With the cumulative update of the database, is there any good plan not to soar to high memory?

as the title, my question is, for example, there is a table with five pieces of data in it, respectively

/* 1 */
{
    "_id" : ObjectId("5b50807f334ca1071be530aa"),
    "a" : ,
    "a" : 2
}

/* 2 */
{
    "_id" : ObjectId("5b508089334ca1071be530ac"),
    "a" : ,
    "b" : 3
}
/* 3 */
{
    "_id" : ObjectId("5b508092334ca1071be530ae"),
    "a" : ,
    "b" : 5
}

/* 4 */
{
    "_id" : ObjectId("5b508092334ca1071be530ae"),
    "a" : ,
    "b" : 4
}

first of all, my purpose is to sort. Let"s take our usual query condition of certain data to be higher than the others. Take the third item, Zhang San, for example. (suppose the query is found according to the b field and then sort the output.) now his b field is 5. I want to mention it to 1, OK, update it right away, and then Zhang San goes up, and then I have to accumulate and sort other data at once, so, Here comes the problem

I have to find all the other data except Zhang San"s data from the b field first, and then sort them (this is not the point). When the key points come, I have to update them in turn. All I can have is for, and then accumulate. In such a scheme, we have to deal with less data. If there are tens of thousands of items, isn"t it a pill? Memory skyrocketing, direct GG. Moreover, the nodejs,v8 memory is already tight. To do so is to commit suicide by poisoning yourself. If you have any plans, please give us some advice