This Week’s Highlights
- Twitter in the Social Networks category takes the overall title this week with a CASC score of 9.77.
- UK in the UK Government is the fastest API this week with a median latency of 71 ms.
- Environmental Systems Research Institute in the US Government category takes the overall bottom spot again this week with a CASC score of 5.59, down from 5.95 last week.
- We take a good, hard look at what Twitter is doing so well and where it can still improve.
Something of interest
Twitter tops the Social Networks category this week and takes the overall crown with an excellent CASC score of 9.77. What’s it doing right? And are there any areas it could improve? 
What’s interesting in terms of Total Time of the API is that every region except South Asia has two or more good clouds. This is why you need to be actively monitoring your APIs. Twitter is a global service. Its users in South Asia should be receiving as good an API service as those in other regions. And it is noteworthy that though the fastest location is Azure North America, in all other regions other Europe, Azure is by far the slowest cloud. Many Twitter users are going to accessing the Twitter API via that popular cloud and are going to encounter hundreds of additional latency in some cases. That’s not acceptable. So optimizing network performance between clouds and regions is a priority, especially given that Twitter does have quick locations in what are often slow regions (East Asia, Oceania, South America and South Asia). If they can do it from some locations, they should be able to do it from all. 
It’s not DNS though. Twitter does very well here with aa DNS Lookup Time of 4 ms or 12 ms everywhere. 
We see a big issue though with TCP Connection Time for Azure, where it is 125 ms or more from East Asia, Oceania, South America and Oceania. Something is clearly out of kilter here as the TCP Connection Time outside South Asia is everywhere 29 ms or less and in 8 out of 22 locations 8 ms or less. Surely, something could be done to optimize performance for this component for AWS. 
We see exactly the same pattern with Azure for Handshake Time. As with TCP Connection Time, it’s not just Azure in Oceania, it’s IBM Cloud too. It’s by monitoring your API from the end user perspective that you can see exactly performance users are experiencing from different location. Everything looks great from AWS and Google. But what about Azure and IBM Cloud? 
Process Time interestingly doesn’t show much variation by cloud, but it does by region. North America is very much the quickest here and South Asia the slowest, although by nothing like as much as the extra latency Azure user are experiencing. But it’s clear that though Twitter is a first rate API and has some latency components are well-optimized such as DNS Lookup Time, there’s definitely work to be done in understanding why there are major issues with TCP Connection Time and Handshake Time for Azure (and IBM Cloud) in some regions. 
API Analysis: Tops in Overall Quality
Week Ending 3 May 2021
Category
Organization
CASC score
Corporate Infrastructure
GitHub
9.46
COVID-19
covid19api.com