After running a set of queries, it was found that, there are various values of TotalIOCount.
These values range from few hundreads to as high as 23 * 10^6.
On a system with 14 nodes and 504 AMPs, how to we say whether these values are significant, not - significant etc?
This truly depends on what is meant by significant.
The TotalIOCount is the amount of data involved in a query. It depends on the size of the tables, and the query specifics (joins, etc).
It is easier for CPU because you can compare AMPCpuTime with the maximum CPU capacity your system can deliver in one hour (number of CPU-seconds ), thus make a decision what number of AMPCpuTime is large and what is not.
For I/O it is not that straightforward because you have blocks of different size, some queries do block reads, some queries do full cylinder reads, etc.
What you can do is check the largest tables that are involved in the query, understand their size (using dbc.tablesize view), and start from there.