ביצועים על DigitalOcean Premium droplets חדשים

סיכום מבוסס בינה מלאכותית

Users are discussing the performance of DigitalOcean’s new Premium Droplets, which feature NVMe SSDs and faster CPUs, particularly for Discourse hosting. marianord asks if anyone has tested them, noting they can’t resize their current instance yet but are eager to if gains are significant. sam suggests RAM speed and CPU clock speed will likely yield noticeable improvements, and marianord agrees, wondering whether faster RAM or higher clock speed benefits Discourse more. They consider stress testing tools like K6 or JMeter, but sam recommends using discourse bench for accurate results, noting it’s tricky to set up and suggesting a Dockerfile could help. Falco shares real-world rebuild timing results across four droplet types: Basic ($5) took 5m18s, Basic Intel Premium ($6) took 4m19s, Basic AMD Premium ($6) took 3m45s, and CPU Optimized ($40) took 3m33s. The data suggests that upgrading from a $5 Basic droplet to a $6 AMD Premium droplet offers a substantial performance boost for a small cost increase, making it a compelling option for Discourse users seeking better performance without a large budget jump.

Hello everyone,

I’m curious to know if someone has already tested the new premium droplets of DigitalOcean that offer both NVMe SSD and quicker CPUs. As I asume mostly have their Discourse deployed on Droplets (as do I) probably someone already did the resizing.

I am not able to resize my instance and will only be capable to test in the weekend, but if the gains are big enough I might change the region just to do it. :sweat_smile:

More details: Introducing Premium Droplets with faster Intel and AMD CPUs and NVMe SSD | DigitalOcean

Reading the specs it is likely there will be a noticeable difference

Ram speed makes a reasonable difference, top clock speed as well

I was thinking exactly that. Discourse should benefit a lot from these. Now to see what is better, slightly lower clock speed for faster RAM or viceversa.

Other than rebuilding, what could be a good way to test it? I’m thinking on some stress testing with something like K6 or JMeter

The ideal test is probably discourse bench:

Setting it up is a bit tricky, I should at some point make a Dockerfile for it.

One easy way to test the all around hosting performance @codinghorror uses is timing rebuilds. I went ahead and did a 4 way test in Digital Ocean:

Droplet Type Price Time
Basic $5 5m18.476s
Basic Intel Premium $6 4m19.710s
Basic AMD Premium $6 3m45.489s
CPU Optimized $40 3m33.580s

So, it looks like, for people on the $5 droplet, moving to a $6 AMD one is a pretty good deal.