Have you found an answer yet? I don't know myself, and would have to research it.

I know this thread is hella old, but I need to know this, too, so let's put it somewhere Google will find it!

From the source:

printf("%4d, %d(%d):%.1f, %d(%d):%.1f, %d(%d):%.1f, %d(%d):%.1f, %d(%d):%.1f\n", time_count, ( s[0] + l[0] - prev_s[0] - prev_l[0] ), ( l[0] - prev_l[0] ), rt90[0], ( s[1] + l[1] - prev_s[1] - prev_l[1] ), ( l[1] - prev_l[1] ), rt90[1], ( s[2] + l[2] - prev_s[2] - prev_l[2] ), ( l[2] - prev_l[2] ), rt90[2], ( s[3] + l[3] - prev_s[3] - prev_l[3] ), ( l[3] - prev_l[3] ), rt90[3], ( s[4] + l[4] - prev_s[4] - prev_l[4] ), ( l[4] - prev_l[4] ), rt90[4] );

The Array indices are defined a couple lines earlier: New Orders, Payments, Order-Status, Delivery, and Stock-Level transactions. The line prints out every 10s and records these numbers for each of the above types: number of transactions processed successfully (plain printed), number of transactions that responded late (enclosed in parens ()), the 90th percentile response time for those transactions (after the colon : ).

So the output is something sort of like:

A, B(C), E(F):G, H(I):J, K(L):M, N(O):PWhere:A = time elapsed in secondsB = New Orders processedC = New Orders processed too lateD = 90th percentile response time for New Orders (in seconds)...

repeat for Payments (E-G), then Order-Status (H-J), Delivery (K-M), and finally Stock-Level (N-P)

I read somewhere the TPCC benchmark's metric is the number of new orders processed, so it isn't exactly correct to draw conclusions based on the sum of the rows... but w/e I don't really understand it either.