News Feed
Lighting in the Cloud: Low-latency Applications and the Role of Network Operators: Specific point-to-point and low latency services within the cloud: who’s buying? Banks and financial services. Banks have had latency requirements, now from International perspective we’re more concerned with latency. Friends with PhDs in engineering all being hired by large banks. Iceberg transactions, not detected by other traders. It’s a science (black art?) in forex and commodities trading, about making decisions faster than competitors. In software is buying/selling transactions, size is indeterminate. It’s about the speed (in milliseconds). If software isn’t working, don’t switch; it’s not worth having.
What kind of premium are these firms paying? Solutions are optimized for route and circuits, between co-location services and with managed services, other parts of portfolio of services, commands a 20-30% premium for speed and reliability (a premium route). They’re buying customer care, sometimes over specific routes and local loops. Capacity: 100M ethernet, they don’t want variability of switching services or other points in network. Besides guarantee on availability, they also look at packet loss, other variables; screen local providers for reliability. “When it works, they want 99.99x reliability; otherwise I don’t want it.” They get upset if service goes below a specific level.
In Asia, predominant locations are Korea, Shanghai, big financial centers. Larger market for high speed network requirements drive us as cable builders to focus on short routes, low latency, sub-sea operators. Similar in capacity requirements to last panel: optimal latency (but on more precise level).
When talking about cloud, how do you optimize? New cables being built, older routes variable and not always optimal. Cloud is not that new, but for carrier: first need network to deliver services, next can open for engineering, then how to marry network with co-location facilities.
Physical resiliency of networks: in Asia it’s a particular concern (man-made patches, natural disasters). How do you respond to that? We had a mesh topology in place so it was transparent for our users. We lost lots of resiliency, how to get around new breaks? Lately, cables need to be near, fast and straight. Each operator has a choice to decide where their mesh network is built (create consortiums, other approaches to network). Asia got recognized because it’s a bunch of islands. It’s going to be a lot of sub-sea cables connecting them, environmental effects significant. Build network that can withstand uncertainty. Now customers are wanting a route that avoids certain regions (as single points of failure to avoid). Customers are getting smarter, and have large network teams that can help identify, avoid the risks and alternatives.
Single worse place in last 18 months to operate: Singapore (large # of cable breaks), also areas where political systems just turn the network off, dynamite fishing, etc.
What other kind of institutions other than banking? Large enterprises like Microsoft (very specific path), other companies. The more computing power you put in the cloud, the more demand/requirements you’ll have for high speed. Gambling (peaks and long non-use).
Questions: one area where it’s critical: last mile in developing countries. Single operators don’t own last mile. How do you address peering problem? It’s about resiliency and you get what you pay for. Price per risk assessment. Options are available in most countries to have multiple sites.
News Feed