IMHO the slowdown wasn't because of total number display (is that joke?) but because of addons used for infinite scrolling or custom filtering and users who use them (oh you want to filter some GAs by level so how about to download everything with client js addon just to show 2 GAs after, instead this must be done by sg on server-side sql with conditions, you still select data to display anyways aren't you?).
The paging bar changed so that broke some addons or something else.
Just after addon devs fix some shit you'll start to experiencing slowdowns again.
I guess there's no big deal for them to download all results now just to display the counter kek.
Comment has been collapsed.
Have yourself a merry little bump with this. Good work.
Comment has been collapsed.
Nice job cg. Awesome to see continued improvements.
Comment has been collapsed.
13 Comments - Last post 12 minutes ago by devotee
51 Comments - Last post 27 minutes ago by shak59
6 Comments - Last post 32 minutes ago by shak59
16 Comments - Last post 1 hour ago by trentjaspar
24 Comments - Last post 1 hour ago by Maclord40k
1 Comments - Last post 1 hour ago by pb1
3,843 Comments - Last post 1 hour ago by fr0zenX
2,193 Comments - Last post 54 seconds ago by VernoWhitney
85 Comments - Last post 5 minutes ago by MarshallBanana0815
31 Comments - Last post 18 minutes ago by LieEater
363 Comments - Last post 31 minutes ago by schmoan
757 Comments - Last post 31 minutes ago by LieEater
217 Comments - Last post 58 minutes ago by HustlaOG
10,584 Comments - Last post 1 hour ago by HustlaOG
Hi SG, as you might have noticed, we started running into slow load times last week during peak hours. We moved to a new host recently (AWS), and I setup our database on an instance with 122 GiB of RAM. This allowed our dataset and indexes to fit in memory, which solved a number of issues we experienced in the past couple of months as our database grew in size. Nonetheless, we've been hitting some limitations on the CPU side of things the past week, as seen here. The graph shows four days of CPU load on our database instance, and the waves represent the rise and fall of traffic during peak and off-peak hours. Once the instance reaches 100% load, we obviously see a sudden drop in performance as tasks begin to queue up.
I spent most of the past week better analyzing our traffic, and rolling out a wide variety of improvements. The one change you'll likely notice and dislike is the homepage no longer showing the total number of results. However, this is a needed change and one that's going to help us keep the load on the instance down. Other changes include adjustments to how the homepage results and featured giveaway are selected, and how winners and giveaway feedback are stored and retrieved. At the end of the day, you should notice the site being more responsive. The archive page will load twice as fast, the homepage will load 40% faster, user profiles, messages, and trades searches 30% faster, and individual discussions will load 15% faster. Other pages received some improvements as well, but the above are the ones you'll likely benefit from the most.
Overall, the performance is better than expected, and you can see the results here. As mentioned earlier, the blue line represents CPU load on the database during four days last week, and for comparison, the green line represents CPU load yesterday.
Comment has been collapsed.