# Possible optimisations for large \_source documents **URL:** \ **Category:** Elasticsearch\ **Created:** [January 12, 2016, 2:00pm UTC](https://discuss.elastic.co/t/possible-optimisations-for-large--source-documents/38987 "2016-01-12T14:00:13Z")\ **Posts on this page:** 8\ **Page:** 1
**Author:** ![yanns](https://sea2.discourse-cdn.com/elastic/user_avatar/discuss.elastic.co/yanns/32/7094_2.png) [@yanns](https://discuss.elastic.co/u/yanns)\ **Post date:** [January 12, 2016, 2:00pm UTC](https://discuss.elastic.co/t/possible-optimisations-for-large--source-documents/38987/1 "2016-01-12T14:00:13Z")
Hi! For a query, I observe a performance difference if I include or not the \_source. With: "\_source": true, the query needs about 250 ms to complete. With: "\_source": false, the query needs about 50 ms to complete. In the mapping, I already disabled the \_all. I search about how to optimise this use-case and could not found any interesting information. If you have any tips: - about a possible configuration (disk, kernel, JVM...) - about how to work-around this. For example, has anybody tried to put the JSON as a String into one field, like `{"json": ""}`? Thanks in advance for any pointers, Yann ---
**Author:** ![dadoonet](https://sea2.discourse-cdn.com/elastic/user_avatar/discuss.elastic.co/dadoonet/32/137187_2.png) [@dadoonet](https://discuss.elastic.co/u/dadoonet)\ **Post date:** [January 12, 2016, 2:27pm UTC](https://discuss.elastic.co/t/possible-optimisations-for-large--source-documents/38987/2 "2016-01-12T14:27:40Z")
Note that you can still store the `_source` field but when you query, you can ask for other fields but `_source`. ---
**Author:** ![colings86](https://sea2.discourse-cdn.com/elastic/user_avatar/discuss.elastic.co/colings86/32/44960_2.png) [@colings86](https://discuss.elastic.co/u/colings86)\ **Post date:** [January 12, 2016, 2:37pm UTC](https://discuss.elastic.co/t/possible-optimisations-for-large--source-documents/38987/3 "2016-01-12T14:37:21Z")
The JSON of the indexed document is already stored in a single field (called `_source`) . A couple of questions about your use case: - How many documents are you retrieving? (Maybe you could share you whole search request?) - How much memory have you allowed for the FileSystem Cache? ---
**Author:** ![yanns](https://sea2.discourse-cdn.com/elastic/user_avatar/discuss.elastic.co/yanns/32/7094_2.png) [@yanns](https://discuss.elastic.co/u/yanns)\ **Post date:** [January 12, 2016, 4:01pm UTC](https://discuss.elastic.co/t/possible-optimisations-for-large--source-documents/38987/4 "2016-01-12T16:01:54Z")
yes thx! I know that already. In that case, I'm interested in the whole `_source`. ---
**Author:** ![yanns](https://sea2.discourse-cdn.com/elastic/user_avatar/discuss.elastic.co/yanns/32/7094_2.png) [@yanns](https://discuss.elastic.co/u/yanns)\ **Post date:** [January 12, 2016, 4:07pm UTC](https://discuss.elastic.co/t/possible-optimisations-for-large--source-documents/38987/5 "2016-01-12T16:07:17Z")
> [@colings86](#): > > FileSystem Cache Yes I am speaking exactly about this single field `_source`. When I deactive the retrieving of `_source` in the query results, I observe this difference of performance. I am retrieving 20 documents. I am not sure about what you ask precisely with FileSystem Cache. The `free` command outputs: ```auto $ free total used free shared buffers cached Mem: 15396852 12732848 2664004 4296 153104 3033332 ``` ---
**Author:** ![yanns](https://sea2.discourse-cdn.com/elastic/user_avatar/discuss.elastic.co/yanns/32/7094_2.png) [@yanns](https://discuss.elastic.co/u/yanns)\ **Post date:** [January 12, 2016, 4:07pm UTC](https://discuss.elastic.co/t/possible-optimisations-for-large--source-documents/38987/6 "2016-01-12T16:07:59Z")
I'll try to compress the http responses to check if it makes a difference. If you have any other idea, feel free to suggest. Thx! ---
**Author:** ![cheapsteak](https://sea2.discourse-cdn.com/elastic/user_avatar/discuss.elastic.co/cheapsteak/32/17313_2.png) [@cheapsteak](https://discuss.elastic.co/u/cheapsteak)\ **Post date:** [April 12, 2017, 7:50pm UTC](https://discuss.elastic.co/t/possible-optimisations-for-large--source-documents/38987/7 "2017-04-12T19:50:26Z")
We're running into the same problem, when including `_source`, it takes 2+ seconds for a size of 50, when `_source` is false it takes 400ms This is our query ``` GET /case_centric/case_centric/_search?pretty { "size": 50, "from": 0, "_source": { "includes": ["diagnoses.days_to_last_follow_up"] }, "query": { "bool": { "filter": [ { "nested": { "path": "diagnoses", "query": { "bool": { "must": [ { "exists": { "field": "diagnoses.vital_status" } } ] } } } } ], "should": [ { "bool": { "should": [ { "nested": { "path": "diagnoses", "query": { "bool": { "must": [ { "exists": { "field": "diagnoses.days_to_death" } } ] } } } }, { "nested": { "path": "diagnoses", "query": { "bool": { "must": [ { "exists": { "field": "diagnoses.days_to_last_follow_up" } } ] } } } } ] } } ] } } } ``` ---
**Author:** ![system](https://us1.discourse-cdn.com/elastic/original/3X/1/a/1ac57faf039f6b580b3f104ef42a2a89e41014de.png) [@system](https://discuss.elastic.co/u/system)\ **Post date:** [July 5, 2017, 10:01pm UTC](https://discuss.elastic.co/t/possible-optimisations-for-large--source-documents/38987/8 "2017-07-05T22:01:14Z")