C. Labovitz, A. Ahuja, and F. Jahanian. Experimental Study of Internet Stability and Backbone Failures. In Proceedings of FTCS, pages 278--285, 1999.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....attacks. Some techniques for evaluating calm day workloads are relatively well understood. Request traces have long been used to benchmark systems. And a number of studies have quantified environmental factors such as hardware, maintenance, and environmental failures [13] Internet failures [18, 15, 9, 1], and Internet performance variability [29] Several recent studies have used faultloads derived from such studies to examine end to end service availability [9, 28] To deepen system understanding under calm day scenarios, additional research is needed. On the request load side, for example, it ....

....of the service. For instance, it would be important to consider the rate at which new content is introduced and how clients access different portions of the service as a function of time. On the environment side, although there is a growing body of publicly available data on network availability [18, 15, 9, 16], additional studies are needed to help Internet service designers understand the range of Internet behavior. In particular, we need traces spanning longer periods of time (e.g. months or years v. days or weeks) resolving finer time granularities (e.g. failure duration resolutions of seconds ....

[Article contains additional citation context not shown here]

C. Labovitz, A. Ahuja, and F. Jahanian. Experimental Study of Internet Stability and Backbone Failures. In FTCS99, June 1999.

....focuses on quantifying the prevalence and diagnosing the causes of IP level failures. Our analysis builds on this study by studying these and other traces to determine metrics relevant to end to end service delivery: failure location and the duration of unavailability events. Labovitz et al. [23] examine route availability by studying routing table update logs. They nd that only 25 to 35 of routes had availability higher than 99.99 and that 10 of routes were available less than 95 of the time. They nd that 60 of failures are repaired in a half hour or less, and that the remaining ....

....Most notably, it represents in middle failures as the interruption of connectivity between a single pair of nodes that does not a ect any other pairs ability to communicate. In reality failures in the middle of the Internet infrastructure will typically a ect more than one pair of nodes [23]. Thus, groups of such middle failures are likely to be correlated, and the correlation will depend on details of network topology that would be complex to model. However, we believe that our simple model provides a reasonable rst order approximation for evaluating routing based techniques: an ....

C. Labovitz, A. Ahuja, and F. Jahanian. Experimental Study of Internet Stability and Backbone Failures. In FTCS99, June 1999.

....than staying connected to a given server. Simple server replication does not address the problem. Network failure or congestion after the connection is established may render the service unavailable to the end user. Studies that quantify the effects of network stability and route availability [12, 8] demonstrate that they can significantly reduce the end to end availability of Internet services. Although highly available servers can be deployed, deploying highly available services remains a problem due to connectivity failures. As server identity tends to become less important than the ....

C. Labovitz, A. Ahuja, and F. Jahanian. Experimental Study of Internet Stability and Backbone Failures. In Proc. 29th Symp. on Fault-Tolerant Computing (FTCS), June 1999.

....However, simple replication with TCP does not address the problem of service continuity after the connection is established. Subsequent network failure or congestion may render the service unavailable to the end user. Studies that quantify the e ects of network stability and route availability [17, 9] demonstrate that connectivity failures can signi cantly reduce the end to end availability of Internet services. Although highly available servers can be deployed, deploying highly available services remains a problem due to connectivity failures that may lead to disconnections between a typical ....

C. Labovitz, A. Ahuja, and F. Jahanian. Experimental Study of Internet Stability and Backbone Failures. In Proc. 29th Int'l Symp. on Fault-Tolerant Computing (FTCS), June 1999.

.... of failures, software about 20 , and system management a bit over 50 [15] Our study is similar in spirit to two studies of network failure causes: Mahajan examined the causes of BGP misconfigurations [13] and Labovitz conducted an earlier study of the general causes of failure in IP backbones [11]. 8. Conclusion From a study of more than 500 component failures and dozens of user visible failures in three large scale Internet services, we observe that (1) operator error is the leading cause of failure in two of the three services studied, 2) operator error is the largest contributor to ....

C. Labovitz, A. Ahuja, and F. Jahanian. Experimental study of Internet stability and backbone failures. Fault-Tolerant Computing Symposium (FTCS), 1999.

