New algorithm for managing resources on shared plans

Any news about this?

We´ve been seen a more inconsistent performance lately.

Is this new algorithm working already? Or do we have to do something to update to get this improvements?

Can you provide any update, please?

Couple notes to avoid confusion:

-The new algorithm to prevent any one app from using up too much of the shared cluster’s capacity has been live for about a month now.

-What isn’t live yet is the ability to control how much capacity each app has and pay for more for apps that need more… we’re still wrapping up some technical work on this.

-This algorithm works on an app-by-app basis: each app has a total amount of capacity. If you have more than one app, they are independent of each other. (But the development + live versions of the same app do share the same capacity).

-If you’re wondering whether or not the algorithm is limiting your app, you can look at the “Logs” tab in the editor… the “Periods where the app hit its maximum capacity (%)” graph shows times where your app was up against the limits of its amount. If that graph is at zero, then the new algorithm is not having any effect on your app. (Right now we’re only showing this graph for apps on paid plans).

-Your app is also affected by overall system performance and health, which you can monitor here: http://status.bubble.is/. We’ve seen a huge improvement to our overall uptime and response time since we rolled out the new algorithm, but we occasionally still have issues where performance is affected – we’re working on squeezing them out one-by-one.

-All of the above only affects server performance: how fast we can send data and process workflows. That matters a lot, but the performance that your end-user sees depends on other things as well, such as: how big and complicated a page is, how much data the page displays (both bubble data and things like images), how much work the javascript in the user’s web browser does, how efficient the load-ordering of the page is (can we load everything at once or do some things depend on other things?), how fast the user’s internet connection is, how long and complicated workflows are, etc. So if you’re seeing slowness that’s not explained by either your Logs tab or the https://status.bubble.is page, it might be due to one of those issues. We’re constantly working to make Bubble more efficient, and you can improve things on your end by keeping your pages small and minimizing the amount of data that gets manipulated by your searches and workflows.

8 Likes

Thanks for the explanation @josh. Much appreciated.

I reached out to the Bubble team to discuss our options for dedicated. Is anyone here on a dedicated plan with a server in Frankfurt? I’d like to get in touch to see how the performance is.

Currently, more often than not around this time of day (when the US 'wakes up) the server load increases heavily. It is not uncommon to see a number between 10 and 20 in the server usage max capacity. We have offices in West and Eastern Europe and especially Eastern Europe has a lot of issues with the performance of the app. (their internet connection is fast and reliable).

1 Like

I’m really sorry to having to keep bringing this up and I would like to thank @emmanuel for taking the time to log into our app and test the performance. Unfortunately I still feel these algorithm changes have very badly impacted our app’s performance in a way that is far beyond normal acceptable use.

Currently we are just working in the app with 5/6 users and server capacity is sometimes in red for 45 minutes out of 60. We are opening this app (less data intensive) up for the public soon, but I have a feeling it will totally crash.

Please have a look at this. My team in Eastern Europe keep sending me these kinds of screenshots:

Then when I look at the server usage, I see this:

They are doing very simple actions, changing a field on a Thing. Retrieving a Thing’s Thing’s fields. Not even starting API workflows.

I really feel we have been badly capped as before these algorithm changes this was nowhere near as bad. Yes we have some more users and we have more workflows but we talk about 1/2 users and +coupe thousand workflows.

I have people sitting in an office waiting for minutes (literally) for data to be displayed on screen. And yes I changed a lot of the data sources from parent group’s Thing’s Things to ‘Do a search for’ and it helps, but not enough. Even worse is that they start to panick because they change a field and the other person is not seeing it because of server delay, then they keep messing up because they think the changes have not been processed.

We want to add more external users to our backend of the app, but I really cannot do this now as they will have a terrible user experience. And it looks things get worse whenever a new user is added.

I really need a solution! We cannot afford to go dedicated yet ($635 a month is too much at this point) but on the other hand, this is just not working either.

1 Like