Delayed | | packet | d conn | of timer | impact | | connect | | loss | ection | | | | ion | +---------+--------+---------------+--------+-------------+---------+ | IP | No con | In the scope | No | No impact | No conn | | broken | nectio | of timer | impact | | ection | | | n | | | | | +---------+--------+---------------+--------+-------------+---------+ | IP | Defaul | Default route | Fastes | Fastest | Selecte | | setup d | t | (+ path 1..n) | t path | path | d path | | uration | route | | | | | | depende | | | | | | | ncy | | | | | | +---------+--------+---------------+--------+-------------+---------+ | MP avai | MP_CAP | sum_1..n(MP_C | MP_CAP | max(MP_CAPA | MP_CAPA | | labilit | ABLE | APABLE_n HS) | ABLE | BLE_1 .. MP | BLE HS | | y durat | HS + M | + MP_JOIN HS | HS + M | _CAPABLE_n | + | | ion | P_JOIN | | P_JOIN | HS) | MP_JOIN | | | HS | | HS | | HS | +---------+--------+---------------+--------+-------------+---------+ | Guarant | Depend | Yes | Yes | Yes | Depends | | eeing | s on | | | | on sele | | session | the de | | | | ction | | setup | fault | | | | | | | route | | | | | +---------+--------+---------------+--------+-------------+---------+ IP: Initial Path; MP: Multi-Path; HS: Handshake Table 1: Overview RobE features during initial connection setup 1.1. Terminology This document makes use of a number of terms that are either MPTCP- specific or have defined meaning in the context of MPTCP, as follows: Path: A sequence of links between a sender and a receiver, defined in this context by a 4-tuple of source and destination address/ port pairs. Subflow: A flow of TCP segments operating over an individual path, which forms part of a larger MPTCP connection. A subflow is started and terminated similar to a regular TCP connection. Amend & Kang Expires May 7, 2020 [Page 6] Internet-Draft MPTCP RobE November 2019 2. Implementation without MPTCP protocol adaptation RobE_TIMER, RobE_SIM and RobE_IPS are compatible with the current MPTCP protocol definitions in [RFC6824] and [I-D.ietf-mptcp-rfc6824bis] but may be lack of the full optimization potential which require protocol adaptation in Section 3. Following sections will describe them in detail. 2.1. Re-transmission Timer(RobE_TIMER) In RobE_TIMER, a new connection is initiated by sending a SYN+MP_CAPABLE along the initial path. If this path is functional, the solution will perform identical to classic MPTCP: the initial flow will be established, and subsequent flows can be created afterwards. If however the initial path is faulty, the retransmission will be triggered on another path. This path might circumvent the dysfunctional network, and allow the client to create an initial subflow. The first path is now seen as a subsequent path and the client sends SYN+MP_JOIN messages to create a subsequent flow. In high latency networks, the initial SYN+MP_CAPABLE might be delayed until the client retries on another path. Once the second SYN arrives at the server, it will try to complete the three-way handshake. If the first SYN was delayed by more than the retransmission time plus half a Round Trip Time (RTT) of the second path, it will arrive at the server after the second SYN. The server could now treat the segment as obsolete and drop it. Amend & Kang Expires May 7, 2020 [Page 7] Internet-Draft MPTCP RobE November 2019 Host A Host B ------------------------ ---------- Address A1 Address A2 Address B1 ---------- ---------- ---------- | | | | SYN + MP_CAPABLE(Key-A[*]) | |Timer---------------------------------------->| | | SYN + MP_CAPABLE(Key-A'[*]) | | |------------------------------->| | | SYN/ACK+MP_CAPABLE(Key-B') | | |<-------------------------------| | | ACK + MP_CAPABLE(Key-A',Key-B')| | |------------------------------->| | | SYN + MP_JOIN(Token-B',R-A) | |--------------------------------------------->| | Subflow will be set up as normal MPTCP | | | [*] Key-A in the first MP-capable is related to RFC6824 only and does not exist in RFC6824bis. Figure 2: The Robe_TIMER Solution Immediately after sending the final ACK of the initial handshake, subflows are established on the remaining paths as defined in [RFC6824] and [I-D.ietf-mptcp-rfc6824bis] [Notes: How to set the Timer is TBD. If there is the case that the first SYN on default path arrives earlier than that from the second path, the MPTCP