MongoDB performance

Question

I need to design mongoDB database for save history notifications. And I consider two possible solutions.

1) every user will have 1 document like this:

{
_id: 1234567890abcdefgh,
emailfrom: email1@example.com,
history: {
  email2@example,com: {
    {
      read: 1,
      text: 'hey man',
      time: 2015-11-20-23-05-15
    },
    {
      read: 0,
      text: 'hey whats up',
      time: 2015-11-20-23-10-10
    }
  },
  email3@example,com: {
    {
      read: 1,
      text: 'oooh',
      time: 2015-11-20-23-05-11
    }
  }
}

}

2) second solution is make for every notificaion own document like this:

{
 _id: 1234567890abcd
 emailfrom: email1@example.com,
 emailto: email2@example.com,
 text: 'hey man',
 read: 1,
 time: 2015-11-20-23-05-15
},
{
 _id: 1234567890abcd
 emailfrom: email1@example.com,
 emailto: email2@example.com,
 text: 'hey whats up',
 read: 0,
 time: 2015-11-20-23-05-15
},
{
 _id: 1234567890abcd
 emailfrom: email1@example.com,
 emailto: email3@example.com,
 text: 'oooh',
 read: 1,
 time: 2015-11-20-23-05-11
},

My question is about performance of this two approaches. when there will be thousands and thousands notifications and I will want to select, update or find:

1) find all notifications send from email1@example.com AND read: 0 - I think first approach will be faster

2) I will want to save new notification - I think second approach will be faster

3) I will want to update read 0 -> read 1 emailfrom email1@example.com emailto email2@example.com - I don't know which approach would be faster.

Could anyone help me with this? Which one is the right approach for save this type of data? Thank you for any comment!

Answer 1

I guess there is a typo error in the first approach. The type of "email2@example,com" and "email3@example,com" should be array, right? Whatever, according to the use cases you mentioned, I prefer to choose option 2. Just add proper index to that table, such as compound index for " emailfrom + read " and " emailfrom + emailto " which can ensure you have high performance during query.

The most common case is query by emailfrom and another condition which option 1 cannot help too much. And since there will be many many notification coming in future, you still consider there is a 16M limitation for each document size in mongoDB.

Answer 2

If you think about thousands of notifications in the future, I strongly advice against storing embedded history of notifications in every user's document due to limitation of document size.

HERE is the article shows some approaches to designing a mongoDB schema which seems to be similar to your (inbox).

Additionally, HERE is a great, short guide from docs.mongodb.org which can help you to make decision what MongoDB data model will be the best for you.