....victim. In this case, HCF will not be able to identify every spoofed packet. However, we will show in Section 5 that even with a limited range of hop count values, HCF can be highly effective at identifying spoofed IP addresses. 4 Feasibility of Hop Count Filtering Recent Internet experiments [22, 33] have shown that, despite the large number of routing updates, 1) a large fraction of destination prefixes have remarkably stable Border Gateway Protocol (BGP) routes, 2) popular prefixes tend to have stable BGP routes for days or weeks; and (3) a vast majority of BGP instability stems from a ....

C. Labovitz, A. Ahuja, and F. Jahanian. Experimental study of internet stability and backbone failures. In Proccedings of the 29th International Symposium on Fault-Tolerant Computing, Madison, WI, June 1999.

....undesirable. Addressing this problem on a best effort basis inside NAVTECH is not only possible but straightforward to implement. Whilst IPv6 is bound to improve routing responsiveness, connectivity glitches still last longer than desirable, for example due to the load queues that build up[24]. The above notion of partition can be extended to several sets of partitions. When the ## has additional information on the state of the network, provided by the ## for example, it may be possible to correlate failure of GlobalNet links to the defaulting route, in case of partial partition, or to ....

C. Labovitz, A. Ahuja, and F. Jahanian. Experimental study of internet stability and backbone failures. In Digest of Papers, The 29th Annual International Symposium on Fault-Tolerant Computing, Madison, USA, June 1999. IEEE Computer Society.

....focuses on quantifying the prevalence and diagnosing the causes of IP level failures. Our analysis builds on this study by studying these and other traces to determine metrics relevant to end to end service delivery: failure location and the duration of unavailability events. Labovitz et al. [23] examine route availability by studying routing table update logs. They find that only 25 to 35 of routes had availability higher than 99.99 and that 10 of routes were available less than 95 of the time. They find that 60 of failures are repaired in a half hour or less, and that the ....

C. Labovitz, A. Ahuja, and F. Jahanian. Experimental Study of Internet Stability and Backbone Failures. In FTCS99, June 1999.

....3.4 of the time [18] This study focuses on quantifying the prevailance and diagnosing the causes of IP level failures. Our analysis builds on this study by studying these traces to determine metrics relevant to endto end service delivery: failure location and failure duration. Labovitz et al. [13] examine route availability by studying routing table update logs. They find that only 25 to 35 of routes had availability higher than 99.99 and that 10 of routes were available less than 95 of the time. They find that 60 of failures are repaired in a half hour or less, and that the ....

....Most notably, it represents in middle failures as the interruption of connectivity between a single pair of nodes that does not affect any other pairs ability to communicate. In reality failures in the middle of the Internet infrastructure will typically affect more than one pair of nodes [13]. Thus, groups of such middle failures are likely to be correlated, and the correlation will depend on details of network topology that would be complex to model. However, we believe that our simple model provides a reasonable first order approximation for evaluating routing based techniques: an ....

C. Labovitz, A. Ahuja, and F. Jahanian. Experimental Study of Internet Stability and Backbone Failures. In FTCS99, June 1999.

....1.5 to 3.4 of the time [25] The study focuses on quantifying the prevalence and diagnosing the causes of IP level failures. Our analysis builds on this study by studying these traces to determine metrics relevant to end to end service delivery: failure location and duration. Labovitz et al. [19] examine route availability by studying routing table update logs. They nd that only 25 to 35 of routes had availability higher than 99.99 and that 10 of routes were available less than 95 of the time. They nd that 60 of failures are repaired in a half hour or less, and that the remaining ....

....Most notably, it represents in middle failures as the interruption of connectivity between a single pair of nodes that does not a ect any other pairs ability to communicate. In reality failures in the middle of the Internet infrastructure will typically a ect more than one pair of nodes [19]. Thus, groups of such middle failures are likely to be correlated, and the correlation will depend on details of network topology that would be complex to model. However, we believe that our simple model provides a reasonable rst order approximation 3 Parameter Default value Comment Rate 1.5 ....

