I don’t think Jason has a single valid point on his list. A “normal” website would never need more than a shared hosting environment. If we are talking about a site that gets a large amount of traffic and needs to scale I don’t think you can call that “normal” since most sites don’t fall into that category.
If you are going to use EC2 you should use more than one instance. Use DNS to round robin between the different IPs or use a frontend server outside of EC2 to redirect to instances that are currently up.
You just need to think differently about needing block storage. For a wide number of applications there are very few writes that are critical. For critical writes there is the queueing service.
There are other tricks in the bag outside of hardware load balancing, see #1.
Again you just need to think differently.
From what I’ve seen there are few people who do hard core tweeks to the kernel to get better performance. You can tweek the OS just not the kernel.
This isn’t a valid point at all. There is a 10G limit on the size of an image and I bet in the majority of cases the image size is less than 2G.
A lot of these issues are brought up again and again because people want to use EC2 for a shared hosting replacement and I bet that wasn’t the use Amazon had in mind when they put it together.
I see this as FUD. You don’t need a $100K box and $100K of network infrastructure to do a lot of the stuff that people make you think you need that type of hardware for.
I don’t think Jason has a single valid point on his list. A “normal” website would never need more than a shared hosting environment. If we are talking about a site that gets a large amount of traffic and needs to scale I don’t think you can call that “normal” since most sites don’t fall into that category.
If you are going to use EC2 you should use more than one instance. Use DNS to round robin between the different IPs or use a frontend server outside of EC2 to redirect to instances that are currently up. You just need to think differently about needing block storage. For a wide number of applications there are very few writes that are critical. For critical writes there is the queueing service. There are other tricks in the bag outside of hardware load balancing, see #1. Again you just need to think differently. From what I’ve seen there are few people who do hard core tweeks to the kernel to get better performance. You can tweek the OS just not the kernel. This isn’t a valid point at all. There is a 10G limit on the size of an image and I bet in the majority of cases the image size is less than 2G.
A lot of these issues are brought up again and again because people want to use EC2 for a shared hosting replacement and I bet that wasn’t the use Amazon had in mind when they put it together.
I see this as FUD. You don’t need a $100K box and $100K of network infrastructure to do a lot of the stuff that people make you think you need that type of hardware for.