Amazon cloud or renting hardware in a data center. What are the pros and cons?

Question

Amazon Web Services

It is necessary to find a platform for hosting the "payment system". The system consists of an application server (Java and Apache tomcat) and a database server (Oracle 11g). Since the system is related to finance, high reliability and availability are needed. Share your experience where it is better to mark up a system in the cloud or rent servers in a data center and hire admins to build a cluster. Please state all "FOR" and "AGAINST". Thank you!

Answer 1

Own hardware
possibility of certification of the FSTEC and FSB systems +
safer from the point of view of the client +
possibility of a "cunning" configuration +
many parameters depending on the directness of the hands of the administrator + -
Amazon
cost +
ease of scaling +
"helplessness" in case of hoster problems - In this case, I would choose own iron.

Answer 2

One admin will do quite well at the initial stage. Several admins are needed, rather, to ensure "round the clock", and not because of the abundance of work.
With channels, there are two options, with and without an autonomous system. We did failover with an autonomous system, it’s more convenient, there are no problems with the selection of IPs by the provider and there is a fairly large margin for maneuver. Two duplicating nodes are installed, as soon as the master falls off, the slave rises. Routing is in order, everyone is happy.
If without an autonomous system, it is easier to choose a data center with a large number of operators and connect to several. There is usually order with electricity, the Internet will not drop for everyone at once, the inputs are independent. We chose one of the spb-ix sites for the main node.

Answer 3

The main problem of clouds is disks.
Imagine you have an “interface” with bandwidth S.
On your hardware, you are alone on it, on the cloud there are many. But what if everyone starts generating backups at 0.00 at once?
Well, about the network, everything has already been painted above.

Answer 4

Payment systems are certifications of local and not only local governing bodies, so they won’t let you post this on some “incomprehensible” Amazon - I have a slightly different work profile, but the equipment and server room for the main service provided by the company was very severely checked a short time ago checking state structure.
In the CIS, the requirements are approximately the same for such things.

Answer 5

many parameters depending on the directness of the hands of the administrator +-

Will one admin pull it or do you need a staff of 2-3 people?

“helplessness” with hoster problems - In this case, I would choose my own hardware.

Problems with communication channels can also be on your own hardware. You need to have multiple providers. What do you think, if one of the nodes of the cluster is moved to another DC, in another region, will this help ensure the constant availability and reliability of the system? I wonder how to do it on your hardware and in the cloud?