A fair and honest chat about Performance

@feee Hopefully Bubble team will come up with a solution to keep you as a customer. I am concerned however that for someone paying $2500 a month you are not getting premium, ‘pick up the phone’ support! I have an APP hosted on AWS and pay half that a month, including premium support and have phone support with AWS team and response to enquiries within an hour or two at the most.

I have just launched a beta enterprise app on Bubble and picked up the first of two enterprise clients and I was hoping I would not have to go back down the traditional development route - please no!!!
Although I am well prepared to do so if I have to. Now we have people using the app and trying to make changes I can see some limitations which will require some repetitive data restructuring to keep the speed at an acceptable level before we start upgrading to dedicated. But it looks like we might end up going down the rebuild route once we have proven the idea and of course have the revenue coming in.

3 Likes

@feee it would help us a lot to have a look at your app if its public and you can share. I’ve always been very curious to know how a high traffic Bubble app looks like and what are the real limits.

2 Likes

Hey @vlad thanks for the tip about 3. Empty search expressions.

This has helped speed up the reporting area a second or two.

Cheers!

1 Like

@JonL You posted the same request once already. I think @feee was plenty generous sharing his experience and performance metrics. There could be implications with disclosing his website in this context on a forum that is publicly indexed, especially for a site as large as his. I would like people who feel inclined to contribute to the discussion to be able to do so without feeling pressured to show off their apps. I think that is apart of keeping the discussion fair.

10 Likes

He could have easily said so.

I am just kind of suspicious due to his forum profile. Not very active and he states that he has done everything in his hand to fight performance issues.

Not that it means anything, but the best Bubble websites in terms of performance are owned by active people in this forum. I’m assuming that is because they put the time and effort to learn in this forum all the ins and outs of Bubble.

Everyday I learn something new here that makes my site faster. Maybe he has another account.

I don’t know…it just was suspicious to me that someone inactive in the forum comes out of nowhere bashing Bubble performance and support. We know performance is not in the best shape but I also know it’s not that bad if you understand how Bubble works and build with that in mind.

8 Likes

I appreciate your enthusiastic advocacy for Bubble, but let’s assume people on the forum are operating in good faith. This is an extremely positive community, and while many people participate, others lurk – everyone is welcome and I appreciate when lurkers chime in!

I can confirm @feee is a Bubble customer who is running a high traffic website on a dedicated plan.

I wasn’t aware they were looking to move off of Bubble, and I’m sorry to hear it. As @feee mentions, I’ve helped them with performance issues in the past, but haven’t had as much time to investigate individual Bubble apps lately. Improving performance generally involves understanding how Bubble works end-to-end, and we’re short on people who can do that, which is why my priority right now is hiring and training.

Without spending time on it, I can’t say whether @feee is up against fundamental limitations of the system, or if there are further optimizations we can do. @feee, I don’t know how deep you’ve gone down the re-building off Bubble route: if you’re still willing to consider staying, message me and we can commit to spending a little bit of time late next week evaluating if it’s possible (no promises) to improve your app’s performance in the near-term to make it good enough for your needs. (Probably by taking one particular feature of your app and using that as a proof-of-concept to see if we can get decent performance for it, and probably by giving you suggestions on how to change the app design rather than by changing Bubble’s code, but we’d see).

To address the concerns about support and capacity that several people on the thread are mentioning:

We prioritize support tickets from customers on team, production, and dedicated plans, but we’re still handling them with the same small team that handles all our incoming requests, so while the level of support we provide is better, it isn’t radically different. I would like us to be able to sell premium support the way larger players like AWS does, where we can commit to strict SLAs and provide 24 / 7 coverage, but we’re still working on building out a team that can handle that. Currently, most of the money for large dedicated plans goes to hardware, overhead, and backend costs: we don’t have phone support priced in. (I think our pricing is extremely competitive compared to hiring out the team that would be necessary to achieve similar results through code). I know that for a lot of our users, premium support would make Bubble a lot more appealing, so we definitely want to offer it, but I also think it’s important for support to be able to actually resolve problems, and that means building the team from engineering forward rather than just getting more warm bodies, which is a slow thing to build out.

