What are the pros and cons using this virtual machine compared to buying your own?
A VM in the cloud gives you a quicker, lower-cost startup. You can get running immediately with no machine purchase costs.
Also there are fewer system maintenance issues. If you're short on IT skills or assistance then that's something to consider; esp for 24/7 coverage.
Instant scalability is another advantage.
But obviously the initial costs of your own machine get spread out over time, and there are fewer spikes requiring instant scalability, so the larger your setup, and the longer you intend to run it, the more it's probably worth doing your own installation.
From Experience the cloud option outweighs your own dedicated server.
Why Scaleable - not days or weeks but done in minutes. It can also be downscaled in the future. (go from 10gb to 1tb - 4 core to 16 cores)
As CDBrown mentions - Uptime and very rarely downtime. 99.98% This alone outweighs your own server. (when is goes down 4.59pm on Friday afternoon)
*RAM buy as much as you can afford. 16GB of RAM is going to help. 32GB better.
One big pro of having your own physical machine (... that you can physically access!) is sneakernet data loads.
This is less of an issue if you don't host your own data - but if you have gigabytes or terabytes of data and are constantly updating it, you're going to have to invest time and money [bandwidth] transmitting it "into the cloud". However, if you can walk up to a physical device with another physical device, then a week-long upload could become an hour-long file copy.
As I see it here are some things to consider.
For Amazon Cloud - offsite virtual environment - security and scalable up time - 0 downtime - admin available where ever you are
Against Amazon Cloud - need to place expensive software on someone else's machine - need to place possibly sensitive data on someone else's machine - need to spend a lot of time organising how the services will be updated - periodic payment so if one payment missed services can be down - lastly if everyone has data in the same cloud architecture how long will it be secure
There are no hard and fast rules on this my suggestion would be to look at the Cloud as a start up location and as you get bigger migrate to a local static virtualised server enviornment you can control.
FYI , a white paper of ESRI about Estimating the Cost of a GIS in the Amazon Cloud
Regarding GIS and cloud computing in its current state I think the biggest deciding factors would be how many users you have and where they are located. If the end users who will be accessing the services and data you place in the cloud are from various independent locations with no common network infrastructure then cloud computing might be for you. While you won't need to manage the hardware you will still need to manage the server OS and software just as you would with a local box. On the other hand if all of your users are locates on or in a campus/building with a common network infrastructure then a local install will probably offer much better performance.
We are currently supporting about 30 editors across the country on an m1.large machine in the Northern California region. It costs about $350/month. Our plan is to move to a m3.medium in the Oregon region. This should bring our costs down to less than $200/month. We run standard workgroup version of ArcGIS Server with SqlServer Express, SDE, GIS Services, webserver and FTP site all running on the same box. We have been very happy with the performance and cost.
We currently have in-house physical production hardware, which at times is great (speed, horsepower). At other times it's a nightmare. Like when updates get pushed to your box by someone else and it totally jacks things up. If you can have total control over your environment, you are better off. Unfortunately, that is rarely the case.
We are considering some ESRI/EC2 cloud solutions. Can anyone chime in on costs you've encountered?