Latency to AWS has always been worse for me than to the big name DCs like softlayer. You can not autoscale your server on the fly (I don't think?), you have to redeploy your server image onto a bigger instance. IO on EC2 has always been MUCH worse than a local RAID10 array for me.
I use EC2 extensively for an auto scaling application, spawns more copies of itself when load is high, it's great for that (keeping costs low for variable cpu needs) but I wouldn't use it as my main host.
Dallas Colocation by Incero
e: sales(at)incero(dot)com • 855.217.COLO (2656)
Colocation & Enterprise Servers, SATA/SAS/SSD, secure IPMI/KVM remote control, 100% U.S.A. Based Staff
SSAE 16, SAS70, Redundant Power & Network, Fully Diverse Fiber
I wouldn't go with Amazon. They are pretty good, but I think you could do better. If I were you I would look at stormondemand.com and not Amazon. It's in a Liquid Web data center (and it's a Liquid Web company), so you can expect a quality network.
I believe a very important point is being overlooked.
Amazon operates multiple regions, allowing users to add redundancy to their applications by hosting them in several regions. In a multi-region setup, when one region experiences performance problems, customers can shift workloads to an unaffected region.
The latest outage affected sites generally that were only hosted in "one" region and not taking advantage of the capabilities built into the system.
Redundancy is a factor that influences the "up-time" of any hosted site. Amazon cloud or Softlayer DC, be sure that you have the capability to shift to another installation when there is a problem with one machine / vendor.
It seems that the Amazon "outage" points out the consumer's lack of basic network understanding rather than stability of a given system.
There was a thread a while back about Amazon's total lack of response for a DDOS attack from one of their servers that was taking down a host's server. Multiple tickets and phone calls didn't even get a response, and I remember the host's frustration over the event. There evidently was no tech support over a long weekend, or something.
So what, its not their customer they were ignoring, right? I find it inconceivable that a provider wouldn't care about the rest of the network. Their lack of response to another business they were impacting shows an utter disregard for anyone else, in my view. I suspect that bleeds through to their relations with their customers. If you're happy with SoftLayer, I would stay with them.
That article seems to be centering on how Amazon "broke their promise" but one small part of it hits the main issue:
Of course it's possible to protect against a catastrophic failure (multiple AZs), but for most businesses the additional expense and engineering effort isn't worth it (or may even be counterproductive by introducing additional complexity).
Vendor guarantees should be relied on for reimbursement only, reliability of authored systems remain the responsibility of the author.