Hacker News Reader

I’ve been going down the certification road for a little over a year and should have 7 (now 4) by the end of the year. The company is paying for them and paying me above the local market average (I’m not talking about SV money here) - especially considering that I negotiated not to be made lead.

I started down this road as a dev lead for another company that wanted to “move to the cloud”, but at the time I didn’t know anything about AWS. Studying for the associate cert was mostly to get an overview of AWS so at least I could talk intelligently about from the infrastructure side and know what I didn’t know from the development side. I knew I would have to dig into the SDKs to actually do anything.

They hired a bunch of consultants who were old school net ops guys that didn’t know anything about AWS besides how to mirror on prem infrastructure. They were “lift and shift” guys. Of course it ended up costing more. I didn’t know then that AWS always costs more unless you are willing to change your processes, let a few of your infrastructure people go and let AWS do the “undifferentiated heavy lifting”.

I saw the situation was hopeless and all the company wanted to do was host a bunch of EC2 instances and the netops guys and consultants were so entrenched that there was no hope of me actually getting any real world experience with AWS.

I changed jobs and found a small software company that was already on AWS but the newly minted CTO wanted to actually take advantage of AWS. The MSP they had chosen before he came aboard were also a bunch of lift and shifters.

I was hired with no real AWS experience to lead the effort. After doing a lot of green field small proof of concept projects and fumbling through, I feel badly for anyone who tries to implement anything on AWS with just the certifications, without the industry background, and not working for a company with a business support agreement plan with AWS.

Good info.

You are right, AWS can cost more, but I disagree with "always". I run my personal website for $15/year, and most of that is the domain.

I would say: if you're trying to list and shift to AWS, you're doing it wrong, and it will always cost more. You also can't really fire many people aside from pure network guys. Server admins will still be needed for the virtual servers.

If you take the opportunity to redesign, you can realize cost savings. I still advocate avoiding layoffs, because you can retrain competent ops people to support the new infrastructure. The grizzled old guys that have trouble adapting are still usefull because chances are high you are not fully serverless within 3-5 years after a lift and shift.

As you become more familiar, you realize the tweaks that can be made to save money. We used to spend $600/month on 3-4 concurrent CICD workers. We swapped to docker-machine Spot-based CICD and spend $100-140/month for 12+ concurrent workers (we started at 6 and keep raising this number because the costs simply are so small)

Our front end is all static files (JS driven API calls). The $100/month server is now under $5 month for S3+CloudFront.

So I agree, AWS always costs more if you are trying to replicate a physical data center and ignore all the possibilities for improvement... But with slight tweaks (even starting with auto scaling.. or even more simply, timed lambdas to turn on/off servers overnight), you can reduce costs. Migrating to containers is also a huge cost savings, since you can pack more work into fewer servers without worrying about as much (I lived the nightmare of virtualenv's and managing 12 apps on 1 server.. I prefer containers now as an ops guy)

My goal and marching orders are to use AWS managed services for anything possible. We (the CxOs and myself) agreed that out of all of the business risks we had, our dependence on AWS is the least of them.

For builds we are completely serviceless. We use CodeBuild with mostly the prebuilt Docker images and one custom image to build some .Net Framework code. We use CodePipeline to deploy lambdas with CloudFormation and Code Deploy agents for EC2 instances.

For any new work, we have to justify a need for servers instead of using lambda for both APIs and back end ETL work loads - with step functions. There are a few back end workloads that we will need to transistion to Docker eventually but even then we are looking to use Fargate.

I have liked Corey Quinn's advice about saving money in the cloud. Here's a tweetstorm about the topic:

https://mobile.twitter.com/QuinnyPig/status/1091041507342086...

I could not tell you the names of many of the AWS services, but I have lambda functions running in AWS (written in Go). They monitor my DB loads in RDS. I get email and text alerts when things are not as expected via SNS, but I can't recall the name of the AWS service that I used to schedule that alert... CloudWatch maybe. For sure it was CloudSomething ;)

I'm not AWS certified, but I've been a technologist for many years and I majored in CS, so it all just sort of makes sense. Some names such as ELB, RDS and Lambda are intuitive. But some are overloaded, for example Alexa. They have Alexa for business and Alexa Top Sites (domain name research). BTW, has anyone else used Alexa Top Sites in AWS, it seems way overpriced to me.

I'd be leery of someone who knew all these names, but didn't have much actual experience as a technologist, but in an interview, they'd probably sound better than I would.

I have used Alexa Top Sites but have never paid for it. There is already a free csv you can download daily that contains the top million sites. For more granular lists, Builtwith often satisfies my needs.

Lots of questions about this. Some say it is old, stale data. I don't believe it is current. They charge 25 cents per 1,000 queries in AWS. Why give that away for free?

See thread here: https://gist.github.com/chilts/7229605

It is good enough for my needs even if its stale. I just want the top 100k sites and dont care about trends, and precise rank. I can see how someone else would want to pay for it though.

I maintain they bought that service just for the name (and toolbar installs).

Your response is exactly the right kind for an interview.

"I don't use those AWS services, but let me tell you what I have used". I agree, most people won't know one of: Sumarian or Sagemaker, but chances are you can easily know most features that matter for your job interests. Most new AWS services over the last few years are niche markets. A good company to work for would not expect you to know this stuff without experience. I personally don't expect someone without certifications to answer AWS questions, so it's always a plus when they can. But if they have certifications and can't answer any questions? That's a big red X. I can't speak for everyone, buy I usually have 2-3 different questions on different areas, and if you're certified and have used AWS, you should be able to get 2 of them.

Ex: what is the difference between EC2 server types M, R, and C.. my previous lambda question, and usually a question on ECS or CloudFront

Let me also state: If you have a certification and are having troubles landing a job, do a personal project to highlight your capability. It doesn't have to be hard.. or expensive! It barely needs to even look good -- if you're a back end guy, getting into an interview room and taking about the great CICD or script setup you did matters more than if it's a white page with black text... And if you're certified, you should know there's plenty of free it under $1/month options for AWS service. I hired one guy without a cert because he showed me the game server he set up in AWS to play with friends