4 ms·

Ask HN: Why can't Tumblr stay up?

5 points by emilepetrone 16y ago

spooneybarger 16y ago
Are you hoping for a response from a tumblr engineer or just some idle speculation and gossip?
- emilepetrone 16y ago
  A technical response would be valuable to the community. Too often we accept poor performance without questioning what is going on - and learning from it. It's a valid question. Please don't be so cynical. Thanks
  - veb 16y ago
    I see quite often people talk about 'good practices' in web-programming but generally say "don't worry about scale issues... if it becomes an issue, you've just beaten 90% of other startups" umm, isn't this a bad practice? Because when you do start worrying, it's gonna be a bitch to fix...
    - pinksoda 16y ago
      Why would this be a bad practice? This is a GOOD practice. You don't want to add months to the development time before you know if it's going to get traction. I've had several successful sites, but dozens more failures and I'm glad I didn't waste the time making them all easily scalable; they simply didn't need it.
      - veb 16y ago
        I'm just wondering what happens when suddenly you're growing incredibly fast, and you simply do not have the development/money to keep up with the volume of traffic, how does one get on top?
        spooneybarger 16y ago
        that is a better problem to have then shipping too late to matter.
        veb 16y ago
        Fair point.
  - spooneybarger 16y ago
    I'm not being cynical. I find the odds of anyone here having intimate knowledge of tumblr's setup to be really low. Aside from someone with that knowledge, you aren't going to get anything other than speculation and gossip. If you had asked, what are issues that could cause the types of problems that a size w/ tumblr's scope might encounter, that I could understand.
dieselz 16y ago
Anyone looking at their situation is probably saying "with all the resources they have, they should be able to build a system that could stay up." This is true. Its likely that if they could start from scratch, their team could put together a system that would be able to handle the traffic they are seeing and expect to see in the future. The problem is getting the legacy systems to the point where they can scale. It's one thing to have a platform that handles 10s of thousands requests per day. Now convert that system into one that handles 100s of millions of requests daily. Thats a tall order. Developers which are involved with a product early are typically more interested in building a viable product, not getting it to "web-scale," and rightfully so.

5 points by emilepetrone 16y ago

spooneybarger 16y ago

Are you hoping for a response from a tumblr engineer or just some idle speculation and gossip?

emilepetrone 16y ago
A technical response would be valuable to the community. Too often we accept poor performance without questioning what is going on - and learning from it. It's a valid question. Please don't be so cynical. Thanks
- veb 16y ago
  I see quite often people talk about 'good practices' in web-programming but generally say "don't worry about scale issues... if it becomes an issue, you've just beaten 90% of other startups" umm, isn't this a bad practice? Because when you do start worrying, it's gonna be a bitch to fix...
  - pinksoda 16y ago
    Why would this be a bad practice? This is a GOOD practice. You don't want to add months to the development time before you know if it's going to get traction. I've had several successful sites, but dozens more failures and I'm glad I didn't waste the time making them all easily scalable; they simply didn't need it.
    - veb 16y ago
      I'm just wondering what happens when suddenly you're growing incredibly fast, and you simply do not have the development/money to keep up with the volume of traffic, how does one get on top?
      - spooneybarger 16y ago
        that is a better problem to have then shipping too late to matter.
        veb 16y ago
        Fair point.
- spooneybarger 16y ago
  I'm not being cynical. I find the odds of anyone here having intimate knowledge of tumblr's setup to be really low. Aside from someone with that knowledge, you aren't going to get anything other than speculation and gossip. If you had asked, what are issues that could cause the types of problems that a size w/ tumblr's scope might encounter, that I could understand.

dieselz 16y ago

Anyone looking at their situation is probably saying "with all the resources they have, they should be able to build a system that could stay up." This is true. Its likely that if they could start from scratch, their team could put together a system that would be able to handle the traffic they are seeing and expect to see in the future. The problem is getting the legacy systems to the point where they can scale. It's one thing to have a platform that handles 10s of thousands requests per day. Now convert that system into one that handles 100s of millions of requests daily. Thats a tall order. Developers which are involved with a product early are typically more interested in building a viable product, not getting it to "web-scale," and rightfully so.