Transferring Bulk Data over the GeneRic Autonomic Signaling Protocol (GRASP)
draft-carpenter-anima-grasp-bulk-03

Abstract

This document describes how bulk data may be transferred between Autonomic Service Agents via the GeneRic Autonomic Signaling Protocol (GRASP).

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 July 11, 2019.

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
• 2. General Method for Bulk Transfer
• 3. Example for File Transfer
• 4. Loss Detection
• 5. Maximum Transmission Unit
• 6. Pipelining
• 7. Other Considerations
• 8. Possible Future Work
• 9. Implementation Status [RFC Editor: please remove]
• 10. Security Considerations
• 11. IANA Considerations
• 12. Acknowledgements
• 13. References
• 13.1. Normative References
• 13.2. Informative References
• Appendix A. Change log [RFC Editor: Please remove]
• Authors' Addresses

1. Introduction

The document [I-D.liu-anima-grasp-distribution] discusses how information may be distributed within the secure Autonomic Networking Infrastructure (ANI) [I-D.ietf-anima-reference-model]. Specifically, it describes using the Synchronization and Flood Synchronization mechanisms of the GeneRic Autonomic Signaling Protocol (GRASP) [I-D.ietf-anima-grasp] for this purpose. However, those mechanisms are limited to distributing GRASP Objective Options contained in messages that cannot exceed the GRASP maximum message size of 2048 bytes.

There are scenarios in autonomic networks where this restriction is a problem. One example is the distribution of network policy in lengthy formats such as YANG or JSON. Another case might be an Autonomic Service Agent (ASA) uploading a log file to the Network Operations Center (NOC). A third case might be a supervisory system downloading a software upgrade to an autonomic node. A related case might be installing the code of a new or updated ASA to a target node (see the discussion of ASA life cycles in [I-D.carpenter-anima-asa-guidelines]).

Naturally, an existing solution such as a secure file transfer protocol or secure HTTP might be used for this. Other management protocols such as syslog [RFC5424] or NETCONF [RFC6241] might also be used for related purposes, or might be mapped directly over GRASP. The present document, however, applies to any scenario where it is preferable to re-use the autonomic networking infrastructure itself to transfer a significant amount of data, rather than install and configure an additional mechanism. The basic model is to use the GRASP Negotiation process to transfer and acknowledge multiple blocks of data in successive negotiation steps.

The emphasis is placed on simplicity rather than efficiency, high throughput, or advanced functionality. For example, if a transfer gets out of step or data packets are lost, the strategy is to abort the transfer and try again. In an enterprise network with low bit error rates, and with GRASP running over TCP, this is not considered a serious issue. Clearly, a more sophisticated approach could be designed but if the application requires that, existing protocols could be used, as indicated in the preceding paragraph.

NOTE: This is an early draft of a solution. As the specification becomes more mature, the authors expect it to become precise enough to be placed on the standards track.

2. General Method for Bulk Transfer

As for any GRASP operation, the two participants are considered to be Autonomic Service Agents (ASAs) and they communicate using a specific GRASP Objective Option, containing its own name, some flag bits, a loop count, and a value. In bulk transfer, we can model the ASA acting as the source of the transfer as a download server, and the destination as a download client. No changes or extensions are required to GRASP itself, but compared to a normal GRASP negotiation, the communication pattern is slightly asymmetric:

1. The client first discovers the server by the GRASP discovery mechanism (M_DISCOVERY and M_RESPONSE messages).
2. The client then sends a GRASP negotiation request (M_REQ_NEG message). The value of the objective expresses the requested item (e.g., a file name - see the next section for a detailed example).
3. The server replies with a negotiation step (M_NEGOTIATE message). The value of the objective is the first section of the requested item (e.g., the first block of the requested file as a raw byte string).
4. The client replies with a negotiation step (M_NEGOTIATE message). The value of the objective is a simple acknowledgement (e.g., the text string 'ACK').

The last two steps repeat until the transfer is complete. The server signals the end by transferring an empty byte string as the final value. In this case the client responds with a normal end to the negotiation (M_END message with an O_ACCEPT option).

