Statistics causing Shared Memory Problems

I am having an interesting problem after changing some of our update scripts. Each night we load mainframe data and update existing tables and insert new information. In the past, I have been updating statistics in individual fields to high, but changed this to update the whole table - about 250,000 recs - 120 fields. This takes over an hour, and forces online to grab new shared memory segments. Well, while it is doing this, users get locked out of the system - the login's hang, and I get unable to allocate memory errors during query processing in dbaccess. It grabbed so many shared memory segments one time that the online.log said the kernel parameters had been overloaded. Is this normal behavior??? The speed increases on the table after it is done are incredible!!! I hate to revert back to the old way - or can I do this once a week?? James Saville