Personally Amazon AWS but also Softlayer, the multi-nationals use these and are running a 24x7 for Tradestation & eSignal. Was setup by some specialists and pings sms (60m+) or iphone twitter messages (<5m) when signals are hit.
It looks like it's the foothold in to automated trading for small retail given their pricing structure, low latency is all very nice but you're never going to compete with HFTs. Basically you get some support, we just receive a timed ping from the server instead, currently every 6hrs workday but could be 1m or 10s if wanted.
Almost any Windows box plus RDC will work, it comes down to what infrastructure you need behind it, both of your choices will give you a level of certainty, but can never match enterprise. What guarantee do you need 80%, 95%, 99%, 99.9%, 99.99%, the higher up you go the more it costs if things go wrong but the faster you're back online or in the higher options, not offline as it was pre-emptive, it's that simple.
When volatility spiked we just chose a faster instance, took <1m, and can downgrade or switch it off over the weekend via automated restarts. There will be nothing wrong with the ones above, but they are the start, as your knowledge and requirements increase so will your need to move quicker.
We just went with the end result first rather than relearn each time moving to a new instance or environment. If you're not confident about your strategy then probably best to go with the cheapest that provides the most options medium term while you go through the process.
