| Description | Because the responsiveness for throughput-oriented
transactions is divided by the elapsed time, it can be very
sensitive to short-term performance variations for
transactions that take a short period of time. For example,
when downloading a very short file, a single dropped packet
could double or triple the total response time.
Further, throughput is usually examined for applications that
transfer a lot of data, and when doing so it is helpful to
conceptualize transaction costs that are proportional to the
amount of data separately from those costs that are relatively
fixed (i.e., independent of the amount of data). For very
short transactions, these fixed transaction costs (handshake,
setup time, authentication, round-trip time) may dominate the
total response time for the transaction, resulting in
throughput measurements that aren't really proportional to the
network's, server's and client's combined data throughput
capability.
This object controls the minimum number of seconds that an
throughput-based transaction must exceed before an exception
can be generated for it. If this object is set to zero, then
all throughput-based transactions are candidates for
exceptions.
The value of this object must persist across reboots. |