connection will be initialized on the path of the first SYN.The server could treat the second SYN as obsolete and drop it.] 2.2. Simultaneous Initial Paths Simple Version (RobE_SIM) RobE_SIM is a sender only implementation and no negotiation is required. In RobE_SIM, the MPTCP connection setup benefits from the fastest path. As shown in Figure 3, host A initiates the connection handshake on more than one path independently (SA1 and SA2). The paths selected for RobE_SIM and referred to as potential initital flows, can belong to the number of interfaces on the device or a subset selected on experience. When Host A receives the first SYN/ ACK back from Host B (SA3), the path carrying this message is identified as the normal initial path. Host A sends then immediately a TCP RST message (SA6.1) on any other path used for simultaneous connection setup causing an immediate termination of assigned flows (break-before-make). The terminated ones are merged as subsequent subflows following the JOIN procedure described in [RFC6824] and Amend & Kang Expires May 7, 2020 [Page 8] Internet-Draft MPTCP RobE November 2019 [I-D.ietf-mptcp-rfc6824bis]. The process is equivalent to any other scenario where the SYN/ACK arrives on an other path than depicted in Figure 3. Host A Host B ------------------------ ---------- Address A1 Address A2 Address B1 ---------- ---------- ---------- | | | | SYN + MP_CAPABLE(Key-A[*]) | (SA1) |--------------------------------------------->| (SB1) | | SYN + MP_CAPABLE(Key-A'[*]) | (SA2) | |------------------------------->| (SB2) | | | (SA3) |<---------------------------------------------| (SB3) | SYN/ACK + MP_CAPABLE(Key-B) | (SA4) | |<-------------------------------| (SB4) | | SYN/ACK + MP_CAPABLE(Key-B') | | | | | ACK + MP_CAPABLE(Key-A, Key-B) | (SA5) |--------------------------------------------->| (SB5) | | RST | (SA6.1) | |------------------------------->| (SB6.1) RobE SIM | | SYN + MP_JOIN(Token-B, R-A) | (robust) | |------------------------------->| | | MP_JOIN Process... | [*] Key-A in the first MP-capable is related to RFC6824 only and does not exist in RFC6824bis. Figure 3: MPTCP RobE_SIM Connection Setup 2.3. Heuristic Initial Path Selection (RobE_IPS) 2.3.1. Architecture Figure 4 provides the architecture for RobE_IPS and employs an "Initial Path Selection" logic which can be integrated into the MPTCP stack or exists as an isolated module in the terminal. The IPS logic has access to a set of transmission status information for each available path or its belonging network interfaces. When an application starts a first communication, IPS selects based on the available path transmission characteristics the path with the highest probability to succeed. Amend & Kang Expires May 7, 2020 [Page 9] Internet-Draft MPTCP RobE November 2019 +-------------------+ +-------------------+ | Terminal | | Server | | +-------------+ | | +-------------+ | | |Application n| | | |Application n| | | +-------------+ | | +-------------+ | | | | | | | | +-------------+ | | | | | |Initial-path | |-------+ | | | | | Selection | | | | | | | +-------------+ | | | | | | | | +--------+ | | | | +-------------+ |--|Internet|--| +-------------+ | | | MPTCP Stack | |--+--------+--| | MPTCP Stack | | | +-------------+ | | +-------------+ | +-------------------+ +-------------------+ Figure 4: Architecture for Initial-path Selection 2.3.2. Typical Scenarios Two typical RobE_IPS scenarios are presented in this section. Figure 5 shows that the "Initial Path Selection" logic executed for each MPTCP connection establishment. On the other hand Figure 6 describes that "Initial Path Selection" in case no path information are available. Considering the fact that no heuristics are given before a recent MPTCP connection was established, the default initial path can be adopted. Further combinations and implementations with more or less sophisticated heuristics are possible. Amend & Kang Expires May 7, 2020 [Page 10] Internet-Draft MPTCP RobE November 2019 +---------------+ | Application | | Request | +---------------+ | V +---------------+ +--->| Initial-path |<---+ | | Selection | | | +---------------+ | | | | | V |Info | +---------------+ | | | Set initial |----+ | | path | | | for MPTCP | | +---------------+ | | | V | +---------------+ |No |Establish MPTCP| +----| Connection | +---------------+ |Yes V Figure 5: RobE_IPS for each connection establishment Amend & Kang Expires May 7, 2020 [Page 11] Internet-Draft MPTCP RobE November 2019 +--------------+ | Application | | Request | +--------------+ | V +--------------+Yes | First |------------+ | Connection? | | +--------------+ | |No | V | +--------------+ V +----->| Initial-path |<-+ +-------+ | | Selection | | |Default| | +--------------+ | |initial| | | | | path | | | | +-------+ | V |Info | | +--------------+ | | | | Set initial |--+ | | | path | | | | for MPTCP | | | +--------------+ | | | | | V | |No +--------------+ | +------| Successful? |<-----------+ +--------------+ |Yes V Figure 6: RobE_IPS using default route when no meaningful heuristic available Figure 7 shows the process flow of "Initial Path Selection". Upon a request from an application, the IPS logic will acquire transmission status information which represents the transmission performance of each available path or network interface and evaluate it. The transmission status information is characterized by at least one of the parameters: signal strength, throughput, round-trip time (RTT) and link success rate. In this way, the path with the best transmission performance can be determined and used for connection establishment. Amend & Kang Expires May 7, 2020 [Page 12] Internet-Draft MPTCP RobE November 2019 | V +---------------------------+ |Acquire transmission status| | info for available paths | +---------------------------+ | V +---------------------------+ | Evaluating the status | | for available paths | +---------------------------+ |No V +---------------------------+ | Determining an available | | path with better | | transmission | | performance | +---------------------------+ | V +---------------------------+ | Using the network | | interface | |corrresponding to the path | | with better transmission | |performance for connection | | establishment | +---------------------------+ | V Figure 7: Implementation process for Initial Path Selection 2.3.3. Path decision information The level of heuristic can be mainly divided into three layer: application level, transport-layer level and link-layer level based on the information acquisition method. For example, RTT can be calculated for each path within a MPTCP connection and belongs thereof to the transport-layer level. The transmission status information for each available path SHOULD be characterized by at least one of the parameters: signal strength, throughput, RTT and link success rate. Application level information are more seen for statistical purposes. Amend & Kang Expires May 7, 2020 [Page 13] Internet-Draft MPTCP RobE November 2019 o Application level: application name, domain name, port number and location. o Transport-layer level: RTT, CWND, Error rate. 2.3.4. Initial Path Selection use local RTT information Figure 8 presents a "Initial Path Selection" logic based on RTT, e.g. assuming two paths over LTE and WiFi access. RTT calculation on the transport layer usually reflect the time when an information is sent and an related acknowledgment received. For an asymmetric usage (e.g. download only) of a communication it might happen that recent RTT calculation is only available on sender side which is possibly not the side which employs the IPS logic. A solution for this can be found in Section 3.2. Instead of using the most recent RTT value of a path a filtered value consisting of several measured RTTs can be used. A RTT can also be derived from link layer information but may has a limited meaning when it does not picture the end-to-end latency. +-------------------+ | New Session | +-------------------+ | V +-------------------+ No |Running Connections|-----------+ |(LTE.RTT| Any data for | No | | Initial Path |----------+ | | Selection? | | | +--------------+ | | | | | V V | +--------------+ +--------+ | | Initial-path | |RobE_SIM| | | Selection |<-+ +--------+ | +--------------+ | | | | | | | V |Info | | +---------------+ | | |No |Establish MPTCP|-+ | +------| Connection |<--------+ +---------------+ | V No +---------------+ <------| Successful ? | Network+---------------+ Problem |Yes V Figure 9: Combination of RobE_SIM and RobE_IPS 2.5. Combination of RobE_TIMER and RobE_IPS Since RobE_IPS solely does not guarentee that session can be set up on the