13 comments

[ 2.4 ms ] story [ 64.5 ms ] thread
>Since we added Solid Cache to Basecamp, the 95 percentile request duration fell from 375ms to 225ms. There’s no magic here, just the effect of a bigger cache!

A few more numbers to understand the load characteristics would be helpful, but that top line number is pretty impressive.

Mainly fragment cache, so cached HTML parts. It wouldn't make sense for caching simple objects (for example caching a query-record-by-id)
>It wouldn't make sense for caching simple objects (for example caching a query-record-by-id)

Yes, caching database queries in a database would be a groundbreaking technology

Really love how the 37signals guys are turning conventional wisdom on its head - moving from cloud back to own data center, getting rid of SPAs with Hotwire, and now this. More power to DHH and co.
And just to add - they've never taken VC funding; and they've been remote from the get-go before it was a thing. Truly a unique organization.

Note - I don't work there or use any of their products (other than Ruby on Rails). These are just my honest observations.

I agree.

I understand DHH pisses off many people when he speaks, and some times he could probably be a bit less harsh, but I admire him a lot because he's one of the few people I know in this industry critically thinking, applying common sense, and not blindly following fads and all the bullshit that's going on at most places.

I think overall it is a net gain to follow and listen what he says, even if not said in the most "nice" way, I agree on almost everything he says tech wise.

And using Rails after years of pure "modern" SPAs, feels really refreshing. I'm sad most of us have forgotten how simpler things could be, and terrified about new people joining the industry that will never know it (for the time being at least).

The software business constantly rediscovers old technologies and techniques, renames them, and calls it innovation.
Are the response times from a client point of view? If they’re internal, is 225-375 standard for this kind of stack?
wonder what would happen if they'd add a bigger Redis cache
Well I doubt you could beat Redis for pure performance but bigger Redis cache would likely cost more to run.

So the real question is about cost/performance and picking the one that best fits your use case.

I also wonder how tweaking the backing DB's in memory cache will effect things.

Either way it's great to get this stuff out of the box with Rails and then be able to swap for Redis later if you observe that being a better fit.