Errors of any kind are handled with the normal GRASP mechanisms, in particular by an M_END message with an O_DECLINE option in either direction.

The block size must be chosen such that each step does not exceed the GRASP message size limit of 2048 bits.

This approach is safe since each block must be positively acknowledged, and data transfer errors will be detected by TCP. If a future variant of GRASP runs over UDP, the mandatory UDP checksum for IPv6 will detect such errors. The method does not specify retransmission for failed blocks, so a failed transfer will need to be restarted.

An observant reader will notice that the GRASP loop count mechanism, intended to terminate endless negotiations, will cause a problem for large transfers. For this reason, both the client and server must artificially increment the loop count by 1 before each negotiation step, cancelling out the normal decrement at each step.

If network load is a concern, the data rate can be limited by inserting a delay before each negotiation step, with the GRASP timeout set accordingly. Either the server or the client, or both, could insert such a delay. Also, either side could use the GRASP Confirm Waiting (M_WAIT) message to slow the other side down.

The description above concerns bulk download from a server (responding ASA) to a client (requesting ASA). The data transfer could also be in the opposite (upload) direction with minor modifications to the procedure: the client would send the file name and the data blocks, and the server would send acknowledgements.

3. Example for File Transfer

This example describes a client ASA requesting a file download from a server ASA.

Firstly we define a GRASP objective informally:

["411:mvFile", 3, 6, value]

The formal CDDL definition [I-D.ietf-cbor-cddl] is:

  mvfile-objective = ["411:mvFile", objective-flags, loop-count, value]
  
  objective-flags = ; as in the GRASP specification
  loop-count = ; as in the GRASP specification
  value = any

The objective-flags field is set to indicate negotiation.

Dry run mode must not be used.

The loop-count is set to a suitable value to limit the scope of discovery. A suggested default value is 6.

The value takes the following forms:

• In the initial request from the client, a UTF-8 string containing the requested file name (with file path if appropriate).
• In negotiation steps from the server, a byte string containing at most 1024 bytes. However:
  • If the file does not exist, the first negotiation step will return an M_END, O_DECLINE response.
  • After sending the last block, the next and final negotiation step will send an empty byte string as the value.

• In negotiation steps from the client, the value is the UTF-8 string 'ACK'.

Note that the block size of 1024 is chosen to guarantee not only that each GRASP message is below the size limit, but also that only one TCP data packet will be needed, even on an IPv6 network with a minimum link MTU.

We now present outline pseudocode for the client and the server ASA. The API documented in [I-D.ietf-anima-grasp-api] is used in a simplified way, and error handling is not shown in detail.

Pseudo code for client ASA (request and receive a file):

requested_obj = objective('411:mvFile')
locator = discover(requested_obj)
requested_obj.value = 'etc/test.pdf'
received_obj = request_negotiate(requested_obj, locator)
if error_code == declined:
    #no such file
    exit

file = open(requested_obj.value)
file.write(received_obj.value) #write to file
eof = False
while not eof:
    received_obj.value = 'ACK'
    received_obj.loop_count = received_obj.loop_count + 1
    received_obj = negotiate_step(received_obj)
    if received_obj.value == null:
        end_negotiate(True)
        file.close()
        eof = True     
    else:
        file.write(received_obj.value) #write to file

#file received
exit

Pseudo code for server ASA (await request and send a file):

supported_obj = objective('411:mvFile')
requested_obj = listen_negotiate(supported_obj)
file = open(requested_obj.value) #open the source file
if no such file:
    end_negotiate(False) #decline negotiation
    exit

eof = False
while not eof:
    chunk = file.read(1024) #next block of file
    requested_obj.value = chunk
    requested_obj.loop_count = requested_obj.loop_count + 1
    requested_obj = negotiate_step(requested_obj)
    if chunk == null:
        file.close()
        eof = True
        end_negotiate(True)
        exit
    if requested_obj.value != 'ACK':
        #unexpected reply...

