查询的mongodb'count'非常慢

Question

大家好,我使用mongodb 2.4.6版本和Windows 2008 64位版本.

我有一个有200万条记录的集合,需要在客户端进行搜索和分页.

db.products.find({"catalogs":1205}).skip().limit() is very fast .

但是当计算总记录数时:

db.products.find({"catalogs":1205},{"_id":1}).count() is too slow.

>> 442312 records.

>>[log] Sat Sep 28 00:20:01.566 [conn10] command products.$cmd command: { count: "products", query: { catalogs: 1205.0 }, fields: { _id: 1.0 } } ntoreturn:1 keyUpdates:0 locks(micros) r:460681 reslen:48 460ms

这个计数命令经过的时间是460ms,太慢了.如果我们有很多请求那么可怕.

我为'catalogs'字段创建了一个索引,并且无法使用$inc命令,因为查询可能非常复杂.

谷歌搜索一些像这个问题,发现这个'计数'性能错误已经修复mongodb 2.4版本.

来自http://docs.mongodb.org/manual/release-notes/2.4-overview/

Improvements to count provide dramatically faster count operations. Counting is now up to 20 times faster for low cardinality index based counts.

有什么方法可以提高数量？谢谢.

更新一些信息

> db.products.getIndexes()
[
    {
            "v" : 1,
            "key" : {
                    "_id" : 1
            },
            "ns" : "products.products",
            "name" : "_id_"
    },
    {
            "v" : 1,
            "key" : {
                    "catalogs" : 1,
                    "created" : -1
            },
            "ns" : "products.products",
            "name" : "catalogs_1_created_-1"
    }
]

查询和已用时间:

>db.products.find({"catalogs":1205},{"_id":1}).limit(20)
>>Tue Oct 01 15:39:19.160 [conn2] query products.products query: { catalogs: 1205.0 } cursorid:277334670708253 ntoreturn:20 ntoskip:0 nscanned:21 keyUpdates:0 locks(micros) W:5045 r:1017 nreturned:20 reslen:704 1ms

查询exaplin:

>db.products.find({"catalogs":1205},{"_id":1}).explain()

{
    "cursor" : "BtreeCursor catalogs_1_created_-1",
    "isMultiKey" : true,
    "n" : 451466,
    "nscannedObjects" : 451466,
    "nscanned" : 451466,
    "nscannedObjectsAllPlans" : 451466,
    "nscannedAllPlans" : 451466,
    "scanAndOrder" : false,
    "indexOnly" : false,
    "nYields" : 2,
    "nChunkSkips" : 0,
    "millis" : 2969,
    "indexBounds" : {
            "catalogs" : [
                    [
                            1205,
                            1205
                    ]
            ],
            "created" : [
                    [
                            {
                                    "$maxElement" : 1
                            },
                            {
                                    "$minElement" : 1
                            }
                    ]
            ]
    },
    "server" : "WIN-O47CO6C2WXY:27017"

}

Answer 1

此计数查询不是特别快的原因是它必须扫描索引中的 451466 个条目才能对条目进行计数。换句话说，相对于满足查询的条目的索引和大小，您的查询不是很有选择性。