Arcade.gs using SQL.js http virtual filesystem (Static SQLite)
08:21 Share

Okay, quick follow up to the sequel, JS HTTP virtual file system post. I both the new website to test out a different data set. This is a 10 MB data set of arcade games from archived.org, their Internet arcave project their spread across flash games. In my as titles and name arcade games, this is about 11 or 12,000 records in a sequelightte database. Most things are indexed and I built a view next friend in deposit now, if I test things out so performance, why is the CDN performance just about the best I've ever seen him it's a little faster than anything else, I've deployed to either city and I've used. I have to look into a little bit. More just gave me a good idea of kind of where the 60s and where this fails, so the website is pretty blazing fast, like I have infinite pagination going on for each category, and I am it loads 30 records at a time in normal paginated. Sql queries like limit offset and it's pretty instantaneous like. Basically, you can't see the light indicator at all and that's on it pretty fast Internet connection, but like each of the subsequent request for additional pages are so fast that I don't like 3040. Ms, I keep you. Can you can't perceive it really so you end up with the system. It'S really really really fast. However, it doesn't up with a pretty hefty coldstart problem. It'S web assembly see you have to load the web assembly module before you make your first request that takes about an hour, let's see anywhere from 900, ms to 1 1/2 seconds from the tasks I've done. It'S a 1.3 MB by Ineri Wassom file that contains basically sequel, eight recycled address as a project name for the driver. Anyway, that's about a second just to say you can hide that basically behind you can use as a future to use a high that I am but I'll just start with that also the requests required to perform. My first query on the status are in the Bing about a 1.2 MB presentational 1.2 MB of that 10 MB database and an additional requests end up being just blazing fast, like somebody's come in at like five or 10. Others must be cash. Some of these come on. Let'S see yeah these ones have ask HR they end up. I will just not being really really fast. I guess I'm not 4050, ms 70. Ms, for that, one wrong impression about 30 to call bytes of data. So what is our website? Look like well basically, when users light on the homepage. I can come hide that initial coldstart problem, since it happens, asynchronously the webpage Lowe's in under a second. If it's just a normal knock, stop serving basically start a countdown for the homepage and then, when the page loads. It kicks off that download in the load to initialize the Wasem file and takes a second and then IT by making a query from the homepage. To get that big one megabyte query out of the way, and so basically as long as it takes the user a couple of seconds to click to the category they want to browse, it just seems very, very quick and fast. However, it's probably not possible to optimize too much further. Like you, I think, if you're going to use this approach, do you have to be looking at a larger data set, probably larger than 10 MB? That is index in a way that you're gon na have a lot of subsequent queries, but you can come home hide that cold start problem was actually thinking about it on lake a website like Netflix or the HBO max app. Since I worked on that before I was kind of thinking about that contacts like conceivably, you could take the entire public database of titles that were available to the user and you could stick all the public meditator. Just like you know, URLs to images and descriptions. Titles things like that you can stick all of that in a database if you have, if you had the coldstart problem on the student up being super fast, I guess one trade-off is this. Data ends up getting cash, so your theoretical maximum usage of like memory in your JavaScript deep, is basically the entire size of the database. So that's something to keep in mind, especially for lake Ella webOS. I don't think webOS could handle. Maybe FiOS FiOS, some android TV could, I bet Apple TV definitely could but like why Apple TV doesn't have a technologies is matter. You have to reinvent everything from scratch. It'S probably not worth it. Blake say you implement in webOS for LG TV is. That will probably end up being yeah. I don't think you want your your data set in the consuming 5060 MB of memory in that platform in my name, become an issue in lower and android devices, especially Amazon devices. But I don't have an interest in technique, especially if you use her desktop like I'm. Just imagine you can deploy website directly, can get home to a static. Cnn Kinda, not really think about the problem of a database server while still and I'm getting the unlimited ability to carry a SQL database and read only mode. That'S pretty neat and I'm sure, there's used cases for it, especially data science, are exploring data set alarm.