How to optimize the performance of large documents?

I'm trying to build a program for Online Excel, but I'm having performance issues.
In my implementation, a workbook is a document. The json format of the document is as follows:

{
    "activeSheetId": "aaa",
    "sheets": {
        "aaa": {
            "cells": {
                "0": {
                    "0": {
                        "value": "123"
                    }
                }
            }
        },
        "bbb": {
            "cells": {}
        }
    }
}

op format is as follows.

//set value
{ type: 'setValue', sheetId: 'aaa', row: 0, col: 0, value: '123' }
//add worksheet
{ type: 'addSheet', sheetId: 'bbb', sheetJson: { cells: {} } }

The performance issue is that the workbook's json is so large, and each time applyOp reads the entire workbook's json from the database.

I expect each worksheet to be a subdocument, and each applyOp reads only the main document (workbook) and the corresponding subdocument (worksheet) based on op.

Start at applyOp and trace how the workbook JSON is loaded from the database for each operation. Evaluate the proposed workbook-and-worksheet subdocument model for setValue and addSheet operations; done means operations read only the main document and the affected worksheet without loading the full workbook.