My forum is scoring very badly in Lighthouse performance category and is being penalized on Google
Im wondering is there any way to improve this speed? my site has around 2k members, 100k posts, and 5k topics. I am using 4gb ram plan from DigitalOcean.
In discourse, I am using all default settings only (with Cloudflare template enabled).
Though I do have quite a few plugins installed
Is there any tips to improve speed of my Forum?
Beacuse of that? I’m totally sure it is not.
According to Google Search Console, in the performance section it shows that the site’s loading time is slow. I’ve heard that PageRank considers this when ranking websites.
And yet it is not slow and Google is not penalting you. Out there is huge amount of mis- and disinform about PageRank and this claim is one of that category.
Otherwise optimizingquestion is really is easy. You can’t do much. On Wordpress there would be options because PHP and how sites are working but regarding an app as Discourse there isn’t similar tools.
Sure, you can start searching faster VPS and/or DNS but that’s it and you will not getting actual and real worth to your money.
This isn’t accurate. Page speed has been a search signal for a pretty long time now, albeit not a high ranking one. Google used data from both their crawler view, and CrUX - I’m not sure if that’s still the case.
Yes, but not in that secondlevel in the meaning it woulf affect PageRank and true/meaningfull SEO.
It does. Evaluating page experience for a better web | Google Search Central Blog | Google for Developers
Search has added a variety of user experience criteria, such as how quickly pages load and mobile-friendliness, as factors for ranking results.
If you have a way to check the speed with all these disabled, that might be quite informative. Discourse isn’t inherently slow - my lighthouse score is 100.
If you can, measuring with half of them enabled will allow you to narrow down perhaps to a single plugin which is causing performance problems. But it might not be just one.
Hello 
Just an other approach… I always had problem with DO this type of servers. It was slow for my community activity. So I move the server to Vultr. I think Vultr High Frequency servers are the best in this price. Discourse much prefers the single core faster CPU then multicore slower ones. As I know, only Vultr offers 3GHz+ vCPU on there HF lines. I never have problem with speed on these servers. Super fast and stable! 
I use them 2vCPU, 4GB ram, 128GB NVMe SSD server. I have been using Vultr HF servers for years. Definitely worth to try, I highly suggest it.
Thanks for the tip. Do you have measurements on Vultr servers somewhere?
Compared with DO I haven’t but I will make a rebuild with timer once I can, if I remember correctly it usually about 5min and also check the server actual performance. I will update this post. 
CPU info:
Intel Core Processor (Skylake, IBRS)
cpu MHz: 3695.998
processor : 0
vendor_id : GenuineIntel
cpu family : 6
model : 94
model name : Intel Core Processor (Skylake, IBRS)
stepping : 3
microcode : 0x1
cpu MHz : 3695.998
cache size : 16384 KB
physical id : 0
siblings : 2
core id : 0
cpu cores : 1
apicid : 0
initial apicid : 0
fpu : yes
fpu_exception : yes
cpuid level : 13
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx rdtscp lm constant_tsc rep_good nopl xtopology cpuid tsc_known_freq pni pclmulqdq ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand hypervisor lahf_lm abm cpuid_fault invpcid_single pti ssbd ibrs ibpb fsgsbase bmi1 avx2 smep bmi2 erms invpcid xsaveopt arat
bugs : cpu_meltdown spectre_v1 spectre_v2 spec_store_bypass l1tf mds swapgs itlb_multihit srbds mmio_stale_data retbleed
bogomips : 7391.99
clflush size : 64
cache_alignment : 64
address sizes : 40 bits physical, 48 bits virtual
power management:
---
processor : 1
vendor_id : GenuineIntel
cpu family : 6
model : 94
model name : Intel Core Processor (Skylake, IBRS)
stepping : 3
microcode : 0x1
cpu MHz : 3695.998
cache size : 16384 KB
physical id : 0
siblings : 2
core id : 0
cpu cores : 1
apicid : 1
initial apicid : 1
fpu : yes
fpu_exception : yes
cpuid level : 13
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ht syscall nx rdtscp lm constant_tsc rep_good nopl xtopology cpuid tsc_known_freq pni pclmulqdq ssse3 fma cx16 pcid sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand hypervisor lahf_lm abm cpuid_fault invpcid_single pti ssbd ibrs ibpb fsgsbase bmi1 avx2 smep bmi2 erms invpcid xsaveopt arat
bugs : cpu_meltdown spectre_v1 spectre_v2 spec_store_bypass l1tf mds swapgs itlb_multihit srbds mmio_stale_data retbleed
bogomips : 7391.99
clflush size : 64
cache_alignment : 64
address sizes : 40 bits physical, 48 bits virtual
power management:
The rebuild time with 13 plugins, using s3 object storage and YJIT enabled. So it’s probably can be faster without or less than these or on a new install.
time ./launcher rebuild app
real 5m49.787s
Discourse who’s online plug-in can slow things down.
I found this site to be very useful when debugging
It is able to check the page and view it the same way a user does and attempts to diagnose any issues.
In particular, FCP and LCP are what you probably want to optimize. After isolating my problems to some custom plugins I was able to improve the speed back down to the “good” category for nearly all my pages.
Thanks for all the insightful tips, I will definitely try Vultr out!
How reliable are those metrics with app like Discourse? It has been done for more traditionally web pages. And exacly how we can improve things when there isn’t options such loading PHP later etc?
All we can do is
It was critical in helping me isolate multiple issues on my site. Specifically by comparing the site with or without certain theme components or plugins turned on. It’s also good at isolating large files and identifying any content shifting.
Another option I found helpful was to use the built-in performance testing in Chrome that is accessible in the inspector panel
Same issue for me, I changed from SAS disk (300 IOPs with burst to 1000) to SSD disk (1500 IOPs with burst to 3000) and performance improved ten-fold. (those IOPs are restrictions by the cloud provider I use)
The test looks interesting, but it doesn’t say much about the Discourse benchmark on Vultr.
My best tip is Linode, I’ve speedtested with all of these servers and I’ve found DO the slowest, second place Vultr, and first place Linode (from Akamai)
