A. Battou, "Connection Establishment Latency: Measured Results," ATM Forum Document, ATM_Forum/96-1472, Oct. 1996.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....between different threads. The minimum latency for is 22.5, average latency is 24.9, and the maximum latency is 213 ms. The 5 and 95 quantile are 23.1 and 35.1 ms, respectively. Early published connection establishment latency for pointto point call establishment for one hop using UNI signaling [1] has the following performance: minimum latency is 48.99, average latency 53.20, and maximum latency is 67.60 ms. Later improvements bring the latency for call setup latency for a similar LAN environment down to the range between 10 50 ms, depending on the switching software and hardware platform ....

A. Battou, "Connection establishment latency: Measured results," ATM Forum Document, ATM-Forum/96-1472, Oct. 1996.

....new flow generation, new flow test and new flow start up delays to a minimum. In a na ve simulation, 93 the authors may ignore such values: treating them as zero. Although, when compared to a real world implementation, aiming for a delay of zero could be seen as unnecessary several authors [Battou96, Niehaus97] noting that in commercial ATM switch systems the flow setup process for a new flow can take 20 200 ms. Such a quoted value for the delay does not include the additional time required for the end system to become active. In the work to reduce the delay in the AC test environment, the ....

....experiments where every flow attempt is started. It would be unusual for every flow to be accepted in an AC experiment; every flow that was not accepted would only incur a small part of the total delay overhead. When compared with the earlier stated values for commercial flow setup of 20 200 ms [Battou96, Niehaus97] a mean flow setup period of less than 9 ms is excellent and even 24 ms is acceptable. In real terms this means that, in the hypothetical worst: where all flows are admitted, in an experiment of 6000 flows, with 95 certainty, up to 96 of those flows will be adversely affected by ....

A. Battou. Connections Establishment Latency: Measured Results. ATMForum T1A1.3/96-071, October 1996. (pp 94, 95)

....generation, new connection test and new connection start up delays to a minimum; the smaller the delay achieved the closer to the values used in a na ive simulation. When compared to the real world implementation, aiming for a minimal delay could be seen as unnecessary several authors [2, 28] note that in commercial ATM switch systems the call setup process for a new connection can take 20 200ms. Such a quoted value for the delay does not include the additional time required for the end system to become active. In the work to reduce the delay on the CAC test rig, the delay includes ....

....is a useful worst case boundary. The value of 119:3 new connections per second being a maximum for the number of new connections that will be able to be made within this period implies a call set up time of 9ms. When compared with the earlier stated values for commercial call setup of 20 200ms [2, 28] a call setup period of less than 9ms is quite acceptable. In real terms this means that, in the worst case, for an experiment of 6000 connections, less than 18 of those connections will be affected by system delays where the delaying of one connection will delay the next connection entering ....

Abdella Battou. Connections Establishment Latency: Measured Results. ATM-Forum T1A1.3/96-071, October 1996.

....connection generation, new connection test and new connection start up delays to a minimum; the smaller the delay achieved the closer to the values used in a na ve simulation. When compared to the real world implementation, aiming for a minimal delay could be seen as unnecessary several authors [11, 12] note that in commercial ATM switch systems the connection setup process for a new connection can take 20 200ms. Such a quoted value for the delay does not include the additional time required for the endsystem to become active. In our work to reduce the delay in the CAC test rig, this delay ....

....less than this because this full delay impact is incurred only on new connections that are accepted into the system. For new connections that are rejected from the system, the delay is substantially smaller. When compared with the earlier stated values for commercial connection setup of 20 200ms [11, 12] a connection setup period of less than 9ms is quite acceptable. In real terms this means that, in the worst hypothetical case, for an experiment of 6000 connections, less than 18 of those connections will be a ected by system delays where the delaying of one connection will delay the next ....

Abdella Battou. Connections Establishment Latency: Measured Results. ATM-Forum T1A1.3/96-071, October 1996.

....qGSMPd qGSMP daemon 5 14 tency for is 22.5ms, average latency is 24.9ms and the maximum latency is 213ms. The 5 and 95 quantile are 23.1ms and 35.1ms, respectively. Early published connection establishment latency for point to point call establishment for 1 hop using UNI signalling [1] has the following performance: minimum latency is 48.99 ms, average latency 53.20ms and maximum latency is 67.60ms. Later improvements bring the latency for call setup latency for a similar LAN environment down to the range between 10ms and 50ms, depending on the switching software and hardware ....

