Easy Amazon EC2 Instance Comparison | Hacker News Reader

Easy Amazon EC2 Instance Comparison

ec2instances.info

285 pointsobi1kenobi35 comments

Use this all the time -- I kinda wish it could remember what columns I select or turn off. I don't often care about windows pricing, and tend to think about monthly cost vs hourly.

Up vote my issue.

https://github.com/powdahound/ec2instances.info/issues/201

that's why I love Google's custom machine types, fully customizable. https://cloud.google.com/compute/docs/instances/creating-ins... and all with fast networks.

Yep, as much as I love to hate on google, I recently trialed their vms, aws and azure and google offered the best experience by far.

I have no idea why people like aws so much. To spin up a simple vm, I had to go to the correct web interface for the specific location, go through a wizard full of weird nicknames for all services while browsing multiple pages of documentation to figure out that I need "ec2" and "t1.micro" or "m2.macro", how much they cost, what the hell an "ebs" is, then download the key they generated (which was only valid for that one location as it turned out later...)

At google, I just selected "vm", specified the location, cpu, ram, storage, entered my own key into a text field, it immediately quoted a price and I was done. 2 clicks and 5 input fields, no documentation needed, all made sense.

Not to take from noob experience but google just has less features and services. When/if they have as many it will be similar mess. It's already getting there. Also, you could import your own key in aws. Though I agree that aws could benefit from global settings across regions like keys, cloutrails, config and so on. With more regions coming up it becomes unnecessary noise to manage each region separately.

What features is google cloud missing that's stopping you from switching today.

A lot of Googlers lurk these forms and I'm sure they'd love the feedback.

It does not answer your question straight but here are some things that stared at me in the first week though perhaps there is a way I don't know about:

1. No tag based billing ala aws. The only way to separate bills seems to be using projects. Which means separate networks. Which means more vpn needed to tie it all together to corp network?

2. Need more granular IAM scopes. I.e. I want a user that gets readonly for compute, disks and snapshots and create/delete for snapshots.

3. Tags are labels are tags? For me that was a bit of wtf moment. It might have its reasons but I want to add labels without possibility to trigger firewall rules. For that matter why labels are alpha feature? It's not like it's a new service.

ALL. 1. billing - is stoped accepting my card a year ago (when card expired) and didn't work so far. I wrote to support, called - nothing. 2. Highly unintuitive interface, I spent 2 days to run a single instance, WTF ???. 3. Useless support 4. Limited API 5. Limited access/roles system 6. Almost no support of external deploy/release tools 7. No logs/events collecting/processing service

Google is not suitable for development at all.

How much does the experience of your first 15 minutes really matter when you're picking a platform for the next several years?

A lot. We're running a small <10 team on a tight budget, we don't have time [or money] to waste.

Looking towards the next several years:

* https://cloud.google.com/container-engine [docker in the cloud]

* https://cloud.google.com/bigquery [sql for analytics]

* https://cloud.google.com/ml [tensor flow]

* https://cloud.google.com/dataflow [2nd gen map-reduce]

* https://cloud.google.com/bigtable [large scale semi-structured storage]

* https://cloud.google.com/preemptible-vms [cheap VMs]

* https://cloud.google.com/compute/docs/networks-and-firewalls [fast networks]

Quite a lot of Google's infrastructure crown jewels have become part of their Cloud offering. Not sure where to start looking for competing AWS services, forget about looking closer at design / docs / scale / reliability.

[xGoogler here, used to work on GCP]

I would say if you're going to be locked in to one vendor for 15 years, at the start before you're locked in is the /only/ time the experience matters :)

I would say quite a bit. If it takes 15min to get up and running, any major change at Google would still only require a new 15min to get to know the new system.

If you need a month to properly get to know AWS, any major change will most likely cause another month of getting to know the new system.

Or it could go the other way. No certainty there

GCP just feels so much better "put together" from UX to product features. It had the advantage of not being first-mover.

This is pretty amazing. Navigating the EC2 pricing documentation is such a hassle.

I tend to think of costs in terms of monthly, as well, and this is valuable when working out cost-proposal estimates for projects I'm pitching to potential clients.

It is missing is storage and bandwidth costs, which can be a nontrivial chunk of an EC2 bill.

And yes, it is hair pullingly frustrating experience trying to figure out how much an EC2 instance is going to cost you using Amazon's tools.

Can you elaborate a bit? Is it the cost calculator? (https://calculator.s3.amazonaws.com/index.html)

What do you find frustrating? What can we improve? Let us know!

(Disclaimer: I work for AWS.)

Not OP. But Amazon's pricing is very complex compared to Google's and also more expensive when it comes to VMs.

Huh. Apparently, running Postgres on RDS is tad bit more expensive than MySQL. Why is that?

The RDS FAQ reveals this:

Q: Why is the pricing for each RDS database engine different?

The pricing for each database engine of RDS varies because our costs are different for each. These costs include many operational components in addition to software licensing. We will continue to work hard to reduce costs and pass on those savings to our customers.

My guess is that Postgresql required them to write much more internal tooling to "get right", and the costs reflect that.

Always see Amazon price reductions being promoted on HN, I didn't relise it's still so expensive, compared to my hetzner dedi.

Yes, it's not apples to apples at all, but AWS runs about 5x the cost without considering bandwidth. If you use a lot of transfer, even between AZs (same price as between regions!), it will dwarf the hardware costs. Compare e.g. the Skylake EX51-SSD at $63 USD / month amortized for one year vs R3 High-Memory Double Extra Large at $300/month 1-year no-upfront. Both have 8 hyperthreads, AWS has ECC memory (good) and hypervisor overhead (bad), Hetzner has a newer processor and higher clock speeds.

Depending on your needs, dedicated hardware can run a lot cheaper. You can over-provision by 4x and you're still saving money. There are a lot of other benefits to AWS though, it's a huge ecosystem, and it's much more flexible.

I'm all for dedicated servers but please don't compare against non-ecc ram.

I've been using this site for the past few weeks whilst migrating, is an invaluable resource given the ec2 pricing isn't particularly easy to work out in the docs.

So funny, I just left that page to come here. Thank you, unsung hero/creator!

Fun fact-- it was started by one the founders of HipChat.

FYI https://news.ycombinator.com/user?id=powdahound is the same person on GitHub and HN.

Some Postgres RDS reserved instances (1yr no upfront) are more expensive than on demand. Is this correct? m4.large, m3.large

Would be awesome if it stored your columns, cost, and preference settings in local storage/cookie.

I think it does, I just opened the page up in a new tab and my previous search was still there.

It stores some data in the url, but not all.

I opened up a PR (https://github.com/powdahound/ec2instances.info/pull/202) to store the rest.

Why doesn't amazon make it this easy to compare? Right now looking at the comments a lot of people in charge of server decisions are referencing a random internet person instead of their docs.

It's worth noting that the performance of a given EC2 instance type can vary depending on what hardware you end up with, and that this is not consistent with the 'instance type'. One 'instance type' can be associated with multiple underlying CPUs that have different performance:

http://blog.robertelder.org/a-weird-old-tip-about-ec2-instan...

It can vary, though it should be predictable. M3s, for instance, can launch as either Ivy Bridge or Sandy Bridge, depending. Same with the older M1s. But unless listed, you should get the processor expected.

https://aws.amazon.com/ec2/instance-types/

Been using this site for years.

Really nice, is it a shiny app? They are ridiculously easy to set up.

Reply on news.ycombinator.com