Does creating too many indexes slows down mutation and reading?


(Rohan) #1

I have a schema which looks like this :

Using this schema I find related users with this query :

Can I optimize the schema even further to improve latency of mutations and reading ?

My Current Setup is :

CPU and Ram: Ec2 4 (Intel® Xeon® CPU E5-2686 v4 @ 2.30GHz) Cores 16Gb RAM
Edges : edge_count:33.60M [Output after bulk loading]
Single Server Single Alpha

Currently it takes
For 1 Transaction : 2 secs [3 searches and 3 mutations]
For query : 100 ms to 500 ms depending upon how much connected the user is.


(Michel Conrado) #2

Does creating too many indexes slows down mutation and reading?

Not exactly. Only if your query needs a lot of indexes. Or you’re doing a bulk. In general, don’t.

In theory, yes, but are you using HDD or SSD/NVMe?
IMHO Strategic indexing are great. Define your query predicates too (don’t use expand_ALL all the time or even in prod), you need predictable responses.

Instead of a single Alpha, why don’t you try exploring different setups? BTW, if so, each Alpha would work better having their own physical storage. Case if you use the same machine.


(Rohan) #3

What are the other different setups ? multi and single VM cluster u mean ?


(Michel Conrado) #4

No vm, I mean something like 2 groups with 3 alphas each.