When a backend is slow but its responses are the same for many users, cache them in Nginx. Declare the cache store at the http { } level, then switch it on per location.

At http { }:

proxy_cache_path /var/cache/nginx
    levels=1:2
    keys_zone=app_cache:10m   # 10m of keys ≈ ~80k entries
    max_size=1g
    inactive=60m
    use_temp_path=off;

In the server { }:

location / {
    proxy_pass http://127.0.0.1:3000;
    proxy_set_header Host $host;

    proxy_cache app_cache;
    proxy_cache_valid 200 301 302   10m;   # cache OK/redirects for 10 min
    proxy_cache_valid 404           1m;
    proxy_cache_key   "$scheme$request_method$host$request_uri";

    # Serve stale content while refreshing or if the backend errors.
    proxy_cache_use_stale error timeout updating http_500 http_502 http_503 http_504;
    proxy_cache_background_update on;
    proxy_cache_lock on;                    # collapse a stampede into one upstream fetch

    add_header X-Cache-Status $upstream_cache_status;   # HIT / MISS / EXPIRED / STALE
}

How it behaves:

The first request is a MISS (goes to the backend and is stored); subsequent identical requests are HIT (served from disk) until proxy_cache_valid expires.
X-Cache-Status lets you verify with curl -I — look for x-cache-status: HIT.
proxy_cache_use_stale + background_update keep the site fast and resilient: users get slightly stale content instantly while Nginx refreshes in the background, and stale content is served if the backend is down.
proxy_cache_lock on; prevents a thundering herd — only one request repopulates an expired entry.

Don’t cache personalised pages. Bypass the cache when a session cookie is present:

    proxy_cache_bypass $cookie_sessionid;
    proxy_no_cache     $cookie_sessionid;

Also respect upstream Cache-Control: no-store/private — Nginx won’t cache those by default.

sudo nginx -t && sudo nginx -s reload

Cache upstream responses with proxy_cache