selection of initial path, it can also be combined with RobE_TIMER which generates less overhead compared to the combination with RobE_SIM in Section 2.4 and guarantess session setup. RobE_TIMER can be introduced to optimize the control of path switching when the initial path selected by RobE_IPS is dysfunctional. When the system enables RobE_IPS and uses the selected initial path for session establishment,it sets the timer for path switching. When timer is expired, the system will change to another path to re-establish connection according to Section 2.1. Amend & Kang Expires May 7, 2020 [Page 15] Internet-Draft MPTCP RobE November 2019 +---------------+ | Application | | Request | +---------------+ | V +---------------+ | Initial Path | |----->| Selection | | | and Set Timer | | +---------------+ | | | V |Yes +---------------+ +------| Timer is up? | +---------------+ |No V +---------------+ |Establish MPTCP| | Connection | +---------------+ | V No +---------------+ <------| Successful? | Network+---------------+ Problem |Yes V Figure 10: Combination of RobE_Timer and RobE_IPS 3. Implementation with Bi-directional MPTCP Support Solutions which requires bi-directional support between two MPTCP hosts promise to have better and possibly more features. However, they cannot be defined without extending current standards in [RFC6824] and [I-D.ietf-mptcp-rfc6824bis]. The RobE_SIM and RobE_IPS approach are both capable of profit from an explicit support of the remote end host and defined within this section. 3.1. Simultaneous Initial Paths Extended Version (RobE_eSIM) RobE_eSIM extends RobE_SIM by reusing the potential initial flows. This eliminates the overhead from RobE_SIM by introducing a new option MP_JOIN_CAP and accelerate the transmission speed by early availablity of multiple paths. Further it relaxes the dependency on a reliable third ACK of the 3-way handshake in Amend & Kang Expires May 7, 2020 [Page 16] Internet-Draft MPTCP RobE November 2019 [I-D.ietf-mptcp-rfc6824bis]. Remote endpoint support can be negotiated in two ways, an implicit in Section 3.1.1 or explicit in Section 3.1.2 3.1.1. RobE_eSIM implicit Negotiation and Procedure Similar to RobE_SIM in Section 2.2 the establishment process of [RFC6824] or [I-D.ietf-mptcp-rfc6824bis] is applied independently on multiple path simultaneously. In Figure 11 this is shown in SA1 and SA2. The first path which returns a SYN/ACK (e.g. SA3) is selected as the initial path and proceed with the traditional establishment process (SA5). Any other path which has to send the final ACK of the 3-way handshake includes a new option MP_JOIN_CAP (see definition in Section 3.1.3.2) instead of a MP_CAPABLE (SA6.2). Host A Host B ------------------------ ---------- Address A1 Address A2 Address B1 ---------- ---------- ---------- | | | | SYN + MP_CAPABLE(Key-A[*]) | (SA1) |--------------------------------------------->| (SB1) | | SYN + MP_CAPABLE(Key-A'[*]) | (SA2) | |------------------------------->| (SB2) | | | (SA3) |<---------------------------------------------| (SB3) | SYN/ACK + MP_CAPABLE(Key-B) | (SA4) | |<-------------------------------| (SB4) | | SYN/ACK + MP_CAPABLE(Key-B') | | | | | ACK + MP_CAPABLE(Key-A, Key-B) | (SA5) |--------------------------------------------->| (SB5) | | | (SA6.2) | | | (SB6.2) RobE EXT | | ACK + MP_JOIN_CAP(Key-A, HMAC) | (+fast) | |------------------------------->| [*] Key-A in the first MP-capable is related to RFC6824 only and does not exist in RFC6824bis. Figure 11: MPTCP RobE_eSIM implicit Connection Setup Following the possible process in Figure 11, two further constellations are imaginable and elaborated below. 1. In the flow diagram Figure 11, A1<->B1 is assumed to be the initial flow. A2<->B1 shall be recycled and the ACK is sent with MP_JOIN_CAP. Furthermore, the MP_CAPABLE arrives first at Host B Amend & Kang Expires May 7, 2020 [Page 17] Internet-Draft MPTCP RobE November 2019 (SB5) and the MP_JOIN_CAP afterwards (SB6.2). When the MP_JOIN_CAP is received, Host B has to iterate over the connection list once (like MP_JOIN) and check for Key-A availability. If a Key-A connection is found, this one is validated against the HMAC value. The validation has two reasons: first, several Key-A can exist, because different hosts may choose the same Key-A by accident. Furthermore, no one can join a connection by just recording/brute-forcing Key-A and duplicating the request. 