sql server - Making sense of the number of reads/writes in SQL Profiler -

see sql statements indicated in sql profiler view. these events coming 1 client machine busy long-running process, working through couple thousand rows. each processing of row takes 6.5 seconds, sql profiler showing time between logouts, though actual update statement takes 1ms. each login/logout uses same spid. see between given login , logout event, sql statements indicate reads count of 17 , writes count of 0. yet, logout event indicates total reads count in excess of 2million , writes count in excess of 10k. events need profiling try figure out statement causing reads/writes, because suspect ones causing 6.5 second delay, yet can't see them happening?

the reads/writes figures given audit:logout event cumulative totals duration of connection. in itself, values don't tell detail - if run 10 commands within lifetime of connection, you'll seeing total figures 10 commands within session.

to know breakdown statement, need @ sql:batchcompleted (or sql:stmtcompleted) events recorded same spid between starting audit:login event, , ending audit:logout event.

update: looking @ image, thing looks bit odd (to me @ least), after each audit:logout, reads value not reset, , each time being incremented number of reads statement being executed (17). i'm not sure 100% therefore @ point figure reset - basic point figure cumulative , have been built on time/a number of statements doesn't mean have heavy query hitting server!

i suspect reads/writes figures audit:logout behaving cumulatively above. duration bucks trend. looks duration not cumulative. descriptions msdn:

duration : amount of time since user logged in (approximately).
reads: number of logical read i/os issued user during connection.
writes: number of logical write i/os issued user during connection.