Generating new names for big databases is much too slow

[westram]

• example : mbiocluster:/home/ludwig/16s/6_silva_15jan07_ext.arb
• remove names file (to force regeneration)

⇒

• 99% CPU load caused by arb_name_server
• runs more than 2 hours (with constantly increasing time-prediction).

Reasons:

• A hash used for name lookup had fixed size.
• Name lookup used a linked list (omg).

Solutions:

• Replaced list by hash-
• hash- grow if needed

Runtime now 20% of old-runtime (for 10000 new species, far better speedup if more).