C. Labovitz, A. Ahuja, and F. Jahanian. Experimental Study of Internet Stability and Backbone Failures. In FTCS99, June 1999.

....handling may be common to a number of higher layer network services. d) Fault tolerance on a layer can only handle fault at its level and on the levels below. Hence, some fault tolerance is required at the highest level. A closing question: with the observations of the current internet in mind [33, 22], is the currently so popular IP over light scenario able to shortcut todays multilevel fault handling and provide a fast enough reconguration and a stable rerouting in a large network where failures are the rule more than the exception Table 1. Summary of resource handling issues for the ....

Craig Labovitz, Abha Ahuja, Farnam Jahanian, Experimental Study of Internet Stability and Backbone Failures, Proc. 29th International Symposium on Fault-Tolerant Computing (FTCS-29), pp. 278 - 285, June 1999.

....or to the composition of P out , in the case of total partition. This is not only possible but straightforward to implement in the Internet. Whilst IPv6 is bound to improve routing responsiveness, connectivity glitches may last longer than expected, for example due to the load queues that build up[54]. As an example of how modules of the architecture can use the run time environment, the SF can use TTCB services to help improve the assessment of a problem, namely to determine whether it is a partition or a site failure, as it can reliable detect the failure of sites by detecting failure of ....

Craig Labovitz, Abha Ahuja, and Farnam Jahanian. Experimental study of internet stability and backbone failures. In Proceedings of the 29th International Symposium on Fault- Tolerant Computing (FTCS 29), Madison, USA, 1999.

....volume of BGP routing message was several orders of magnitude more than expected and a majority of the information was redundant and pathological. In a subsequent paper [13] they identi ed reasons behind many of those unexpected routing messages and described remedies for them. Labovitz et al. [11, 12] have also shown that signi cant correlation exists between network usage and instability observed for BGP. Varadhan et al. 21] showed that there are routing policies that can cause BGP routes to oscillate and never converge to a stable con guration. Govindan et al. 6] subsequently described an ....

....of scheduling, queueing, and bu ering mechanisms in order to improve the stability of routing protocols. We plan to extend the work in this paper by making the traf c and loss models more realistic. We also plan to tie these results with actual measurements in the Internet. Labovitz et al. [11, 12] have shown that there is a signi cant correlation between the measured BGP instability and network usage. It will be interesting to see what fraction of these instabilities is due to congestion and routing packet losses. We also plan to extend our models to include other deployed routing ....

C. Labovitz, A. Ahuja, and F. Jahanian. Experimental Study of Internet Stability and Backbone Failures. In Proceedings of the 29th Annual International Symposium on Fault-Tolerant Computing, June 1999.

No context found.

C. Labovitz, A. Ahuja, and F. Jahanian. Experimental Study of Internet Stability and Backbone Failures. In Proceedings of FTCS, pages 278--285, 1999.

No context found.

C. Labovitz, A. Ahuja, and F. Jahanian, "Experimental Study of Internet Stability and Backbone Failures", Proceedings of 29th International Symposium on Fault-Tolerant Computing (FTCS), Madison, Wisconsin, pp. 278-285, June 1999.

No context found.

C. Labovitz, A. Ahuja, F. Jahanian. Experimental Study of Internet Stability and Backbone Failures. Proc. of FTCS-29, Wisconsin. 1999.

No context found.

C. Labovitz, A. Ahuja, F. Jahanian. Experimental Study of Internet Stability and Backbone Failures. In FTCS-29, Madison, USA. IEEE CS Press, 1999.

No context found.

C. Labovitz, A. Ahuja, and F. Jahanian. Experimental study of Internet stability and backbone failures. In Proceedings of IEEE International Symposium on Fault-Tolerant Computing (FTCS), Madison, WI, June 1999.

No context found.

C. Labovitz, A. Ahuja, and F. Jahanian. Experimental study of Internet stability and backbone failures. In Proceedings of IEEE International Symposium on Fault-Tolerant Computing (FTCS), Madison, WI, June 1999.

No context found.

C. Labovitz, A. Ahuja, and F. Jahanian. Experimental Study of Internet Stability and Backbone Failures. In FTCS99, June 1999.

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