10
votes

Issue

I have an Express (Node.JS) + MongoDB app with a server response load time of 4 - 7 seconds on average (slow).

I understand that the average server response time is under 200ms as per google pagespeed tools.

This app is fetching data from a mongoDB asynchronously but the roundtrip times to the database is extremely slow with each call averaging about 500ms - 1s. These calls are simple findAll calls to retrieve data of less than < 100 records.

Context

  • Mongoose version: 4.13.14
  • DB server's MongoDB version is 3.4.16
  • DB server is hosted on MongoDB Atlas M10 in AWS / Oregon (us-west-1)
  • Web server is hosted with now.sh in SFO1 (us-west-1)
  • Have performed recommended indexes as advised by MongoDB Atlas's performance advisor
  • Data fetching perfectly fine in local environment (local server + local db) as data is queried in a matter of few ms
  • Mongoose logs for the affected page can be found in this gist

Mongo Server configuration

  • Mongo Atlas M10
  • 2GB Ram
  • 10 GB Storage
  • 100 IOPS
  • Encrypted
  • Auto-expand storage

Attempted solutions:

I have checked my DB metrics, they looked fine. There are also no slow queries. These are simple findAll queries. Performance advisor on mongo atlas reports nothing unusual.

The production application and database are both hosted in the same region.

I have already tried optimising the application layer of the query (mongoose) by running .lean()

Question:

Where else should i look to improve the database latency? How can a simple query take so long? Otherwise, why is my server response time taking up to 4s when the expected is about 200ms?

3
Are your database calls going through an ORM/Driver? e.g Mongoose? That's the only thing I can think ofProdigle
Sounds like there are a few areas of concern, the DB taking ~500-1s for < 100 records is pretty poor so you might want to think about indexing. However, outwith that, it's very difficult to diagnose a problem with no hard details like where it's hosted, who it's hosted with, what level of hosting (tier) it's running on, what physical hardware it's on, how distributed the app is.....and so on.James
@James Thanks! I've updated the questionclodal
The roundtrip time is 0.5 -> 1s only? And 2-3s to wait for DB processing?Mạnh Quyết Nguyễn
Can you enable debug mode and post the logs here? Also have you tried the mongo shell to connect to your remote database directly? What is the response time when you try Mongo shell?manishg

3 Answers

2
votes

Hey you can try hosting your server and database in the same region. I think the network is creating a overhead in this case. If the server and the database are in the same region, They are on the same network which will reduce the latency significantly. there is a diagram on aws for this enter image description here

0
votes

I add some problem like yours with an app that i developed in my master degree. I add to put a node.js api running online to present it in class room.And i realized that every time i wanted to make a call in the api the response was taking allot of time. I realized that one of the problems was the school network because of the firewalls. Also the place where i put the server heroku.com was giving some delay as well. What i did was use Redis ( https://redis.io/ ) to improve the performance, also heroku was giving me some problems because of the requests being http and not https.

Make a test running the app and data on your localhost and see the performance. if you don´t have any issue try to check if nothing is messing with your request like the place where you host your node server.

Let me know if this helps or if you still have issues so i can try to help you out better.

0
votes

I had the same issue once with my nodejs code using the same development stack(mongodb,nodejs), I got into trouble of late response from api, and after spending a lot of time I found my server the real culprit I then changed from heroku to amazon aws EC2 instance and things started working fast and amazingly fast, so probably your web server is culprit


to make sure mongodb is not culprit, write an api endpoint where you can just return some json response without making any query to database.