Abdella Battou, "Connection Establishment Latency: Measured Results," ATM Forum Document, ATM_Forum/96-1472, October 1996.

....details of the various implementation efforts can be found in [23, 24] 12 latency. 25, 26] and more recently [9] have proposed similar techniques. Even, without these optimising techniques the Hollowman compares quite favourably in terms of performance with more conventional signalling system [27, 28], for example point to point connection creation over a single switch with the Hollowman running on Sun Ultras under Solaris 2.5 requires approximately 11 ms. This is equivalent to the best published figures for standards based implementations [28] The results of our experiments indicate that the ....

A. Battou, "Connections Establishment Latency: Measured Results," ATM Fogrum T1A1.3/96-071, October 1996.

....is below 1 ms on a Sparc Station 20. The delay caused by PNNI routing stack is about 0,7 ms. Additional delay experienced by the setup message depends on the operating system and device drivers of a switch and cannot be determined by the PNNI Emulator. Related measurements have been reported in [9]. 5. CONCLUSIONS AND OUTLOOK In this paper, a performance evaluation of PNNI routing mechanisms has been presented. A PNNI Emulator running in a distributed workstation environment has been used to carry out detailed experiments. Although the focus has been on the single peer group model of ....

A. Battou: "Connection Establishment Latency: Measured Results". ATM Forum, Contribution 96-1472, October 1996.

....of the system may actually improve under heavy loading. 4. 5 Performance Comparison With ATM Forum Signalling The call setup delay of our system is compared with that of ATM networks using the ATM Forum UNI PNNI signalling for their end to end signalling, reported in the literature [32] [34]. These values are shown in Table 27. The values measured in our system correspond to 1 and 3 hops connections. Our experiments show that every additional hop in a connection corresponds to an extra VirtualSwitch BIB invocation and it contributes approximately 15 ms to the end to end call setup ....

A. Battou, "Connection Establishment Latency: Measured Results," ATM_Forum/96-1472, Oct. 1996.

....expectation that they will be reused, also with the objective of reducing latency. 24] 25] and more recently [9] have proposed similar techniques. Even without these optimising techniques the Hollowman compares quite favourably in terms of performance with more conventional signalling systems [26], 27] for example point to point connection creation over 8 a single switch with the Hollowman running on Sun Ultras under Solaris 2.5 requires approximately 11 ms. This is equivalent to the best published figures for standards based implementations [27] The results of our experiments ....

A. Battou, "Connections Establishment Latency: Measured Results, " ATM Fogrum T1A1.3/96-071, October 1996.

....Running on a cluster of SUN Sparc 10s, initial results indicate that the system can support a throughput of about 100 calls sec (or 360,000 calls hour) with average latency of 200ms. The intrinsic 16ms call set up time is substantially shorter then the latency data published by the ATM Forum [5]. The caching principle proposed leads to interesting performance questions. How much resource should be cached or, equivalently, how should resources be partitioned Resources here refer to both name space, bandwidth and buffer space. There are essentially two main categories of algorithms that ....

A. Battou, "Connection Establishment Latency: Measured Results," ATM Forum Document, ATM_Forum/96-1472, October 1996.

....needs and environment; this may involve selecting control architectures with different control functions, e.g. mobility, groupcast, third party signalling, or with different non functional requirements, e.g. trading efficiency against reliability. Table I summarises some of the published results [13], 2] 14] for the latency in point to point connection creation across a single switch. Note that direct comparison of the absolute performance numbers is not sensible, because the results were obtained from a number of unrelated experiments. However, it is enough to demonstrate that there is no ....

A. Battou, "Connections Establishment Latency: Measured Results," ATM Forum T1A1.3/96-071, October 1996.

No context found.

A. Battou, "Connection Establishment Latency: Measured Results," ATM Forum Document, ATM_Forum/96-1472, Oct. 1996.

No context found.

Abdella Battou, "Connection Establishment Latency: Measured Results," ATM Forum Document, ATM_Forum/96-1472, October 1996.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC