That is rather weird, since you don't seem to be using the stats search command at all (although "stats" could be "chart" in your case, I'm not sure how that works exactly.)
The approach that Simeon suggests is a good one.
I would suggest that you make the following change to your search (it should be much more efficient):
index=_internal source=*metrics.log splunk_server="*" | eval MB=kb/1024 | search group="per_host_thruput" | ...
Would be much faster if written as:
index=_internal source=*metrics.log splunk_server="*" group="per_host_thruput" | eval MB=kb/1024 | ...
The reason for this is there are tons of metrics events, and only some of them contain the term per_host_thruput , so by moving that to your first search you let splunk search for that within the index, instead of searching for it on a secondary pass over the events. (Of course, that wouldn't cause the error you are seeing, but it should make your search fastser and more efficient)
... View more