You do realise that people have being doing this for ages and amazon is just
matching what people have being doing on their private networks and offering
it as an option?
You do realize I have no interest in re-inventing AND maintaining
AND documenting that wheel? Maybe it’s different if programming
is just a hobby for you, but no thanks.
So outside of some very specific use cases AWS is a horrible idea
I’d reverse that, if you’re actually running a business.
unless that business can tolerate unknown amount of downtime or data loss and consumes such a small amount of resources that AWS pricing is not affecting the bottom line sure 
unless that business can tolerate unknown amount of downtime or data loss
Not our experience. 3+ years, zero downtime, zero data loss - maybe
you’re doing something wrong?
and … AWS pricing is not affecting the bottom line sure
Again: AWS saves us money and time. And time is money
My viewpoint is that once a company really starts using/abusing the whole Amazon catalog of products it becomes an increasingly big task to replace all that. Given the sheer amount of services and their interplay, it’s extremely foolish to try to duplicate all of that.
Theoretically, it’s possible, but it’s not close to feasible for big deployments. And yes, theoretically, you can have downtime on AWS, but given enough redundancy across regions it’s certainly not likely, depending on your planning.
Is AWS (or other cloud services like it) for everything? No, most services don’t need it. If you’re building a system that doesn’t (or shouldn’t) need any supervision (and thus doesn’t need people on call) and shouldn’t need an ops team for it, they’re good places to start and I think they’re pretty much the only endgame for that scenario.
That’s a very strong point of evidence and good way to evaluate risks. I have 0 projects where I make the decisions that are dependent on single vendor. There are projects external to us that are AWS only they have being affected by 11 hour outage in 2011 (it took AWS 5 days to fully restore all services) severe service degradation due to Netflix mass node migration I think either same year or 2012?. The recent s3 outage had such services down or severely degraded Heroku, Dockerhub, CircleCI, TravisCI, Quora and on and on.
Please go ahead. You can query Hex’s HTTP API to find all package versions that have docs and you can download doc tarballs from https://repo.hex.pm/docs/PACKAGE-VERSION.tar.gz.
I have set up region replication to Europe for the S3 bucket and added automatic failover to Fastly (our CDN in front of S3) so this shouldn’t happen again unless both AWS regions go down.
They have single tenant consuming 30% of resources they do not have spare capacity to absorb a significant issue with that tenant. It has happened before it definitely will happen again.
For AWS VPC we’ve actually recently set up an Illumos cloud that does this, we can trivially spin things up and down on demand.
The Security groups, based on what I am reading, seems to already be controlled by the Illumos software as well?
And apparently AWS has a variety of ACL’s, since the rest is about networking I’m guessing this is Network ACLs? If so then again, Illumos is already handling.
Identity management here is currently controlled via an LDAP server, however later this year it is being replaced by something that handles other security handlers, which will still include LDAP, but will also add quite a few others. Right now everything is using LDAP though, old but reliable.
© Copyright Elixir Forum | Terms | Privacy & Cookies
