info_tfgrid/collections/problems/internet_infra/cloud_issues.md
2024-01-26 06:06:57 +03:00

2.5 KiB

high CDN cost = Content Delivery

  • CDN's cost between 20 and 60 on major CDN's from Amazon, Google, ..., after negotiation and willingness to park lots of money and commit per month it mught be certain discount.
  • Other specialized CDN's can be around 10 per TB if services from e.g. Europe or US, smaller CDN's are much more expensive.
  • Managing these CDN's is expensive and not easy job.

this means to service 10m people it would cost about 1 to 2m USD per month just for CDN alone.

High cost of operations

  • cloud servers
  • managing those servers, what if mistakes are made

Ability to be shut down on the public cloud

  • its highly probable that within months there would be attempts to shut down the CDN's or often it happens more softly (give bad service and make your product look bad).

apps typically not ready to scale

  • ...
  • there are optimizations which can be done e.g. better indexing, better queries, more optimization strategies e.g. using redis
  • more knowledge is required to optimize the app

Ability to shutdown the mobile apps

  • google/apple might chose to shutdown apps
  • not easy to know how fast this will go, it might just be overnight

GDPR

  • are very annoying constraints in e.g. Europe, need to be careful to comply
  • its important to have good terms & conditions and think which legal entity will be the counterpart of the T&C
  • there will be legal requirements like how to shutdown / recognize bad content

Redundancy / Uptime

  • is the site redundant, can data be lost?
  • what happens if a datacenter goes down?
  • or what happens if a DB server crashes?
  • what happens if e.g. DB gets corrupted?
  • how to make sure people always deserve the service they need
  • if it kind of works now, will it work if 10x more people?
  • is everything monitored?
  • if an issue is detected are there people available 24h/day 7/7 to fix
  • do the people who will fix have the right knowledge, where is that knowledge stored
  • is the monitoring system itself monitored, very often monitoring by itself will stop working

Performance

  • how to see performance is not good enough for customers
  • how to make sure we can easily fix it, can be region specific
  • how to relocate services?

protect against human error

  • mistakes are and will be made this might have huge impact on uptime and if not careful loose data
  • truck factor: what happens if someone goes away? can org easily take over and continue
  • level of automation & documentation?
  • how is version control done