2. Like above, but MP_JOIN_CAP arrives before last MP_CAPABLE at Host B * [I-D.ietf-mptcp-rfc6824bis]; Based on Key-A, Host B will iterate over the connection list, but it will not find a match, because Key-A of the previous selected initial flow (SA3, SA5) has not arrived yet. So it will continue with a fast iteration only over the connections which are still in establishment phase using the 10 bit Key-B fast hash (crc16(Key-B) & 0x3FF). If it matches against a (precomputed) existing Key-B_fast_hash in the connection list, it will validate the request using the HMAC(Key-A+B+B') to ensure legitimation. If successful, both, the initial flow and the MP_JOIN_CAP flow, can be immediately established. This is true, because without the knowledge of Key-B, Host A could not calculate the HMAC. So it is clear, that Host A had received the SYN/ACK (SB3). This also mitigates the exchange of a reliable ACK during the handshake process. MPTCP sends the Key-A only with the last ACK and therefore prevents subsequent flow establishment until successful reception at Host B. Using RobE_EXT, the reception of a MP_JOIN_CAP ([I-D.ietf-mptcp-rfc6824bis]) is sufficient to establish both, the path carrying Key-B and Key-B'. * [RFC6824]; Can match based on Key-A, same effort as for a MP JOIN. 3. A2<->B1 is selected as initial flow, because the respective SYN/ ACK returns earlier at Host A. It is the same as above, just the other way round. 3.1.2. RobE_eSIM explicit Negotiation and Procedure The process of an explicit negotiation of RobE_eSIM follows Figure 11 but uses the ROBE_eSIM_EN option Figure 13 additionally during the handshake procedure. Amend & Kang Expires May 7, 2020 [Page 18] Internet-Draft MPTCP RobE November 2019 Host A Host B ------------------------ ---------- Address A1 Address A2 Address B1 ---------- ---------- ---------- | | | | SYN+MP_CAPABLE+ROBE_eSIM_EN(Key-A[*]) | |----------------------------------------------------->| | | SYN+MP_CAPABLE+ROBE_eSIM_EN(Key-A'[*]) | | |--------------------------------------->| | SYN/ACK+MP_CAPABLE+ROBE_eSIM_EN(Key-B) | |<---------------------------------------------------->| | | SYN/ACK+MP_CAPABLE+ROBE_eSIM_EN(Key-B')| | |<---------------------------------------| | ACK+MP_CAPABLE(Key-A,Key-B) | |----------------------------------------------------->| | | | | | ACK+MP_JOIN_CAP(Key-A,HMAC) | | |--------------------------------------->| | | | [*] Key-A in the first MP-capable is related to RFC6824 only and does not exist in RFC6824bis. Figure 12: MPTCP RobE_eSIM explicit Connection Setup 3.1.3. Protocol Adaptation 3.1.3.1. ROBE_eSIM_EN Option 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 +---------------+---------------+-------+-------+---------------+ | Kind | Length |Subtype| (reserved) | +---------------+---------------+-------+-------+---------------+ Figure 13: ROBE_eSIM_EN_OPTION 3.1.3.2. MP_JOIN_CAP Option Amend & Kang Expires May 7, 2020 [Page 19] Internet-Draft MPTCP RobE November 2019 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 +---------------+---------------+-------+-------+---------------+ | Kind | Length |Subtype| | ADDR_ID | +---------------+---------------+-------+-------+---------------+ | Sender's Key-A (64 bits) | | | | | +---------------------------------------------------------------+ | HMAC (>=96 bits) | | | | | : : +---------------------------------------------------------------+ Key-B_fast_hash = crc16(Key-B) & 0x3FF -> (10bit) HMAC_keys = HMAC(Key-A+Key-B+Key-B') -> (>=96bit) HMAC = (HMAC_keys & ~0x3FF) | Key-B_fast_hash -> (size HMAC_keys) Figure 14: MP_JOIN_CAP Computational effort on receiver side is most often expected to be the same as with MP_JOIN. Key-A ensures identification of related flows Key-B_fast_hash enables MP session even when selected initial flow is not fully established yet (slight computational overhead). HMAC authenticates relationship of initial and potential initital flows. 