I've worked with AWS at my job for around two years. I don't proclaim myself an expert or something alike, but surely I've learned a lot during this time.
A couple of months ago I decided to commit and study to get an AWS certification- AWS Solutions Architect. That's why I'm creating a series of posts such as tutorials, guides and tips/tricks on AWS.
There are tons of things I need to learn and deeply understand, and I think teaching them is the most effective way to learn them.
Would love any feedback, corrections and/or suggestions on the series' posts! Also, I'm going to re:Invent 2018 by myself, if anyone wants to hang out there feel free to talk to me!
Thanks a lot, fam! ❤️
I will be updating this post whenever I create a new post for the series.
- IAM
- S3
Top comments (11)
Tip #1 on AWS: If you don't absolutely depend on some of their proprietary systems (e.g. Kinesis), then really consider twice before using AWS.
You will find Google Cloud Platform and Microsoft Azure excellent options, in terms of absolute price comparison, being more user friendly, and having more ready-to-use products rather than raw building blocks.
Amazon clearly is in a dominant position as they have little to no interest on improving their systems, and instead basically tell people to learn to live with the fact that their systems are awful if they want to use them.
I'm not an amazon fanboi, but I believe we should be fair. This is exactly what IBM did in the 70s, m$ in the 90s and apple is currently doing.
In fact, amazon is handling this pretty well, imo; free tier, 10s of new GA services every year? Google isn't even trying to compete - they are maintaining the services they created in past years, not putting much new stuff out. Azure is a mess, they don't need my comments to see they need to fix it. Trying to be everything to everyone, hoping people will play along, like in the good old days.
Amazon is in this clearly dominant position because only they cared enough to take the risk. Now the rest are playing catch-up. It is shameful that the way to "compete" with them is to try and pull them down.
Amazon's usability is incredibly poor, the web console is barely functional and gives incomprehensible and unhelpful error messages regularly, in addition to just bugging out. It has tons of special cases only documented on some impossible to find product FAQ pages in the style of "this type of a load balancer can only connect with the instance targeting to instances of types ...".
It has incredibly useless features for networking and other common features that literally everyone else does better - boot up an instance in EC2, give it a name "web", then boot up another instance in the same VPC with the name "db", how do you connect between them? You've got two very idiotic ways - direct IP, or with the long "hostname" which also includes the IP and is thus objectively just worse. Alternatively you spend your own precious time in implementing a way to make the machines discoverable to each other using a combination of VPC specific private zone, and setting up startup scripts to register your instances there, giving them decent hostnames, setting up resolv.conf, etc., etc.
Their startup scripts, alert limits, etc. are instance specific and done clearly in wrong ways. Startup script only runs once, and is managed using an incredibly clumsy format. Alert limits often are set in absolute values, which causes issues if you'd just want an alert when e.g. your CPU credits on t* instances drop below 20%.
The only area where others are playing "catch-up", is market share, they are far ahead of AWS in every other way that matters to most people. Their services simply work, and are cheaper.
Azure provides excellent tools that are integrated with each other for easy use for many common purposes, e.g. logging, various error tracing and analytics things, etc. whereas Google provides incredible pricing and simplicity while providing the necessary tools to build easily scalable systems on top.
Free tier is not really a thing that is good about AWS, Google and Azure do better on that regard too, giving a lot more free credits and discounts on e.g. test & dev environments.
The amount of new services doesn't really matter if you never need them, most of them are really bare bones and useless as is, and in more cases you will spend more time making them just barely work for your needs than you would by just using some other system, e.g. that garbage tier monitoring & alerting setup they offer.
So their virtually unchallenged dominion over the market share is not an indication of "success"? How did they corner the market if people care about other things? Are you telling me people chose them because they had to, like IBM in the 70s and m$ in the 90s?
I don't want to start a mud-slinging game, and try to defend an emotionless, ruthless company like amazon, but m$ has done even worse. They are both companies, and have done (and will do) cold, calculated things to make more money. We need to be fair, though, I believe.
Bottom line (for me);
azure console is something you'd only spend time on if you're forced to (like ng6)
as i said, gcloud is not even trying, they have state-of-the-art services like bigquery anyway, they are looking to perfect them - when aws learns things the hard way, they'll likely pull a m$ on them and copy the services they need - why pay for r&d?
aws free tier is something that's automatically given - compare that with how you can get the same perks from azure. I'll request a credit/discount, it will be reviewed by someone (or something) and I'll be granted one if seen fit. With aws, it's just given to me - I don't have to prove myself to someone for it.
p.s. I don't plan on any more answers, I don't want to waste more time on this. Each to his own.
I have no interest in your random corporate ethics discussion. I'm talking about service quality, what is the best platform for the typical person to build their typical application on.
Obviously you have no idea how a capitalistic world functions, "success" does not mean "better", it simply implies their marketing etc. has done a better job and there's a lot of idiots in the world making decisions. Nobody ever got fired for choosing AWS as their platform, sort of like Java. That doesn't make AWS or Java the best choice, just the safe choice if you're unable to think for yourself and afraid for your job.
Also you guys shouldn't listen to the FUD about the AWS free tier here. Just check the facts:
cloud.google.com/free/
cloud.google.com/developers/startups/
azure.microsoft.com/en-us/offers/m...
azure.microsoft.com/en-us/free/fre...
azure.microsoft.com/en-us/pricing/...
And actually, even if you do depend on some of their proprietary systems I'd still prefer to host most of my stack elsewhere.
Thank a lot, Iwas looking for a lot stuff about the certification path, now I can follow the serie to complement my study.
This is great. You've got yourself a new student 🦄
Yeeeii!! :D Please, do let me know if something is not clear. Thanks!
Thanx for your effort. How long would it take for this series (approx.) to cover the topics necessary for the certification?
I would say an approximate of three months. I want to go deep into each service before getting into the architectural design questions