Few days ago we were chatting in our corporate Campfire room and one of the guys asked me what do I think about our hiring process for Rails developers, what questions I’d ask a candidate if I was interviewing and so on. Those questions sparkled a really long and interesting discussion and I would like to share my thoughts on the topic in this post.
Few months ago I’ve switched one of our internal projects from doing synchronous database saves of analytics data to an asynchronous processing using starling + a pool of workers. This was the day when I really understood the power of specialized queue servers. I was using database (mostly, MySQL) for this kind of tasks for years and sometimes (especially under a highly concurrent load) it worked not so fast… Few times I worked with some queue servers, but those were either some small tasks or I didn’t have a time to really get the idea, that specialized queue servers were created just to do these tasks quickly and efficiently.
All this time (few months now) I was using starling noticed really bad thing in how it works: if workers die (really die, or lock on something for a long time, or just start lagging) and queue start growing, the thing could kill your server and you won’t be able to do something about it – it just eats all your memory and this is it. Since then I’ve started looking for a better solution for our queuing, the technology was too cool to give up. I’ve tried 5 or 6 different popular solutions and all of them sucked… They ALL had the same problem – if your queue grows, this is your problem and not queue broker’s :-/ The last solution I’ve tested was ActiveMQ and either I wasn’t able to push it to its limits or it is really so cool, but looks like it does not have this memory problem. So, we’ve started using it recently.
In this small post I’d like to describe a few things that took me pretty long to figure out in ruby Stomp client: how to make queues persistent (really!) and how to process elements one by one with clients’ acknowledgments.
Read the rest of this entry →
Since the day one when I joined Scribd, I was thinking about the fact that 90+% of our traffic is going to the document view pages, which is a single action in our documents controller. I was wondering how could we improve this action responsiveness and make our users happier.
Few times I was creating a git branches and hacking this action trying to implement some sort of page-level caching to make things faster. But all the time results weren’t as good as I’d like them to be. So, branches were sitting there and waiting for a better idea.
Read the rest of this entry →
Today I was doing some work on one of our database servers (each of them has 4 SAS disks in RAID10 on an Adaptec controller) and it required huge multi-thread I/O-bound read load. Basically it was a set of parallel full-scan reads from a 300Gb compressed innodb table (yes, we use innodb plugin). Looking at the iostat I saw pretty expected results: 90-100% disk utilization and lots of read operations per second. Then I decided to play around with linux I/O schedulers and try to increase disk subsystem throughput. Here are the results:
How often do you think about the reasons why your favorite RDBMS sucks? Last few months I was doing this quite often and yes, my favorite RDBMS is MySQL. The reason why I was thinking so because one of my recent tasks at Scribd was fixing scalability problems in documents browsing.
The problem with browsing was pretty simple to describe and as hard to fix – we have large data set which consists of a few tables with many fields with really bad selectivity (flag fields like is_deleted, is_private, etc; file_type, language_id , category_id and others). As the result of this situation it becomes really hard (if possible at all) to display documents lists like “most popular 1-10 pages PDF documents in Italian language from the category “Business” (of course, non-deleted, non-private, etc). If you’ll try to create appropriate indexes for each possible filters combination, you’ll end up having tens or hundreds of indexes and every INSERT query in your tables will take ages.