3.1.4. Fallback Mechanisms 3.1.4.1. Fallback mechanism for implicit RobE_eSIM [TBD] 3.1.4.2. Fallback mechanism for explicit RobE_eSIM This mechanism considers that both sides support MPTCP capability but the receiver does not equipped with RobE_eSIM. MPTCP session with RobE_eSIM negotiation will seamlessly fallback to normal MPTCP process. [Requires further check how an unaware Host B reacts on possible ROBE_eSIM_EN; Ignore or RST? See also RFC6824 Sec. 3.6 "Shoukd fallback [...] the path does not support the MPTCP options"] Amend & Kang Expires May 7, 2020 [Page 20] Internet-Draft MPTCP RobE November 2019 Host A Host B ------------------------ ---------- Address A1 Address A2 Address B1 ---------- ---------- ---------- | | | | SYN+MP_CAPABLE+ROBE_eSIM_EN | |------------------------------------------>| | | SYN+MP_CAPABLE+ROBE_eSIM_EN | | |---------------------------->| | SYN/ACK+MP_CAPABLE | |<----------------------------------------->| | | SYN/ACK+MP_CAPABLE | | |<----------------------------| | ACK+MP_CAPABLE | |------------------------------------------>| | | RST | | |---------------------------->| | | SYN+MP_JOIN | | |---------------------------->| | | MP_JOIN Process... | | | | Figure 15: Fallback to MPTCP when missing RobE_eSIM support 3.1.4.3. Fallback to regular TCP when missing MPTCP support When the receiver is not MPTCP enabled, MPTCP session with RobE_eSIM negotiation will seamlessly fallback to regular process which is illustrated in this section. Amend & Kang Expires May 7, 2020 [Page 21] Internet-Draft MPTCP RobE November 2019 Host A Host B ------------------------ ---------- Address A1 Address A2 Address B1 ---------- ---------- ---------- | | | | SYN+MP_CAPABLE+ROBE_eSIM_EN | |------------------------------------------>| | | SYN+MP_CAPABLE+ROBE_eSIM_EN | | |---------------------------->| | SYN/ACK | |<----------------------------------------->| | | SYN/ACK | | |<----------------------------| | ACK | |------------------------------------------>| | | RST | | |---------------------------->| | | Regular TCP Process... | | | | Figure 16: Fallback to TCP without MPTCP support 3.1.5. Comparison Robe_SIM and RobE_eSIM Potential initial flows in RobE_SIM Section 2.2 and RobE_eSIM Section 3.1 guarantee MPTCP session establishment if at least one selected path for session establishment is functional. Figure 17 makes the differences between both approaches visible and points to the latest decision possibility during session setup when RobE_SIM or RobE_eSIM can be selected. Until SA5 in Figure 17 traditional MPTCP connection setup is independently applied on multiple paths simultaneously and offers to select the initial flow later (potential initial flows). The final decision which path is selected as the main one and the handling of the remaining flow(s) differs in SA6.1 when RobE_SIM is applied or instead SA6.2 RobE_eSIM. Amend & Kang Expires May 7, 2020 [Page 22] Internet-Draft MPTCP RobE November 2019 Host A Host B ------------------------ ---------- Address A1 Address A2 Address B1 ---------- ---------- ---------- | | | | SYN + MP_CAPABLE(Key-A[*]) | (SA1) |--------------------------------------------->| (SB1) | | SYN + MP_CAPABLE(Key-A'[*]) | (SA2) | |------------------------------->| (SB2) | | | (SA3) |<---------------------------------------------| (SB3) | SYN/ACK + MP_CAPABLE(Key-B) | (SA4) | |<-------------------------------| (SB4) | | SYN/ACK + MP_CAPABLE(Key-B') | | | | | ACK + MP_CAPABLE(Key-A, Key-B) | (SA5) |--------------------------------------------->| (SB5) | | RST | (SA6.1) | |------------------------------->| (SB6.1) RobE SIM | | | (robust) | | | ------------------------------------------------------------------- RobE EXT | | | (+fast) | | ACK + MP_JOIN_CAP(Key-A, HMAC) | (SA6.2) | |------------------------------->| (SB6.2) [*] Key-A in the first MP-capable is related to RFC6824 only and does not exist in RFC6824bis. Figure 17: MPTCP RobE_SIM and RobE_eSIM connection setup 3.1.6. Security Consideration [Tbd, however no differences to [RFC6824] and [I-D.ietf-mptcp-rfc6824bis] are expected] 3.2. Heuristic Initial Path Selection with remote RTT Measurement 3.2.1. Description Usually the path RTT can be determined by a time difference between sending a package and an ACK and is integrated into the TCP protocol. For asymmetric transmission, the latest RTT for TCP flows is calculated by the side which sends data at latest and possible does not correspond to the site which employs RobE_IPS. This problem is already elaborated in Section 2.3.4 and can be solved by transmitting the RTT information per subflow. The negotiation procedure is Amend & Kang Expires May 7, 2020 [Page 23] Internet-Draft MPTCP RobE November 2019 depicted in Figure 18 and uses the MPTCP option L_RTT_EN defined in Section 3.2.2. Host A Host B ------------------------ ---------- Address A1 Address A2 Address B1 ---------- ---------- ---------- | | | | SYN+MP_CAPABLE+L_RTT_EN | |------------------------------------------>| | SYN/ACK+MP_CAPABLE+L_RTT_EN | |<------------------------------------------| | ACK+MP_CAPABLE | |------------------------------------------>| | ACK+DSS+L_RTT_EN(latest RTT)+Data | |<------------------------------------------| | | SYN+MP_JOIN | | |---------------------------->| | | MP_JOIN Process... | | | | Figure 18: Negotiation procedure for RTT exchange A successful negotiation allows the exchange of the measured RTT value from one subflow of a MPTCP host to another using the "Latest RTT" field within the L_RTT_EN option. 3.2.2. Protocol Adaptation Calculating the "Latest RTT" by a remote host in an asymmetry transmission scenario should be transferred from remote host to the client running RobE_IPS. So a new MPTCP subtype option named L_RTT_EN is allocated for this function. During the three-way handshake L_RTT_EN is used for negotiation of remote RTT measurement capability between client and server (in Section 3.2.1). When both parts support the usage of remote RTT measurment, the "Latest RTT" field in L_RTT_EN is applied for carrying the value of latest RTT computed by the remote host. Amend & Kang Expires May 7, 2020 [Page 24] Internet-Draft MPTCP RobE November 2019 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 +---------------+---------------+-------+-------+---------------+ | Kind | Length |Subtype| (reserved) | +---------------+---------------+-------+-------+---------------+ | Latest RTT (32 bits) | | | | | +---------------------------------------------------------------+ Figure 19: ROBE_L_RTT_EN OPTION 3.2.3. Fallback Mechanism When the receiver is not L_RTT_EN capable, MPTCP session with L_RTT_EN negotiation will seamlessly fallback to normal MPTCP process. [TBD, Need same checks as Section 3.1.4.2] Host A Host B ------------------------ ---------- Address A1 Address A2 Address B1 ---------- ---------- ---------- | | | | SYN+MP_CAPABLE+L_RTT_EN | |------------------------------------------>| | SYN/ACK+MPTCP_CAPABLE | |<------------------------------------------| | ACK+MPTCP_CAPABLE | |------------------------------------------>| | | SYN+MP_JOIN | | |---------------------------->| | | MP_JOIN Process... | | | | Figure 20: Fallback to MPTCP without RobE_IPS 3.2.4. Security Consideration [Tbd] 4. IANA Considerations This document defines three new values to MPTCP Option Subtype as following. Amend & Kang Expires May 7, 2020 [Page 25] Internet-Draft MPTCP RobE November 2019 +-------+--------------+--------------------------------+-----------+ | Value | Symbol | Name | Reference | +-------+--------------+--------------------------------+-----------+ | TBD | ROBE_eSIM_EN | RobE_eSIM enabled | Section | | | | | 3.1 | +-------+--------------+--------------------------------+-----------+ | TBD | MP_JOIN_CAP | Join connection directly in | Section | | | | RobE_eSIM | 3.1 | +-------+--------------+--------------------------------+-----------+ | TBD | L_RTT_EN | Server RTT enabled | Section | | | | | 3.2 | +-------+--------------+--------------------------------+-----------+ RobE Option Subtypes Table 2: RobE Option Subtypes 5. Informative References [I-D.ietf-mptcp-rfc6824bis] Ford, A., Raiciu, C., Handley, M., Bonaventure, O., and C. Paasch, "TCP Extensions for Multipath Operation with Multiple Addresses", draft-ietf-mptcp-rfc6824bis-18 (work in progress), June 2019. [RFC0793] Postel, J., "Transmission Control Protocol", STD 7, RFC 793, DOI 10.17487/RFC0793, September 1981, . [RFC6824] Ford, A., Raiciu, C., Handley, M., and O. Bonaventure, "TCP Extensions for Multipath Operation with Multiple Addresses", RFC 6824, DOI 10.17487/RFC6824, January 2013, . Authors' Addresses Markus Amend Deutsche Telekom T-Online-Allee 7 64295 Darmstadt Germany Email: Markus.Amend@telekom.de Amend & Kang Expires May 7, 2020 [Page 26] Internet-Draft MPTCP RobE November 2019 Jiao Kang Huawei D2-03,Huawei Industrial Base Longgang District Shenzhen China Email: kangjiao@huawei.com Amend & Kang Expires May 7, 2020 [Page 27]