You don't need to be an 'investor' to invest in Singletrack: 6 days left: 95% of target - Find out more
Anyone use these public platforms for making your own apps out of pick'n'mix services? IBM Bluemix, Heroku etc?
Or PaaS in general?
Just starting out with Mendix in our dev team, to try and speed development of 'quick wins' - have used elements of AWS in the past to 'fake it' a bit (DBaaS with EC2 custom installs talking to them) but often there are more cost effective ways to run something long term. If you're building and burning continually, and don't have infra or budget to run a dev ops environent I can certainly see the appeal - really good for startups who just have good developers and an idea. Think Heroku offers a free time to try it?
I work with Kubernetes and AWS for a living and we have some quite big customers using it. Feel free to ask questions. 🙂
I use Platform.sh quite a lot. Good for Drupal, WordPress, Symfony type stuff.
I love the workflow
Rachel
I know at the Drupal Association we use AWS to host our testbots. Means we can allow many more concurrent bots during heavy load times, like global sprints and conferences. The Drupal core is approaching 35000 tests now...
Rachel
When you say AWS you are talking about IaaS mostly - renting VMs and so on, right?
@molgrips Yes, with a PaaS ontop. Google offer GKE which is Kubernetes as a service i.e you ask for a 4 node cluster and it builds it out for you. You just bring your workload. Given Google wrote most of Kubernetes it's pretty robust.
Using a few
Airtable for data, Glitch for node apps and absolutely love [url= https://zeit.co/ ]zeit[/url] for deploying apps. Their developer experience is second to none.
Google offer GKE which is Kubernetes as a service
Still looks like IaaS to me. You then install your own app server/whatever on it, right?
PaaS to me is something like OpenWhisk or Cloudant - it's already there you just deploy your code.
we use PaaS but I can't tell you any more or I'd have to kill you
"Apache OpenWhisk is a serverless, open source cloud platform"
Serverless ! = Platform as a Service.
Clearly - but there are PaaS serverless offerings, along with other types of PaaS.
What're you wanting know about them? I've used Heroku in a production environment, did things things that we needed.
I'm not sure where in the scale I'd put kubernetes as it handles scaling for you and you just tell it what docker image(s?) to spin up so you just push your image to the right place.
I'm quite interested in things like terraform and amazon cloudformation, not had a chance to play with them yet but they look very cool.
We use Bluemix, API stuff mostly although we've done some POC chatbot stuff too using Node.
What're you wanting know about them?
Just curious to see who uses them and what people think. How much of your IT operations are on these platforms - just a few side projects or main business critical stuff?
Have had a quick muck around with Heroku. It wasn't obvious to me that it was actually any easier to set up than a VPS for light use stuff. Possibly for critical service it MIGHT be more robust or take the load balancing issues off your desk, but no easy way for me to know/test if it will. Never pursued it commercially as not clear to anyone exactly how much it would cost (this was 18 months ago - no idea if it is still the same).
Finally worth bearing in mind that if Heroku or any other PAAS/IAAS is outside the EEA, or failover reverts to a a system, or there is a backup outside EEA [as I recall Heroku was EC2 based with no obvious / easy option to define the local infrastructure) then you may need to get your users informed consent for personal data transfer outside the EEA (depending on many factors).
Wow. Getting into some actual development with IBM Cloud and having a great time. So much clicking and things just working. Toolchains are a particular highlight. Instant devops setup.
We use Heroku in production for business critical stuff. It's very easy to use and reliability is not far off the underlying AWS. They've had a couple of embarrassing oppsies. We hide behind a CDN to help performance and reliability, degrade gracefully and have a static site hosted on s3 just in case.
But it is also very, very expensive - our Heroku bill is just under a million dollars per year - and like all cloud stuff it's easy to create things and not destroy them afterwards. Whether it's worthwhile really depends on how much money you could be making in the time you've saved.
I personally love how easy RDS, SQS and DynamoDB are to use. It would be very hard to see past them if they do what you want unless you're on a really tight budget.
Have used commercial IaaS for enterprise level stuff, big Oracle apps, warehouses etc. Not a pleasant experience. End up with 3 or more tiers of IT admin to wade through and performance terrible.
More recently smaller projects using AWS or Azure both in house and via contractors. Smaller IaaS and PaaS. All good. Also now looking at SaaS for data processing and BI.
IME best not try to reinvent the wheel.
How much of your IT operations are on these platforms – just a few side projects or main business critical stuff?
There has been a big shift the last 2 years, the big cloud providers with more transparent pricing are really highlighting the poor decision making and rip-off contracts for hosted stuff. Contracts are not getting renewed and the playing field has been leveled. Is good for the small guys who now have a reliable platform to back them up.
<span style="color: #444444; font-size: 12.8px; background-color: #eeeeee;">But it is also very, very expensive</span>
Cheaper or more expensive than doing it yourself?
Depends what "it" is. Building your own competitor to Heroku would be hard, and you end up chasing features getting further and further behind. Too many internal "private cloud" projects die this way.
A neat deployment system that provisions hosts, builds and deploys multi-region with fail-over for you? Jenkins on steroids? Maybe. We tried and abandoned in favour of Heroku. Heroku was just so much easier to work with. Actually we've found even raw CloudFormation is better (sometimes, CloudFormation can be a real pain and then manual provisioning is easier). I have a little shell script that will create 4 ec2 boxes, copy my app over, configure an ELB and off it runs about a minute later. Of course you don't get auto-scale out of the box and you can't just turn the Dyno knob to meet demand.
For us we're in a growth area. So our developers build something, it gets an extra X revenue and they move on to the next thing which will generate YÂ revenue. Where X and Y are greater than the Heroku cost. If you're in a more stable environment where there's less new revenue associated with developments then it would make sense, in my view, to use a cheaper, slower to deploy system. I've been meaning to crunch the numbers and find the point it makes sense to switch for years and never done it. Same argument applies to AWS Lambda.
Containers may change the game a bit. We also run a large cross-region Kubernates cluster. That, in itself, isn't as easy as the hype would have you believe but progress in the toolchain is very rapid and it's becoming much more accessible. EKS looks exciting.