1.2.1
We are happy to announce that we are close to ending performance optimization. When we launched Headspin platform, we did it without many performance optimization and as thought that could be really big problem when we get tons of users and parts of site could be broken due aggressive optimization.
We are listening our users, and we made clear that we’ll take serious on this.
After launch we have introduced facet caching which improved filtering results, but we were not fully satisfied with that alone. Our measurements showed that page generation time is not that great but facets are now cached and filtering performance was much better.
We have tested few solutions, and we have moved site to another provider which gave us dedicated threads (vCPU) with around 20% single core performance boost based on benchmarks we could find. In our tests dedicated resources and faster single core processor performance shaved page generation time by half based on our internal tests. On eye tests it flies.
The last piece of puzzle is cache optimization. We took some effort to see what is actually our site doing and what would be ideal approach.
While we have combination of service and plugins, strictly speaking pattern of our site is membership-like site. We are having different user categories, and activating just private cache would generate tons of cache, and would mean less cache hits.
We have decided to custom code own membership solution which we are later using for caching logic. We have divided users in caching groups, and when you are logged in you should share on some pages cache with all users within you category. This way instead copy per user we can have one copy which could include 200 users which should have same cache, pages should get more hits, TTL could be higher and feel could be much faster.
We had report that for one user it was taking 12s to load our wireframes page. We were in contact with user to see why this happens, and that we are not happy with speed, but it takes 5.5s at the time for use.
There was our error that user having free and pro tier would see both, and that was due our error absolutely the worst case scenario.
Right now we thing that for the most of users performance should be almost like static page. Loading Wireframes should take less than 1s, filtering should take less than 1s when we are talking about cached scenario,
In cache miss scenario that should be around 2s for each case. For some location TTFB could be worse, but for almost every user it should load under 3s.
If you face any issues, or your experience deviates from mentioned in this article, please let us know, and we will do best we can to fix it. Because of all settings keep in mind there could be some issues, report them to use and we’ll work to make our platform better place.
You can reach us at:
support@headspinui.com
