NetBase for MPE Reference Guide

Welcome to NetBase
v
NetBase Performance
NetBase is fast! Of all the features that NetBase offers, performance is perhaps the most valuable.
Networking a transaction is more costly in terms of elapsed time and machine resources than
performing the same transaction locally. The key to enhanced performance is that NetBase
minimizes the number of network transactions.
Minimizing Network Transactions
When NetBase was developed, great care was taken to ensure that the code was written as
efficiently as possible. However, the main reason that NetBase is efficient is Quest’s approach in
implementing network file access. The major cost associated with accessing remote files results
from the overhead incurred transporting the request and result between computers. Most of this
additional overhead comes not from the actual transfer of data across the LAN, but is incurred in
the preparation of the data to be transported and the handling of the data on the other side of the
line. Sending 1000 bytes of data requires about the same overhead as sending 10 bytes.
To illustrate this point, consider the following: A typical DBGET on an unloaded Series 947 takes
less than 1 millisecond. The difference in elapsed time is the overhead required to prepare and
route the request across the LAN link and return the result to the requesting process. Keeping in
mind that most of the overhead was incurred preparing the transaction and not transferring the
data, a large performance benefit lies in getting the most out of each network transaction. For
example, if the number of network transactions can be cut in half, the throughput will be almost
doubled.
NetBase utilizes two techniques for reducing the number of network transactions. Anticipatory
reading by NetBase yields the highest performance savings. NetBase also piggybacks calls together
whenever possible to further reduce the number of network transactions. Some calls lend
themselves to being serviced by both techniques.