Or What are the two features of SSDs that this simple IO model (as in HDD) fails to capture?

What are the parameters which will change when we will move from HDDs to SSDs.

Basically, as I said, we made two assumptions about disk cost which are fair with HDDs. The first is that HDDs will always read a large chunk of data at each time, as the cost of IOs does not depend much on the size of the IO. This is no longer true with SSDs; as the cost of initiating an IO is low, the cost of the data transfer is relatively much more important, and so it becomes possible to request small amounts of data.

The second assumption is that the cost of reading and the cost of writing are identical. This is indeed true for HDDs. With SSDs, this is not necessarily so, but it is quite disk dependent. For low end disks, random writes can be very expensive, while random reads are very cheap. Sequential reads are generally cheaper than random reads, sequential writes faster than random writes, and reads faster than writes. But this can be quite SSD dependent.