In terms of concerns about whether buying capacity has an effect on performance, the thing to remember is that capacity doesn’t buy speed, it buys scale. Think of checking out at a supermarket: if the supermarket adds more registers, the time it takes for the clerk to ring you up doesn’t change: they still have to scan all your items and add them to the bag. But, the supermarket as a whole can handle more customers at the same time. You might see a performance increase adding more capacity because if there was a long line for a single register, adding more capacity can shorten the line. But the max speed is still going to be the time it takes for a single register to check-out a single customer.

Re:

Great, I’ve added it to our list – I do need to do some more investigation to confirm if it’s feasible, but I suspect it is, and that it’s a relatively quick change rather than a long one.

24 Likes

If you’re paying around $2,500/month for a service that’s expected to speed up or can handle more data at once in your app then that tells me there’s a major flaw with Bubble.

I know by saying this, I may start something

It makes me wonder, does Bubble actually give you a Dedicated Server with the same features that you requested or are they claiming they are and not actually providing it?

Remember that I’m running millions of workflows…
They do provide dedicated server, when there is the occasional bubble outage I’m not affected. Also, you get your private IP when transitioning to dedicated.
My issue seems to be all about data queries, I have 500gb of data and repeating groups that are loading a lot of that data. (also, some of my data may not be organized the best way possible)

I’m actually super excited right now because it may just be that Search 2.0 (Algolia) could solve the majority of my issues. I’m referring to this post on the forum: Search 2.0 - Summarizing what I've learned exploring Algolia

I’m building this in as a test right now, will let you all know how it goes.

17 Likes

Hey!

Thanks for sharing your Bubble scaling issues and thoughts. I’m not asking you to reveal your exact website or anything that would give it away but could you give more insight as to what the repeating groups are displaying that takes so long?

How many results are usually returned in repeating groups? Are you doing data searches, how complex are they?

What is your average load times from start to fully loaded?

Thank you for sharing, interested in hearing.

3 Likes

Do you think social media apps would fall under CRUD style apps? Could one possibly create a new social media platform on Bubble, ie would the current performance and stability of Bubble allow for a Twitter like app (since one of Bubble’s famous example is a Twitter like app)?

From my experience, you’ll have to switch over from Bubble when your social app grows and gets users. It would be cheaper per month, however, you’ll need to pay for a developer or try to do it yourself. Bubble will get slower overtime because there would be more data for it to handle and if there are many users on your app at once, they would have a bad experience.

As someone who has been on Bubble for quite a while I feel I must point out a couple of things. How many of us would even be thinking about our own app or app needs if it wasn’t for Bubble? To even dream of beginning to make an app like Twitter or Facebook a few years ago would have been disregarded as needing a bunch of capital and a one way ticket to the Valley. With Bubble an experienced user can have it running in a month or less from anywhere, even a one stoplight town in Ohio;)

As for performance I can only say that from a strategic perspective @josh and @emmanuel have to know that consistent performance improvement is vital to their stated vision. I have no doubt that they are working towards those improvements daily. I also have no doubt that Bubble will only get faster and more powerful with time. If you have time then build and improve your app alongside of their progress. Don’t let Bubble do the job of catching up to your app. Keep steamlining and at the end of the day you will have a tightly built, well designed app that will only get faster with bubbles improvements.

So Josh and Emmanuel, keep improving this platform and I will keep improving my app. Our success will go hand in hand.

One last suggestion would be to provide a performance improvement letter every month that outlines what improvements have been made. Don’t worry whether it is overly technical. Interested people will take it as a cue to learn and the community will be reassured that improvements continue to be made.

18 Likes

My Grandmother always said, "Don’t try to cross a bridge until you come to it."

I’m relatively new to bubble but have just over 30 years in tech, and my business successes meant I was approached to mentor/coach over 1,000 tech entreprners. The number of businesses whose app I have seen hit the @feee volume you could count on your fingers.

