Phone Call to Mars: Why Application Latency Matters
Some of us may remember the time when long distance calls across the ocean were often accompanied by an annoying delay between the time we finished a sentence and the other party replying. We would say, “We're probably connected via a satellite. I like it better when these calls are routed through an undersea cable.”
This half-second delay was caused by the signal traveling 80,000 km at the speed of light to and from a geosynchronous satellite. In technical terms, this delay is called latency. And if you think that those satellite calls were bad, think about the call latency to your significant other on assignment in a future Mars colony! It will be about three minutes, and there won’t be a damned thing you can do about it.
At the other extreme, traders on the world’s major stock exchanges frequently place their servers, which run automated trading systems, as physically close as possible to the exchange’s servers to get an edge on their competitors. Every hundred meters closer gives them about an extra microsecond, which can make all the difference in placing an order first.
Up till now, call latency has not been a major concern. Today, we are connected by a global mesh of fiber optics, and the latency of even a trans-ocean connection is approximately 50 milliseconds. This
has no discernible effect on voice calls or on interaction with Internet applications like hosted services, Web surfing or storage backup.
One side effect is that internet applications can be delivered “over the top” of existing communications service provider (CSP) networks, without a discernible impact on customer experience. The downside of this of course, is that networks simply provide the pipes between the end-users and the data-centers where the Internet applications are hosted, but do not participate in the revenue streams from the Internet services.
However, this situation is beginning to change with the emergence of a new and expanding set of machine-to-machine (M2M) based communication applications. These broadly fit under the umbrella of the Internet of Things (IoT), and are also starting to form distinct categories like the Smart City (e.g. for traffic monitoring and light control), and Augmented Reality, where what you see and do is augmented by other computer-based information and controls.
In these applications, latency most definitely does matter. This may not be to the same degree as high-frequency electronic trading, but there is a requirement for response times to be within the millisecond range. Practically, this means you must deploy the supporting servers less than 20 kilometers from the intelligent devices with which they are communicating.
Here lies the potential to open up a world of new opportunities for CSPs, who, for totally different reasons, are starting to deploy reconfigurable computing technology at the edge of their networks to support telecommunication services. CSPs can start coupling their network-edge server technology with the tens of thousands of micro data center platforms (technically called hyper-converged clouds), which need to be deployed for M2M and IoT applications. So, all of a sudden, latency does matter again, and this time it looks like it can work in the favor of CSP business.
For more information about how to take advantage of this new CSP opportunity, contact ECI and discover the benefits of the ELASTIC Network™.