If you were designing a new website, how would you plan and test for scale?

Question

Anonymous · Accepted Answer

Let us say I am creating a website where you do product reviews, blogger style.
Anyone who has a login can enter add a product, it pictures, type and
classification etc. and write a detail review. This will then go through
profanity/sanity check etc. and then gets published. Oter users can publish
comments on the site. Users can request reviews for products on the site as
well. Initially we can plan this without any load balancing and then analyze
traffic to understand where the requests are coming from and then host
accordingly. As the website becomes much popular, this can be re re-analyzed and
planned to be hosted from multiple regions with load balancing in place. Plan
for tracking metrics like latency, upload speeds etc to make sure that the
website is within thresholds. Also use metrics from comparable sites to
calculate scale for instances like surges or growth

Anonymous · Answer

I would start with adding a Analytic tool to my website that would detect the
number of users using the website from each geographic location. It would also
track each of the events that the user performs as part of the entire user
journey on the website. We would get to know which are the most used features on
the site and which are the least used once. This could give us an idea if the
most used features can be efficiently enhanced internally first. This could
involve engineering tactics like efficient query optimisation decreasing the
load time . Also tracking the number of users using the website would give us a
fair amount of idea on the user traffic in particular geographic locations. This
could help eventually in applying proper load balancing techniques to ensure
user traffic is regulated there by reducing the load time.

Anonymous · Answer

In designing a new website, first analyze the content and target audience data
to determine objective and threshold values in terms of site access and minimal
and peak access times. Once you've determined the values you can design an
N-tier server architecture to reduce load on single section by having an SSL
decryption server, a app server, a database server, a content server and then
multiple load balancers set up to distribute access loads based on your
objective and threshold values. In a cloud based architecture, the use of
individual docker containers for app / DB / encryption, etc accomplishes the
same goal, with a kubernetes disturbed framework latency can be reduced and
speed increased.

If you were designing a new website, how would you plan and test for scale?

Answers