On 15-12-2011 18:11, Vlad Khorsun wrote: >> As I said before, ext4 performance with FW=ON is much slower than ext3, >> but I now found something new. >> >> I've verified that with FB 2.5, TCS times was ok (0s - 3s), and only >> with FB 3 they are very bad (first test takes 16s, then 12s, 8s, 7s, 7s, >> etc) >> >> What I found is that: >> >> Times was bad only with the default SharedCache=true. >> >> It happens because in this condition, both the cache_writer and >> garbage_collector calls PAG_init and it then sets FW=ON before the >> database is completely created! >> >> In FB 2.5, only running CREATE DATABASE does not even open it as forced >> write. It's just after some other command calling PAG_init that it happens. > > You mean PAG_header, not PAG_init > Yes, sorry.
>> So now we're spending time while the database is not even completely >> created. > > While we can restore previous behavior (put FW ON flag at header page > when database is created but create db file with this flag switched off) i > prefer > to ask - are we really want to maintain this hack ? Wouldn't it be better to > add > support of the all existing DBP tags to the CREATE DATABASE statement > and adjust TCS and gbak correspondingly ? > Maybe, but I don't think the wish for a faster CREATE DATABASE with FW=ON is a hack. If this is true, seems we have a design problem. Adriano ------------------------------------------------------------------------------ 10 Tips for Better Server Consolidation Server virtualization is being driven by many needs. But none more important than the need to reduce IT complexity while improving strategic productivity. Learn More! http://www.accelacomm.com/jaw/sdnl/114/51507609/ Firebird-Devel mailing list, web interface at https://lists.sourceforge.net/lists/listinfo/firebird-devel
