Sphinx 1.10 版本令人激动，在实时搜索、高并发、分布搜索效率、稳定性、功能、接口上都有不凡的表现：
# added RT indexes support (新功能，很可能会马上测试它)
# added prefork and threads support (workers directives) (非常好的改进，经简单测试，开启 prefork 后，一台双核 Xen 550x，并发从 600 提升到 2500, 数据量 180w; 如果把 poll 换成 epoll, 效率能否再高点？)
# added multi-threaded local searches in distributed indexes (dist_threads directive) (改进 distributed 模式响应时间)
# added common subquery cache
# added string attributes support (存放非整型关键ID，比如 GUID ~)
# added indexing-time word counter
# added CALL SNIPPETS(), CALL KEYWORDS() SphinxQL statements
# added field_weights, index_weights options to SphinxQL SELECT statement
# added insert-only SphinxQL-talking tables to SphinxSE
# added select option to SphinxSE queries
# added backtrace on crash to searchd
# added SQL+FS indexing, aka loading files by names fetched from SQL
# added a watchdog in threads mode to searchd (稳定性)
# added automatic row phantoms elimination to index merge
# added hitless indexing support
# added --check, --strip-path, --htmlstrip, --dumphitlist ... --wordid switches to indextool
# added --stopwait, --logdebug switches to searchd
# added --dump-rows, --verbose switches to indexer
# added "blended" characters indexing support
# added joined/payload field indexing
# added FlushAttributes() API call
# added query_mode, force_all_words, limit_passages, limit_words, start_passage_id, load_files, html_strip_mode, allow_empty options, and %PASSAGE_ID% macro in before_match, after_match options to BuildExcerpts() API call
# added @groupby/@count/@distinct columns support to SELECT
# added query-time keyword expansion support
# added query batch size limit option
# added SINT() function to expressions
# improved SphinxQL syntax error reporting
# improved expression optimizer (better constant handling)
# improved dash handling within keywords (no longer treated as an operator)
# improved snippets (better passage selection/trimming, around option now a hard limit)
# optimized index format that yields ~20-30% smaller indexes (灰常好)
# optimized sorting code (indexing time 1-5% faster on average; 100x faster in worst case) (灰常好)
# optimized searchd startup time (moved .spa preindexing to indexer), added a progress bar (灰常好)
# optimized queries against indexes with many attributes (eliminated redundant copying)
# optimized 1-keyword queries (performace regression introduced in 0.9.9)
# optimized SphinxQL protocol overheads, and performance on bigger result sets
# optimized unbuffered attributes writes on index merge.
From:
http://www.sphinxsearch.com/news/42.html
http://sphinxsearch.com/docs/current.html#rel110
--EOF--