DEV Community

Alin Climente
Alin Climente

Posted on

Python/Mongo website on a small 1GB RAM VPS

I just released a small website which will most probably be taken down in 1-2 years because of lack of users. This was an idea which bugged me for a long time, so I had to do it (a little bit of OCD).

The website is pretty simple and does the following:

  • scrape some data once per day;
  • clear old data once per day;
  • serve that scraped data with Python/FastAPI from MongoDB;
  • added some extra features on that scraped data so users can do something with it;
  • SSR html pages served with Jinja2 with some custom logic;

Docker services running on that small 1GB RAM, 1CPU, VPS (Linode/Nanode):

  • Python/FastAPI web application (serves data to users);
  • Python Worker (which scrapes/cleans the data in the background);
  • MongoDB (holds all data);
  • Caddy2 as a reverse proxy;

The amount of RAM free with the website idle 93MB RAM.

Tasks: 118 total,   1 running, 117 sleeping,   0 stopped,   0 zombie
%Cpu(s):  0.7 us,  1.0 sy,  0.0 ni, 98.0 id,  0.0 wa,  0.0 hi,  0.0 si,  0.3 st
MiB Mem :    969.4 total,    93.2 free,    315.9 used,    519.3 buff/cache
MiB Swap:    512.0 total,    438.7 free,     73.3 used.    506.2 avail Mem 
Enter fullscreen mode Exit fullscreen mode

I tried a bit of load testing using baton (a golang cli load testing tool).

Here are some load testing stats on that VPS with 100 concurent users making 1000 requests:

====================== Results ======================
Total requests:                                  1000
Time taken to complete requests:        14.646065944s
Requests per second:                               68
===================== Breakdown =====================
Number of connection errors:                        0
Number of 1xx responses:                            0
Number of 2xx responses:                         1000
Number of 3xx responses:                            0
Number of 4xx responses:                            0
Number of 5xx responses:                            0
=====================================================
Enter fullscreen mode Exit fullscreen mode

It took around 15 seconds to complete...but no errors which is good!

Now here is the same on my laptop (16GB RAM, 7 CPU's):

====================== Results ======================
Total requests:                                  1000
Time taken to complete requests:         128.916784ms
Requests per second:                             7757
===================== Breakdown =====================
Number of connection errors:                        0
Number of 1xx responses:                            0
Number of 2xx responses:                            0
Number of 3xx responses:                         1000
Number of 4xx responses:                            0
Number of 5xx responses:                            0
=====================================================
Enter fullscreen mode Exit fullscreen mode

It took around 130 miliseconds to complete!

Some stats with 20 requests made by 1 user on the VPS:

====================== Results ======================
Total requests:                                    20
Time taken to complete requests:         1.271508214s
Requests per second:                               16
===================== Breakdown =====================
Number of connection errors:                        0
Number of 1xx responses:                            0
Number of 2xx responses:                           20
Number of 3xx responses:                            0
Number of 4xx responses:                            0
Number of 5xx responses:                            0
=====================================================
Enter fullscreen mode Exit fullscreen mode

The request used for load testing does the following:

  • makes a query to MongoDB for 10 random posts;
  • renders those posts with Jinja2 on the server + some template logic (if else);

Results may differ for your use case and tech used, but I think this is a good average for a Python/MongoDB website.

Top comments (1)

Collapse
 
vulcanwm profile image
Medea

nice!