4. Loss Detection

The above description and example assume that GRASP is implemented over a reliable transport layer such as TCP, such that lost or corrupted messages are not likely. Rarely, an error might be detected via a missing ACK, in which case the transfer would be aborted and restarted. In the event that GRASP is implemented over an unreliable transport layer such as UDP, it would be possible to add a block number to both the data block and acknowledgement objectives, so that missing blocks can be retransmitted, or duplicate blocks can be ignored. For example, the objective in Section 3 would become:

  mvfile-objective = ["411:mvFile", objective-flags, loop-count, value]
  
  objective-flags = ; as in the GRASP specification
  loop-count = ; as in the GRASP specification
  value = [block-number, any]
  block-number = uint

It would also be necessary for the transport layer to detect data errors, for example by enabling UDP checksums.

5. Maximum Transmission Unit

In an IPv6 environment, a minimal MTU of 1280 bytes can be assumed, and assuming that high throughput is not a requirement, bulk transfers can be designed to match that MTU. However, there are environments where the underlying physical MTU is much smaller. For example, on an IEEE 802.15.4 network it may be less than 100 bytes [RFC4944]. In such a case, a bulk transfer solution has several choices:

1. Accept the overhead of an adaptation layer, and therefore assume a network-layer MTU of 1280 bytes. Indeed, the presence of such an adaptation layer may be impossible to detect.
2. Attempt to determine the actual MTU available without lower-layer fragmentation. This however will be impossible without using low-level functions of the socket interface.
3. Attempt to determine a message size that provides optimum performance, by some sort of trial-and-error solution.

These complexities suggest that using a GRASP-based mechanism is unlikely to be optimal in environments with a very small physical MTU.

6. Pipelining

The above description and example descibe a simple handshake model where each block is acknowledged before the next block is sent. For the scenarios discussed in Section 1, this should be acceptable. Therefore we do not suggest adding a pipelining or windowing mechanism. If high throughput is required, a conventional file transfer protocol should be used.

7. Other Considerations

If multiple transfers are requested simultaneously, each one will proceed as a separate GRASP negotiation session. The ASA acting as the server must be coded accordingly, like any ASA that needs to handle simultaneous sessions [I-D.carpenter-anima-asa-guidelines].

Bulk transfer might become a utility function for use by various ASAs, such as those supporting YANG or JSON distribution, log file uploads, or code downloads. In this case some form of user space API for bulk transfer will be required. This could be in the form of an inter-process communication call between the ASA in question and the ASA implementing the bulk transfer mechanism. The details are out of scope for this document.

8. Possible Future Work

The simple file transfer mechanism described above is only an example. Other application scenarios should be developed.

The mechanism described in this document is suitable for simple unicast scenarios where GRASP runs over TCP and can be treated as a reliable protocol. A more sophisticated approach would be needed in at least two cases:

1. A scenario where GRASP runs over UDP, where error detection and retransmission would be essential.
2. A scenario where multicast data distribution is required, so that a mechanism such as Trickle [RFC6206] would be appropriate.

These solutions might also require extensions to the GRASP protocol itself.

9. Implementation Status [RFC Editor: please remove]

A prototype open source Python implementation of simple file transfer has been used to verify the mechanism described above. It may be found at https://github.com/becarpenter/graspy/blob/master/getter.py and https://github.com/becarpenter/graspy/blob/master/pusher.py .

10. Security Considerations

All GRASP transactions are secured by the mandatory security substrate required by [I-D.ietf-anima-grasp]. No additional security issues are created by the application of GRASP described in this document.

11. IANA Considerations

This document makes no request of the IANA.

12. Acknowledgements

Thanks to Joel Halpern and other members of the ANIMA WG.

13. References

13.1. Normative References

13.2. Informative References

Appendix A. Change log [RFC Editor: Please remove]

draft-carpenter-anima-grasp-bulk-03, 2019-01-07:

Added future work section, implementation status.

draft-carpenter-anima-grasp-bulk-02, 2018-06-30:

Update reference, fix TBDs.

draft-carpenter-anima-grasp-bulk-01, 2018-03-03:

Updates after IETF100 discussion.

draft-carpenter-anima-grasp-bulk-00, 2017-09-12:

Initial version.

Authors' Addresses