Round-trip delay (or latency) is the time taken for a packet to travel from one host on the network, to a remote host and back to the initiating host. Round-trip loss is defined as the percentage of these measurements for which a return packet is never received by the initiating host (and is therefore deemed to have been lost). Routine monitoring of round-trip packet delay and round-trip packet loss across the core network of an IP Service Provider is a crucial function for any network operator. Such round-trip delay/loss statistics are key performance indicators and are of great interest, to both the network operator and to its customer's alike. Any variations in round-trip delay and/or sporadic and transient packet loss are indicative of network congestion, and could lead to degradation in the performance of a customer's applications.

How the measurements are made

To generate the data presented on this website, dedicated monitoring hardware has been deployed at key PoPs within the BT Global Services IP backbone network. Each monitoring station is constantly checking the round-trip delay/loss by sending 64 byte packets (several hundred times an hour) to all the other monitoring stations in our network, and to monitoring stations located within the networks of our strategic partners. The results from all the distributed monitoring stations are collected in a central database which processes the raw data and produces the statistics shown on this website.

The origins of round-trip delay and loss

A key part of the measured round-trip delay is essentially static in nature and is due to factors such as: the time of flight of packets from A to B the rate at which an unloaded router can forward packets between the input interface and the output interface. the speed of all the network circuits on the route between the two end hosts. Such delays are unavoidable. However there is also a component to the round-trip delay that exhibits a more dynamic behaviour and arises due to packets queuing up to be processed and forwarded by each router along the path. Such a forwarding delay at any one router is obviously dependent of the number of other packets which are waiting to be forwarded by that router. If the network load is particularly high and too many packets require forwarding at any particular router, then that router will simply have to discard some of those packets (according to a set of pre-defined rules). Such queuing delays (and queuing losses) are strongly dependent on the traffic load on the network. Here at BT Global Services, we are constantly monitoring the backbone network performance and aim to maintain such delays at a level which has minimal impact on the performance of a customer's applications.

The pages on this website

The Hourly, Daily and Monthly Network Details pages present the hourly, daily and monthly average city-to-city round-trip delay and percentage round-trip packet loss between key PoPs within our IP backbone network. There is then a separate set of pages, the Hourly, Daily and Monthly City Details pages, which presents detailed delay/loss statistics for a larger number of cities; some of which are located on our IP Backbone network, and some of which are located on the backbone networks of our strategic partners. The Hourly, Daily and Monthly Network Summary pages present a view of the network performance for the last few hours, days and months respectively by summarising the detailed results shown on the corresponding hourly, daily and monthly details pages in terms of inter-regional averages. For the purpose of these pages, the following inter-regional definitions apply: an Intra-Europe result is defined as the average of the results between the eleven key European PoPs. a Transatlantic result is the average of the New-York (NY) to London (Lo) result and the Washington (Wa) to Amsterdam (Am) result. a Europe-USA result is the average of the results between the two main European PoPs, namely London (Lo) and Amsterdam (Am), and the six USA Nodes. a Europe-Latin result is the average of the results between the two main European PoPs, namely London (Lo) and Amsterdam (Am), and the seven Latin America Nodes. a Europe-Asia result is the average of the results between the two main European PoPs, namely London (Lo) and Amsterdam (Am), and the fourteen main Asia Pacific Nodes. On all the pages on this website, the colour coding of the boxes indicates the degree of acceptability of the result. A green box is the normal colour for any measurement result, and represents a satisfactory delay/loss performance. An orange box indicates a round-trip performance that is showing a noticeable degradation from its baseline performance. A red box indicates performance that we view as unacceptable. For the city-to-city delay results, the green-to-orange and the orange-to-red transitions are set individually on the basis of the known baseline performance on each of those routes. For the city-to-city loss results, the green-to-orange and the orange-to-red transitions are set to occur at 1% and at 5% packets loss respectively. However for the inter-regional averages the green-to-orange and the orange-to-red transitions are set to occur at 1% and 2% packet loss respectively. You can be sure that on the rare occasions that a red box appears on these pages our representatives will be doing everything in their power to rectify the problem. Occasional technical difficulties mean that measurement data may not always be available in which case the relevant box will be marked as NA - not available.