New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Is using Amazon AWS an overkill?
abytecurious
Member
in General
A friend of mine is starting a new business which deals with merchants and payments. His computing needs are going to be around calculating percentages, maintaining ledger entries and invoicing merchants.
The company that is working with him has proposed to create three Amazon instances (Dev -> QA -> Production). I have very limited idea of how Amazon AWS works other than the fact that each aspect of their service is priced (instead of a bundled amount like when you get a Dedicated server).
My question is "Are the Dev & QA instances better off on a different provider (I think a reliable VPS should be fine)? Or are the cost benefits outweighed by the ease in which you can migrate between AWS instances?"
Comments
Don't put all your eggs in one basket.
In this case, AWS is a reliable basket but better to be safe than sorry. Personally I've always found AWS management and pricing to be confusing but if you're comfortable with it then it should "just work".
Generally, within corporates (but SMEs as well), you'll hear the saying that "No one ever got fired for choosing AWS".
This mindset is one of the reasons that AWS is chosen over cheaper alternatives. If AWS falls over, then it's interpreted as Amazon having a bad day. If another VPS provider falls over, then colleagues/clients start to question your choice of provider.
If the business's stack doesn't involve anything AWS-specific, then it would be fine to deploy the DEV/QA environments at another (cheaper?) provider.
However, if the stack does include something other than simple compute instances, then it would be easier to keep everything with AWS even if the auxiliary component is something that could be deployed separately.
The reason for this is simply to ensure that the environments are as similar to each other as possible.
No one ever wants to push something to production only to see if fail because of some AWS weirdness that they failed to account for.
@ramesh_vish,
I've been in situations like this before and trying to persuade the Developers to use AWS alternatives will only grow problems for you. The Developers are too accustomed to AWS and then there is always a "feel good factor" attached. Using a different provider, if there is some problem you are going to hear "this never happened with AWS".
If the company wants to pay... get them AWS. Get thru with your Development, deployment and testing phase. After a quarter or two when things move smooth, try to port your production to a cheaper host.
Keep a close eye on the Configuration settings and optimization tweaks and make notes of how you fix if something didn't work in the first place. This will help you immensely when porting to another provider.
That would be my plan if you ask...! Good luck
AWS is the new IBM.
Even the most over-engineered, '100% uptime' online system, will have some end-to-end blank-spots in connectivity, in the real-world.
Better to just assume 99.9% in your business plan.
Also, make your app cloud-agnostic.
(Always good to have leverage if you have to negotiate. )
AWS really really wants you to use their proprietary 'services' to lock you in.
There's no reason to use AWS for dev instances IMO.
Get a Hetzner box (or two or 10 ).
AWS Bandwidth pricing can murder your bizplan.
At my day job we are currently moving all of our infrastructure to AWS and reworking our code so it works better with the services AWS provides. The benefit of the stability, product offerings and Government accreditations far outweighs the costs.
From the sounds of what he wants to do he should use Lambda for the processing. It will end up being cheaper and you don't need to have different environments as you can do testing and staging under the same roof.
@trewq, agree with you, AWS can offer you so many options to choose from in order to process your daily tasks and to reduce your costs, you just need to know which services to use, a bit of learn will help a lot.
Thank you all for the comments. I think the way to proceed would be to see how the AWS thing works out for a quarter/two (factoring actual utilization & costs), plus time to understand AWS specific nuances. Do a cost comparison with a different provider and finalize the long term plan.
@trewq: Lambda seems to be an interesting feature. I will read more into it. It may be a good idea to use it for "batch" activities like invoice generation which doesn't have to run all the time (again looking at costs, etc.).
For me though, I'd rather do a HA system and failover between nodes. That is just me though, since pretty much you can't go wrong if you have set it up properly - unless all the DCs go lights out at the exact same time.
They power Netflix. It's a pretty darn good compliment on what they've built. It's never been more stable for me. I can't recall the last time I went to turn on Netflix and it wasn't available.
Going to be interesting to see if AWS maintain their lead in the 'cloud' space, or whether Microsoft can start to eat into it as more and more of the old guard start to migrate.
https://twitter.com/Netflixhelps
The last it happened was few days ago, not sure whose fault though.
I laughed. I mean they're doing well it's just that I can't shake the idea that Microsoft is behind it. I can't help but think they'll discontinue it faster than a Google product that only succeeded in South America.
If you have the budget, go for aws. Else get DO and Buyvm slice, set up HA. And in the extreme scenario, get a Buyvm ddos filtered IP. Bulletproof.
I know what you mean, and holy shit they drive me up the fucking wall with the way the roll out updates but trust me, they are really serious about Azure. The impression I get is that it's very much seen as the future of MS.
Still have to set up HA on AWS too. Big misconception that their instances are more HA than others. Perhaps an EBS backed EC2 instance is a bit more but that's about it.
True, but for AWS, it's easier since you can just choose different availability zone and pray that Amazon actually sets them up using seperate connection and power.
Pray? It's not guaranteed?
Yes, it's guaranteed. Different availability zones are in different data centres.
I wa going to say, anything else would be insane.
Set a budget first, and don't "cheap out" on your only presence (assuming your friend is online only). In saying that, allow for overages as your infrastructure is never set in stone.
AWS is great, but if you're new there's a lot of pitfalls to watch out for (that could cost you $$$). My advice, just make sure your sysadmin is familiar with AWS.
It is interesting that you mention this. My friend got an email from AWS that he could opt-in to use longer resource ids (https://aws.amazon.com/about-aws/whats-new/2016/01/longer-format-for-ec2-resource-ids/) and the sysadmin told him that it was a phishing email to get my friend's login details.
What an idiot. If he wants to use AWS he should be able to keep up with the current happenings. If he hasn't been getting emails about the longer ids then he must not have an AWS account.
If you want to talk more about an appropriate AWS setup please PM me. From what you're saying this is going to end in tears once the bill comes in.
You pay more and you get (generally) higher availability
A thing I was working on a while back was costing £300/m for a Windows box on Amazon, hosting with OVH for £35/m was just as good in my case.
I know of a few companies that grew within the garden of AWS and it becomes their biggest outgoing cost. There may be a point in time where it makes sense to move away from them.
On that none, if what you are running is stateless, such as data processing, you can use spot instances and cut the costs by up to 90%.
Sometimes AWS isn't the right choice, especially if you don't have the time to invest in recoding bits of your infrastructure. It can still be hosted on EC2 instances but then what's the point in using such a feature rich platform if you are just going to ignore it all.
Indeed, you're almost paying a price for convenience and balancing that against development costs of not being on AWS.
IIRC Netflix still use Amazon and are worth a few quid. Seems like it'd be too much hassle to move.
In a lot of cases, the big boys will be using the likes of AWS and Azure in large part to one thing - liability. They can afford to stump up serious compensation to cover financial loss as a result of a service outage.
From what I can tell the only alternative to Amazon would be building out their own data centers in every PoP Amazon has. Netflix also has deals with ISPs and puts a cache inside their network, works out well for everyone. Pretty sure google do that too.
Love that kind of scale. Nice problems to solve...
Totally agree about liability.
Netflix actually opted to move to AWS by choice, while they were the household name they are today. This year, actually. You have to think they really crunched the numbers too and realized that was where they wanted to grow.
yeah, I read a bit about that. They've been migrating over for 8 years so must have twigged on back then.
As trewq mentioned... they probably came to the conclusion they'd have to reinvent the wheel.