This volume performance criterion ought to have a weighting that makes it less important (right now) in your decision making than say time to market, flexibility to make tweaks rapidly and inexpensively to reach product-market fit, and availability/cost/time to recruit and train up skilled resources.

My thoughts on this matter are to get the MVP live fast then keep tweaking and changing to get to product-market fit, then attempt to scale by getting more users/customers (this last step usually takes way more time than anyone thinks). And, by the time you’ve got there, bubble may well be at a performance level that is more than acceptable. If not, you will have revenue to hire skilled developers to build a code base … but expect that to take many times longer and cost many times more than your bubble app.

24 Likes

Bubble is the tech part of my project. I count on them to be at their best when needed. I’m confident that the speed issue will be fixed in the near future.

As said on a previous post, Amazon has hit 15,000 requests per second to access their API, and they have also a E2 Unlimited that give you all CPU power in peek time. No more time for isolated servers and slow responses. Are we ready to pay extra money to have performance? Everything is about money, right?

2 Likes

agree x100 :slight_smile:

1 Like

I have to agree with other comments about Bubble’s ability to enable us to create apps that would otherwise be out of our comfort zone or time/cost prohibitive. It’s amazing what we’re able to build and the products/companies that are forming because of Bubble. Bubble will continue to grow as we grow - it’s a symbiotic relationship. They are hiring additional engineers/developers to help and performance improvements are stated at the top of their roadmap, so they are aware and working to improve it.

With any new technology there is going to be obstacles and it’s how the platform team responds, and corrects them, that matters. In those terms, the Bubble team is amazing! They try to be as open as possible with us about limitations and future enhancements. They also take our feedback to heart and implement features that we need, almost on the fly.

While I do think there are areas for improvement, they are happening, we just have to continue to dream and build our products and I’m sure, in time, Bubble will improve.

15 Likes

Just out of curiosity.

Are people aware about the relation between power management and browser performance?

A lot of end-users will often complain about performance without actually knowing that the problem is on their side.

I had a good share of desktop sharing session with users that complain about it just to realise that the CPU was throttling due to a “Power saving” mode on their computers. Because sometimes performance is just about perception or putting it another way…it’s just about 500 ms. So a page fully loading in 4 seconds is slow but loading in 3.5 is OK.

Here you have an example with bubble.is(not logged)

Power save mode:

image

Balanced mode:

image

Performance mode:

image

There is a big difference between Power Save and Balanced/Performance but maybe not enough for someone to say…“Wow! performance is horrible.”

If I do the same on @feee website (I finally found out) I get the following. I took as reference to stop the watch when all the elements were displayed in the browser. Specifically the RGs.

Power save mode:

image

Balanced mode:

image

Performance mode:

image

There is a difference of about 15 seconds between the power save mode and the balanced/performance mode.

So while at bubble.is you have the same 2x time you only go from 3,5 seconds to 7.5 seconds to fully load all elements.

But on @feee website you go from 15 to 30 seconds and while 15 seconds may seem too much I can understand given the amount of elements it’s loading and I wouldn’t mind waiting. But waiting half a minute is just not acceptable.

To summarise. Make sure that when you test and process feedback about performance that the bottleneck is not the CPU.

On the other side @feee I would add a nice animation saying that the page is loading and preparing things. It would make the 15 secs wait less frustrating.

8 Likes

Hi @JonL, how good are those test results?

Performance testing normally involves two levels, with and without browser cached; and then averaged over a number of runs. This would help avoid assumptions based on network speed variations at that particular moment.

It certainly is interesting to see the power save mode slowing things down. I’d expect similar effects across different devices, for example my phone is pretty slow at everything.

Both tests were performed on the same session of a Chrome Incognito window after caching all elements and reloading 3 times the website. Times were taken on the 3rd run.

I didn’t have time to take averages from a big sample of runs but indeed that would be ideal.

I just wanted to attach some evidence to my statement without going in depth.

Another way to simulate this in a more controlled way is to use the Chrome Dev Tools CPU throttling option in the performance tab. You can also throttle the network to control also this variable.

2 Likes

Hi, thank you for the great answer, but what specifically are these “advanced filters” I see people mentioning around?