docvalues和Fieldcache
Fieldcache:
docID->document->fieldvalue
无论是聚类排序关联等,首先都需要获得文档中某个字段的值,通过docID去获得整个document,然后再去获得字段值,term转换得到最终值,FieldCache一开始就缓存了所有文档的某个特定域(所有数值类型以及不分词的stringField)的值到内存,便于随机存取该域值!
Fieldcache实现过程:
http://moshalanye.iteye.com/blog/281379
缺点:
1. 常驻内存,大小是所有文档个数特定域类型大小
2. 初始加载过程耗时,需要遍历倒排索引及类型转换
Docvalues:
docID->fieldvalue
建索引时,建立了document到field value的面向列的正排索引数据结构,直接通过已知的docID定位到字段值,从而无需加载document,亦不需要term转换,遍历term找寻doc等的过程
优点:大约节省三分之一的内存!
缺点:由于是硬盘读取,而非内存模式,对于大批量的使用下,优势明显,速度更优;小量情况下没有内存快!总体会慢15-20%
20 February 2015 - Apache Lucene 5.0.0 and Apache Solr 5.0.0 Available
FieldCache is gone (moved to a dedicated UninvertingReader in the misc module). This means when you intend to sort on a field, you should index that field using doc values, which is much faster and less heap consuming than FieldCache.
LUCENE-5666:Change uninverted access (sorting, faceting, grouping, etc) to use the DocValues API instead of FieldCache
Es中
https://www.elastic.co/guide/en/elasticsearch/guide/current/doc-values.html
Sorl中
https://cwiki.apache.org/confluence/display/solr/DocValues
posted on 2016-09-01 16:00 lovebeauty 阅读(755) 评论(0) 编辑 收藏 举报