For gamers located across Eurasia, servers in the wrong place can be matter of virtual life and death. For enterprises, a laggy virtual desktop is a torture instrument for employees. So imagine you are building data centres across Russia and Central Asia, and are working hard to attract business for these applications.
First you need to show that there is a real customer experience from hosting in-region, rather than in servicing those markets from Europe or the Far East. You also need to know where to locate them, so as to offer the best provide Internet connectivity routes, ones that are capable of the performance needed by your customers.
This is the situation faced by IXcellerate, a major commercial data centre in Moscow. They had hired colleague Pete Cladingbowl of Skonzo Ltd, who offer cloud infrastructure design expertise. He in turn sought help from my associates at Predictable Network Solutions Ltd, who are specialists in high-fidelity network measurement.
The resulting data is one of the most accurate measurements of the Internet ever undertaken and publicly published. You can think of it as the difference between a typical prod from the network doctor asking “does it hurt here?” versus a full-blown functional MRI scan of the system’s dynamic properties.
The measurements capture the instantaneous network performance properties, i.e. they measure ∆Q. You can decompose ∆Q metrics in space (where on the path did the loss and delay accumulate?), as well as in time (what were the causes of the loss and delay?). The latter breaks into three components: geographic (G), packet serialisation (S) and variable contention (V) delay.
The resulting data shows how:
- Some routes are highly indirect (double the “G”) compared to an “as the crow flies” fiber.
- More bandwidth (lower “S”) would have little effect.
- Congestion due to load (the “V”) could have very different effects depending on the route.
A way of interpreting the data is to imagine all times (for G, S and V) are scaled as distance at the speed of light in fiber. The indirect routing from Moscow to Frankfurt, for example, would mean the performance is comparable to a direct route to Morocco!
The bottom line is that some applications would only have satisfactory performance on certain routes. The data from the measurements can be used to ensure the service is sold to the right customers (and not oversold). It can also be used to drive future investment plans, and to help manage the traffic better through scheduling.
Whilst there are many people taking network measurements, the ability to take high-fidelity ones is new and offers a competitive advantage. It gives a robust view into what the true customer experience would be. It also tells you the underlying causes of the (good or bad) customer experience. Without this data, you are running your cloud or network business with blurry vision in the dark.
The data from this state-of-the-art measurement was presented at the Eurasian Network Operator’s Group meeting in Kazan, Russia on 9-10 June 2015. You can view it on SlideShare by clicking the link below.
Show me the Internet quality between Russia and Europe
For the latest fresh thinking on telecommunications, please sign up for the free Geddes newsletter.
Read It Now
