NodeJS, the MEAN stack and Javascript in general are the new hot kids on the block. Everyone’s raving and companies are developing strategies.
What NodeJS did right is use Non-Blocking IO.
What is Non-Blocking IO?
When I go to www.wikipedia.org the following happens:
- My computer connects with the kind servers at WikiPedia’s impressive offices and says “Hey Mustafa wants to access your site”
- WikiPedia being a non-biased organization says “Yea sure. Come on in.” This response and the decision making for this response is done by some of their CDN/LoadBalancer/Web Server programs.And thus a connection is established between us.
- What happens next is totally dependant on what I asked for!Let’s say I requested a *.gif image found on their site … the response will just be a rendering image in my browser. Fairly simple.However, if I had requested one of their usual wiki articles. The following would have to happen:
- The web server would pass along the request by me to their PHP based wiki application with the URL, and other server collected details about me.
- The PHP application would now be in control. It can either also continue to give me a green light and render the content I requested; or decline my request and thus end our connection.Let’s assume the PHP app doesn’t decline and decides to process my request.
- The PHP app will execute several queries to the database server(s) to retrieve the article data I requested. Assuming the database server has the requested data, it will send back the data to the PHP app.
- The PHP application would compose the final HTML page with the all the retrieved data and send it back to the webserver with an “All clear”.
- The web server would then send me back the rendered HTML page in my browser where I would see it.
Each of the above steps take some time. Let’s say the whole process from steps 1 to 8 takes 1.5 seconds; step 6 can take 0.45 seconds of that time.
That’s a large portion of the total process!
In PHP (and most other server side technologies) this entire chain of events described above is blocking and synchronous.
Only when step 1 is complete, step 2 can commence and the sequence goes on to the end.
Web apps usually have excessive communication with their database servers. These back-and-forth communications between the server-side app and database servers has latency costs and account for a noticeable chunk of the entire response time of any web app.
So someone in the NodeJS founding team probably said to themselves “Hey why don’t we come up with a way to make the input/output step non-blocking to the rest of the app’s processes?”
So step 6 in the above workflow; if we make it non-blocking and take it out of the sequence of events necessary to complete the HTTP request; would cut down the entire app’s response time by 30% for the above mentioned request.
That’s what Non-Blocking I/O (input/output) is! Step 6 just occurs asynchronously. Once the app completes step 6 separately, it sends the response back via a callback to the frontend Javascript app (using AJAX); allowing the app to run more smoothly.
Is NodeJS perfect? I don’t think so.
However they analyzed the workflow of web apps carefully and came up with an innovative solution to I/O latency. So bravo to them!
The PHP community (as well as all other tech stacks) should also try to come up with their own solutions to the I/O latency, as it is a real thing.
At Web Dotz, we are trying to come up with our own innovative solution to this problem: Intuitive I/O.
By Mustafa Ghayyur
July 5th, 2018
