My mongo query read run more than 5000 milliseconds - mongodb

Question

I have a mysql as primary db and mongodb as secondary database.
I run a query on production and it run for more than 5 seconds.
Here is my query when i run with explain
{
"cursor" : "BtreeCursor host_1_type_1 multi",
"isMultiKey" : false,
"n" : 1,
"nscannedObjects" : 1,
"nscanned" : 313566,
"nscannedObjectsAllPlans" : 313553,
"nscannedAllPlans" : 627118,
"scanAndOrder" : false,
"indexOnly" : false,
"nYields" : 14,
"nChunkSkips" : 0,
"millis" : 6555,
"indexBounds" : {
"host" : [
[
"",
{
}
],
[
/shannisideup/i,
/shannisideup/i
]
],
"type" : [
[
"ambassador-profile",
"ambassador-profile"
]
]
},
"server" : "mongoserver:27017"
}
i've added an indexes
[
{
"v" : 1,
"key" : {
"_id" : 1
},
"ns" : "db.visitor",
"name" : "_id_"
},
{
"v" : 1,
"key" : {
"host" : 1,
"type" : 1
},
"ns" : "db.visitor",
"name" : "host_1_type_1"
}
]
But i still don't know why is it so slow when running query
db.visitor.find({ host:/shannisideup/i, type:"ambassador-profile" }).limit(1)
FYI i run a different server for my apps and my mongodb server in AWS cloud.
For which mongodb running EC2 m3.medium, i've tried raising the openfile limit as mongodb website suggested. My mongodb is running on separate 100GB disk mounted via /dev/sdf.
I run mongodb version 2.4.5
When my mongodb runs, the CPU load almost always 100%
My MMS stat for opcounter are:
command: 19.11
query: 13.79
update: 0.03
delete: 0.00001
getmore: 0
insert: 0.03
My highest Pagefaults is 0.02
What can i do to optimize my mongodb to less then 1 second?

Answer 1

I'll walk thru a few things that should help, although I want to make clear up front that a case insensitive regex search ("host:/shannisideup/i") cannot use an index so there are limits to what you can do with this data model and search. You can see that it's scanning a large number of objects ("nscanned" : 313566) just to return 1 document.
Things to do:
Upgrade your instance with a faster cpu - the 100% cpu load during
this case insensitive search is a clearly indicator that at times
your database is cpu bound. A faster cpu will help. More memory
(which typically comes with Amazon EC2 instances when you move to a
faster cpu) won't hurt either. Same with a faster disk (SSD).
Lower case your host field before storing it in MongoDB - if you
store all your hosts in lower case and then lowercase your search
string prior to search, you can get rid of the case-insensitive flag
for the regex. If you can combine that with a prefix operator in your
regex you would be able to use the index which would help.
Consider using a dedicated text search product (like Solr or
Elasticsearch) for the host query. MongoDB is limited in it's text
search capabilities, particularly with regard to wild card searches.
You may find that something like Elasticsearch or Solr may provide
better performance.
A few links that you may find useful:
http://docs.mongodb.org/manual/reference/operator/query/regex/
http://selectnull.leongkui.me/2014/02/02/mongodb-elasticsearch-perfect-nosql-duo/