| Slobodan C 2005-04-25, 8:24 pm |
| We are using ASA 8.0.2.4299
Production server is on Solaris 9
I am doing some performance profiling. One of our apps loads entire DB
content into memory caches. DB is about 3 Mb, but expect it to grow to
10ish Mb, maybe more.
Looking at the profiler output I see a lot of I/O buffers and
character stuff (String, CharBuffer, etc.) Also noticed some unicode
conversion calls:
29 0.24% 10.33% 3023 40525
com.sybase.jdbc2.utils.PureConverter.toUnicode
31 0.23% 10.81% 1242 41100
java.lang. StringCoding$Charset
SD.decode
38 0.21% 12.32% 2831 39368
sun.nio.cs.SingleByteDecoder.decodeLoop
....
If I were to upgrade to the latest ASA 9.0 which supports UNICODE, and
were to redefine all the character columns so that they store UNICODE
characters, would conversion processing no longer be necessary? Would
the JDBC driver retrieve UNICODE characters from DB and supply them to
ResultSet without any conversion processing? Would this improve
performance?
I should mention that my schema contains around 100 tables and many
character (VARCHAR) columns. Probably more than half the data are
VARCHAR.
thanks
|