The Stack Archive

Optimising performance to beat the internet’s rush hour traffic

Mon 28 Jul 2014

As IP traffic continues to rocket and the peaks during the busy hour get even spikier, there is a growing need for automation in managing network services says Paul Vian

According to the most recent Cisco Visual Networking Index,  global IP traffic has increased more than fivefold in the past five years and will increase threefold over the next five years. Moreover, traffic during the “busy hour,” or the busiest 60 minute period in a day, is growing 20% faster than that. While cloud services and mobile devices are enabling dynamic new IT environments for businesses, the networks connecting them are emerging as critical pain points in service delivery and user experience.

At the macro level, if the Netflix peering disputes with Comcast and Verizon demonstrate anything, it’s that even if you have a very fast broadband connection to the internet, there are no guarantees you will automatically get a better user experience.

The same holds true for businesses that increasingly rely on the internet, from online gaming companies, e-commerce and big data analytics to mobile and digital advertising platforms and exchanges. Latency, jitter, packet loss and congestion all lead to poor internet performance that has a negative impact on business. E-commerce provides a very tangible example: Summit, the online marketing and e-commerce specialist for retail, recently estimated that a retailer with £10m in annual online sales could lose up to £1m in revenue each year due to consumer impatience with slow-loading web pages.

Part of that problem is that networks are dynamic and the performance of any network services provider (NSP) varies based on its overall capacity, traffic at any point in time, network black-outs/brown-outs and the number of hops required across the network to reach the destination. Combining these variables ensures that no single NSP can provide the best performance all the time – especially with the number of destination networks on the internet increasing dramatically, quadrupling in size since 2004 (see chart below).

To try and tackle this problem, many companies buy bandwidth from multiple NSPs, but they have to make their own traffic routing changes. This is a manually-intensive process with each routing change taking approximately five minutes to complete; as NSP performance can change instantly, it’s next to impossible to fully optimise traffic this way. Automation is key.

There are tools that can use an algorithm to dynamically evaluate performance across several NSPs and route traffic over the best performing path at any given time. Such intelligent software can make millions of calibrations daily to avoid congested networks and, compared to any single carrier, it is capable of up to four times less fluctuation in performance, resulting in a more stable, reliable connection and better overall application performance. To put it into perspective – it would require more than 10,000 man-days of effort to make 1 million optimisations manually.

Technology has evolved to give us more power. Just as the experience of rush-hour driving has improved with fully automated road-side and satellite data collection and analysis that feeds into our ‘sat navs’ giving us near real-time information. Now we can manage our internet experience to get maximum performance through near real-time optimised routing.

Paul Vian is business development director at Internap 

Tags:

analytics broadband Cisco Content Delivery Network (CDN) feature internet Netflix Verizon

---