Wow! The Philippines just overtook Vietnam, Indonesia and Cambodia in Internet speeds. 22 million records formed the basis of the analysis. Chris Ritzo, MLabs supplied the data. And Ms Grace Mirandilla Santos helped with our analysis. Thanks!
Why and who made this happen?
This graph below might answer this. It shows the bandwidth speeds by types of organizations. Public entities like Preginet, DOST, DOF, IRRI, UP Diliman and others are tagged as [Public]. PLDT and GLOBE are tagged as [TELCO]. All other ISPs and networks are labeled as [PRIVATE]. [NULL] tags refer to ISPs/networks that are NOT peered with PHOpenIX.
[PUBLIC] organizations are the ones that powered the bandwidth increases. [TELCO]s and [PRIVATE] orgs showed modest speed increases. While the NON-peered networks [NULL] dragged down the average speeds.
You can examine this from another perspective – Peered and non-peered traffic.
This graph segregates the bandwidth into two. One line charts the bandwidth speeds of peered networks. And another line charts the bandwidth speeds of non-peered networks.
You will notice that peered ISPs are generally faster. But why is peering faster?
Peering or Internet exchange points are physical connections. These connections enable networks to exchange traffic among themselves. They are typically made up of ISPs and Content providers.
Peering points are like bridges. They connect your place to your neighbors across the river. Without those bridges, you need to travel farther using a ‘longer’ route to get to your neighbors. Similarly, in the Internet world, peering provides these ‘shortcuts’. It results in shorter travel time among peered networks. You get better internet when both parties are on peered networks.
If you manage a network, please check out the “PhNOG 2017: Our Philippine Internet”. The registration link is here: https://www.eventbrite.com/e/phnog-conference-2017-our-philippine-internet-tickets-31115373921
Oh by the way, did you ask your internet provider if they are peered yet?
My thanks to Mr Joseph Tabadero Jr for debugging this piece of R code that I used with Tableau. The R code detects and filters out anomalies in the MLab data. For the benefit of fellow Tableau and R users, here is the code:
a <- rep(1, length( as.numeric(na.omit(.arg1))))
a[findpeaks( as.numeric(na.omit(.arg1)),threshold=quantile( as.numeric(na.omit(.arg1)),.95),sortstr=FALSE)[,2]]=0
AVG([datafieldname])) = 0