Designing systems to scale


Comments are closed.

Good overview about his experiences/opinions on development best practices. Not focused enough on scaling, though IMHO. I do not agree on the ORM is evil part. Talk would have benefitted on Michael using a microphone. Next time add more real world stories ;-)

I really liked your talk dude, to the point. Could of used a mic. as it was a tad quiet.

Was a really good talk, like others said, could of used with a mic though!

As mentioned above I felt the presentation was pretty good but it could have used some further information about actually scaling applications to meet the demand of users. The coverage around making applications module and redundant was a good reminder for those who don't already do this though :)

I gained a massive amount from this speaker. I have used the... we just need to get it working route before and have faced many issues.
I like the... build many small services and applications that combine to create a greater whole ...approach.
It has altered my thinking and approach to application design and will hopefully make me a better developer. Thank You.

Anonymous at 16:06 on 9 Oct 2012

The first half was quite good and I thought he would elaborate on scalability issues etc but seemed to tail off into talking about his experiences using various third party components. The ORM statement was a little over-dramatic - people use ORMS because they get sick of having to code in their own mapping and data access layers. If you don't like the full blown stuff though you can always look into using a micro-ORM - they're not all evil!

It's unfortunate that you commented anonymously as I'd love to chat and see what you would like to hear about RE: scalability issues.

The ORM statement *was* over dramatic. I thought long and hard about including it, but it's such a polarising statement (and in my experience, it's true!) that I decided to keep it in. Either it made people think about how they're using them, or it validated those that have been burned in a similar way.

I fully agree that ORM's are good in regards to not having to write your own mapping and data access layers, but that's not what the talk was about. When it comes to scaling your solution they're generally your biggest issue (or so I've found). For high volume systems, the performance/developer time tradeoff isn't worth it. I used an ORM for rapid development of our application, but it was the first thing that we pulled out when trying to increase performance.

If you've developed a system that uses an ORM successfully to process millions of requests an hour (without caching) I'd love to hear about it, as I'm open to changing my mind about them. For now though, I stand by my statement that they're evil.

First up, the good bits...
* The slides looked great - decent font with the right amount of content, it's surprising how many speakers struggle with this.
* Good advice on automating tasks, especially the process of gradually moving to auto for scary tasks.
* Was good to see you weren't afraid to introduce many different techs into one stack - too many people are fixated on PHP.
* Generally interesting and well presented talk.

The not so good bits...
* As mentioned in previous comments, you definitely needed a mic, although I'm not sure if this was a technical issue or a choice you made due to the size of the room? Initially we could hear clearly but your voice tailed off once you weren't concentrating on being heard.
* "Things I've learned" must've popped up about a dozen times. I realize this was probably an intentional style choice but it felt unnecessary. This could just be me being picky though :)
* Need to work on the timing a bit as you had to skip the logging part of your talk.

BTW, there's a typo on slide 27 "if you all have is a hammer..." should be "if all you have is a hammer".

This talk for me was one of the best of the weekend. I have taken a lot away from this talk as I am nearly at the crash and burn stage of my XboxAPI project. Your talk really helped and has gave me a fantastic understanding of where I was going wrong.

Since your fantastic talk I have purchased the 'High Performance MySQL' book and have made plans from your advice to carve up XboxAPI into smaller tasks so that if one thing goes down the whole site does not die.

Personally I cannot wait for the video's to come out so that I can watch them again and show them to friends and help them learn from your mistakes and experience.

A good talk with nice slides and no information overload. Your talk was more of an overview of things you learnt from personal experience than a technical talk about scalability.

Having said that, you got me thinking about application separation and scalability so I would say that was a job well done! I felt you didn't really cover enough about why you would scale and how this model would help you to achieve this.

The ORM comment was controversial, but true depending on the context of the data being worked with. Lots of 'Things I've learned' sections, that seemed unnecessary but it did work!

Its a shame we missed out on the logging section, as I had a question about log aggregation across multiple applications e.g would you have a separate logging application to deal with this?

You probably should have had a mic, not really sure why you didn't!

Well done, I would also be absolutely terrified!

Great slides, lots of stories well told that taught a lot. I'll be watching it again on video, as I'm sure I missed a lot

Nice talk, good speaker. I love the "we screwed up, let's start over" attitude :)