Thank you for all these suggestions; I have done most of them such as
indexing, using caches and memory
My indexes are complicated with 300 thousand word referencing from part
descriptions. To avoid duplicates a new word is checked and then a
secondary index record is built. This means millions of db accesses for
index creation - only done once per update for the whole website.
I think the solution is more fundamental than .NET programming and
connected to DBMS and limits of MS SQL (eg verus DB2, ORACLE).
One solution is to avoid Web updates and run a job direct against the
database which is much quicker. But I dont have server access (apart
from via the WWW), so I cant run a task (eg a BAT job) on the server.
I am interested to know of any hosts offering DB2 or ORACLE as a
standard package. I know ORACLE has powerful facilities but the cost is
higher.