Transaction management We have now applied the end-to-end argument in the construction of the SWalloW distributed data storage system[ 15], where it leads to significant reduction in overhead. SWALLOW provides data storage servers called repositories that can be used remotely to store and retrieve data. Accessing data at a repository is done by sending it a message specifying the object to be accessed, the version, and type of access(read/write), plus a value to be written if the access is a write. The underlying message communication system does not suppress duplicate messages, since a) the object identifier plus the version information suffices to detect duplicate writes, and b) the effect of a duplicate read requ ge is only to generate a duplicate response, which easily discarded by the originator. Consequently, the low-level message communication protocol is significantly simplifie The underlying message communication system does not provide delivery acknowledgement either. The acknowledgement that the originator of a write request needs is that the data was stored safely. This acknowledgement can be provided only by high levels of the SWALLOW stem. For read requests, a delivery acknowledgement is redundant, since the response containing he value read is sufficient acknowledgement. By eliminating delivery acknowledgements, the number of messages transmitted is halved. This message reduction can have a significant effect on both host load and network load, improving performance. This same line of reasoning has also been used in development of an experimental protocol for remote access to disk records[6]. The resulting reduction in path length in lower-level protocols was important in maintaining good performance on remote disk access Identifying the ends Using the end-to-end argument sometimes requires subtlety of analy is of application requirements. For example, consider a computer communication network that carries some packet voice connections, conversations between digital telephone instruments. For those connections that carry voice packets, an unusually strong version of the end-to-end argument applies: if low levels of the communication system try to accomplish bit-perfect communication, they will probably introduce uncontrolled delays in packet delivery, for example, by requesting retransmission of damaged packets and holding up delivery of later packets until earlier ones have been correctly retransmitted. Such delays are disruptive to the voice application, which needs to feed data at a constant rate to the listener. It is better to accept slightly damaged packets as they are, or even to replace them with silence, a duplicate of the previous packet, or a noise burst. The natural redundancy of voice, together with the high-level error correction procedure in which one participant says"excuse me, someone dropped a glass. Would you please say that again? will handle such dropouts, if they are relatively infrequent However, this strong version of the end-to-end argument is a property of the specific application two people in real-time conversation -rather than a property, say, of speech in general. If one considers instead a speech message system, in which the voice packets are stored in a file fo later listening by the recipient, the arguments suddenly change their nature. Short delays in delivery of packets to the storage medium are not particularly disruptive so there is no longer any objection to low-level reliability measures that might introduce delay in order to achieve reliability. More important, it is actually helpful to this application to get as much accuracy as possible in the recorded message, since the recipient, at the time of listening to the recording, is not going to be able to ask the sender to repeat a sentence. On the other hand, with a storage system acting as the receiving end of the voice communication, an end-to-end argument does must use some care to identify the end points to which the argument should be applied ys. ot an apply to packet ordering and duplicate suppression. Thus the end-to-end argument is no absolute rule, but rather a guideline that helps in application and protocol design anSALTZER ET AL. End-to-End Arguments in System Design 7 Transaction management We have now applied the end-to-end argument in the construction of the SWALLOW distributed data storage system[15], where it leads to significant reduction in overhead. SWALLOW provides data storage servers called repositories that can be used remotely to store and retrieve data. Accessing data at a repository is done by sending it a message specifying the object to be accessed, the version, and type of access (read/write), plus a value to be written if the access is a write. The underlying message communication system does not suppress duplicate messages, since a) the object identifier plus the version information suffices to detect duplicate writes, and b) the effect of a duplicate read request message is only to generate a duplicate response, which is easily discarded by the originator. Consequently, the low-level message communication protocol is significantly simplified. The underlying message communication system does not provide delivery acknowledgement either. The acknowledgement that the originator of a write request needs is that the data was stored safely. This acknowledgement can be provided only by high levels of the SWALLOW system. For read requests, a delivery acknowledgement is redundant, since the response containing the value read is sufficient acknowledgement. By eliminating delivery acknowledgements, the number of messages transmitted is halved. This message reduction can have a significant effect on both host load and network load, improving performance. This same line of reasoning has also been used in development of an experimental protocol for remote access to disk records[6]. The resulting reduction in path length in lower-level protocols was important in maintaining good performance on remote disk access. Identifying the ends Using the end-to-end argument sometimes requires subtlety of analyis of application requirements. For example, consider a computer communication network that carries some packet voice connections, conversations between digital telephone instruments. For those connections that carry voice packets, an unusually strong version of the end-to-end argument applies: if low levels of the communication system try to accomplish bit-perfect communication, they will probably introduce uncontrolled delays in packet delivery, for example, by requesting retransmission of damaged packets and holding up delivery of later packets until earlier ones have been correctly retransmitted. Such delays are disruptive to the voice application, which needs to feed data at a constant rate to the listener. It is better to accept slightly damaged packets as they are, or even to replace them with silence, a duplicate of the previous packet, or a noise burst. The natural redundancy of voice, together with the high-level error correction procedure in which one participant says "excuse me, someone dropped a glass. Would you please say that again?" will handle such dropouts, if they are relatively infrequent. However, this strong version of the end-to-end argument is a property of the specific application – two people in real-time conversation – rather than a property, say, of speech in general. If one considers instead a speech message system, in which the voice packets are stored in a file for later listening by the recipient, the arguments suddenly change their nature. Short delays in delivery of packets to the storage medium are not particularly disruptive so there is no longer any objection to low-level reliability measures that might introduce delay in order to achieve reliability. More important, it is actually helpful to this application to get as much accuracy as possible in the recorded message, since the recipient, at the time of listening to the recording, is not going to be able to ask the sender to repeat a sentence. On the other hand, with a storage system acting as the receiving end of the voice communication, an end-to-end argument does apply to packet ordering and duplicate suppression. Thus the end-to-end argument is not an absolute rule, but rather a guideline that helps in application and protocol design analysis; one must use some care to identify the end points to which the argument should be applied