Hi! Martin Ebert wrote:
> > 100 questions per second will work with SUN- 4 Processor-maschine. > Tested; mysql/cachemode. > > The problem is (with mysql) the size of url.db (cachemode). > Works fine up to 1,5 Mio URLs. > > Don't ask me about more than 2 Mio URLs; works not so fine. :-(( Martin, what happens with more that 2 million URLs? How does it look like? Something goes wrong with URL table in MySQL, or something is wrong with word database? > But with given mysql the (our) question is: > Is it better to use file based ISAM for url.db, robots.db and so on? > Alex, any way for this idea? I don't think so. It is better to use third pary party stable products as a storage. Let database developers do their work. They develope their products for years, and they are gurus in storage development. We, instead of inventing new buggy storage, can save our time and spend it for development of indexing/search related things. >>>5) is it ready for a project sililar to google > > "No" is to hard. > > 1) mnogo is the best free idea for this. In the world. > 2) mnogo has concepional (spell?) ideas for load-balancing: > - indexer > - answerer (search.cgi) > but not for: URL-database. However, it is partially possible. You can already merge results from several searchd daemons, but this is not real distributed database, it is merging from separate databases. It our TODO we have an item to make it possible to distribute single search database between many machines. Actually, I think it's not so hard, there are some thoughts how to do it. And we already stared code reorganization to make it easier. > Ok, > 1) mnogo has a lot of these concepts basicly implemented. > 2) this is also our problem.... > ___________________________________________ If you want to unsubscribe send "unsubscribe general" to [EMAIL PROTECTED]