Google-Like Response Times with CIS-Net?
At inception, some 10 years ago, CIS-Net was designed using what were then very innovative concepts in “Computer Networking.” Since that time, the trend has caught up with us and the IT industry came up with buzzwords such as “Web Services” or “Service- Oriented Architecture” to speak about the ability to inter-operate between systems and distribute essential business processes and information among societies.
This model has demonstrated its inherent qualities but, in the meantime, has also demonstrated some of its limits. One aspect that has been particularly under stress these past few years is response times.
Over time, not only have the numbers of users and searches increased, the locations from which CIS-Net is operated have also grown, now covering nearly the entire planet. This causes two main problems for the performance of the network.
Firstly, since CIS-Net relies on the Internet to distribute its components, it suffers from the same “high latency” problems between two end-points that are geographically distant from one another. Typical examples are connections between Europe (where most of the CIS-Net components are actually located) and the Asia-Pacific region, whereby propagation delays are mostly due to extremely long network routes, punctuated with overcrowded hardware infrastructures, which are often bottlenecks. The only way to reduce those latencies is to optimize the route and the most efficient way to do that is just to get physically closer. How is this possible? One solution that is in the pipeline as part of the efforts conducted by CISAC’s Performance Working Group, chaired by Fabrizio Zavagli (SIAE), is the so-called “hybrid cache.” Combined with regional hosting located as close as possible to the end users, the cache will provide CIS-Net with passive replications of node information and therefore reduce the latency by shortening the route while at the same balancing the load from the source node to the cache. A first pilot of the cache system is currently being tested.
Secondly, unlike contributing societies, CIS-Net is up and running around the clock. This doesn’t initially sound like a performance problem, but on second thought, if a problem occurs at a node during off-hours (locally), chances are that this problem won’t be resolved before the next business day (again locally). For instance, when a node goes down in Europe just after business hours — so shortly before start of business day in Asia — it will probably remain unavailable until the next morning, and Asian societies will lose a complete day of work with that node. The hybrid nature of the cache will allow us to automatically switch to the cache whenever a node is monitored as being out of order and users will seamlessly retrieve the information they are looking for from ne source or the other.
Better watch out, Google…
FastTrack Technical Director
« Back to Home