4 ms·
I worked on both cloud computing and on premise project. Before cloud computing the risks were different: - much harder to scale. It was much more common to ov
by avrionov 5y ago
I worked on both cloud computing and on premise project. Before cloud computing the risks were different:
- much harder to scale. It was much more common to over provision and have machines and bandwidth being unused for years.
- when we were hit with very high traffic due to a bug or something else, most of the time it would lead to customer outages. Based on the contract some times it requires to pay back because SLAs were not reached. Also an outage could lead to customers canceling the subscription.
We swapped one type of problems with another.
- suction 5y agoIf you have a bug that renders your product unusable and refunds are in order, the flexibility of handling traffic peaks which a cloud provider offers won't solve that problem for you. It could even aggravate it. If a show-stopping bug is introduced, it would probably be preferable to fail quickly.
- philliphaydon 5y agoIf there was an outage in early 2000, we just went outside to play, or watch tv. Now if Facebook is down for 15 seconds everyone has heart failure like their life is over.
- ipaddr 5y agoAmazon uptime is 99.95 which calculates to 43 seconds of downtime everyday.
- lytefm 5y ago> It was much more common to over provision and have machines and bandwidth being unused for years. But the overprovisioned server might still be a lot cheaper than the cloud bill. It can be totally reasonable to have a server running at 1-5% load 98% of the time if you really need the capacity for the remaining 2%. Also, neither "scaling up" as in "re-deploying the same setup on a beefier instance" nor "scaling out" as in "let's expand to the US and have a server there" is too difficult if the setup is automated (Ansible).