India, the fastest growing mobile market in the world, is the focus of the latest Twin Prime GLAS Insights report.
Our GLAS (Global Location-based Acceleration Strategies) reports analyse the millions of requests coming to Twin Prime and showcase interesting insights about network performance. Understanding which factors impact performance are key to dealing with mobile performance issues--our data exemplifies how variability determines mobile performance.
We analysed hundreds of millions of measurements related to network performance in India and found that LTE is off to a great start. Hyderabad, in particular, deserves its "Cyberabad" nickname!
Recent reports forecast a significant growth in LTE devices in India, which is great news for customers looking to experience rich media such as videos. This is a great opportunity and a challenge for India's booming mobile businesses--in sectors as diverse as m-Commerce, travel and dating--that must deal with incredible variability in how their end-users download content across different networks.
A new approach to mobile content delivery that deals with this variability head-on is a critical necessity in India, if such businesses are to realise their potential (for more on this subject, read the Frost & Sullivan white paper on A Billion Reasons for Inconsistent Mobile Performance And How To Solve For Them).
Impending LTE network upgrade may outperform WiFi
Most people will say they perceive WiFi to be better--faster and more reliable--than cellular networks. Our first Insights blog showed that this is not the case with LTE in many cities globally. How about India?
In India, LTE adoption is just beginning to ramp up--the share of LTE traffic among requests seen by Twin Prime in India is around 2%. But if early trends are an indicator, LTE performs better than WiFi in major metros. Hyderabad lives up to its reputation and beats other metros.
The speed chart presented above computes median download speeds for files in the range of 30KB-60KB (a popular choice of size for downloaded content in India) across hundreds of millions of samples.
Cellular experience is more consistent and getting better
One of the key issues affecting perceived performance is the variability experienced by users. When we examine how variable download completion times are across network technologies, we see a clear trend: cellular performance is getting more consistent with each revision and LTE is certainly the best. In this chart, a higher number means users experience a wider range in performance--good and bad. In some ways that may make the experience seem erratic. Our measure of variability is a ratio of how the 75th percentile download completion time compares to the median across tens of millions of samples.
Note that this trend with Indian networks is consistent with what we see with global data.
With LTE, India could be ready for video on cellular
A good quality video (say, 480p) requires 1.5 Mbps download speeds over large file sizes. Analysing about a million samples leads us to conclude that LTE is well positioned to deliver video. The download speeds presented here are for files larger than 200KB since we are interested in sustained speeds achieved over longer connection durations.
Is there an ideal image size for the latency conscious?
Given such wide variations across networks in India, app developers want to watch their latency budget. If we wanted our file to download within two seconds in most scenarios, how big should it be?
In short, not much. Download times vary so much that one would have to limit the file sizes to low levels to achieve these limits. This is a challenge that app developers need to surmount--
how to deal with this huge variability while not reducing content sizes to a lowest common denominator.
Examining millions of samples, we saw one surprising result. With LTE we observed downloads of up to 80KB landing within the two second mark. In other words, the 75th percentile of the download completion time was less than two seconds up to this size. This is intuitive if we recall the previous chart that showed much lower variability with LTE. With WiFi, the higher variability means that it is harder to guarantee a particular download completion time.
Clearly LTE has started on a good note. With broader adoption, we will watch how these metrics hold up--one would expect changes simply due to how loaded networks would get over time with more usage.