5 ms·

Some back of the envelope math for their caching costs: 54 x R3.2xlarge EC2 instances On demand = $314,571.6/year w/ 1 year term = $166,860/year w/ 3 year t

by eunoia 10y ago

Some back of the envelope math for their caching costs: 54 x R3.2xlarge EC2 instances On demand = $314,571.6/year w/ 1 year term = $166,860/year w/ 3 year term = $110,340/year w/ convertible 3 year term = $150,174/year Is that a lot? Seems like a lot

degenerate 10y ago
For a text link site that lets you vote submissions up and down? Yes that's a lot. I'm sure a lot of the decisions at reddit corp were along the lines of, "well, we can't go back and change that now, so let's do whatever we need to fix it and move forward" AKA lots of RAM and caching all over the place.
- mwpmaybe 10y ago
  If a company is grossing $20MM/year and stands to improve that by more than 0.75% by having a faster site, then yes, it makes sense to spend 0.75% of your revenue to do so.
- flashman 10y ago
  One of my favourite breeds of internet comment is "I'm sure five competent engineers could run Reddit at its current scale."
- praneshp 10y ago
  If you think that's all reddit is (text link, with up and down), you have to go use it a little bit. Subreddits, comments, self-hosted images, etc disagree.
user5994461 10y ago
Hard to tell. Don't know how much volume they have. Going from a 50th site in the world in a niche topic (stack overflow) to a 27th worldwide that target all audience. It's entirely possible that it grows the traffic by 100x. Anyway, who cares about money? 100k a year is peanuts. At this scale, if you have a solution that works well, you're good. Not even worth thinking about it.
- notheguyouthink 10y ago
  I mean, doesn't feel like it's working well - fwiw. The number of times i have seen reddit is down pages, ranging from maintenance, to oopsies, to slow loads and failed requests... i just can't count, they seem part of the normal reddit flow. They're better now then they were ages ago, but it still happens. Combine that with slow load times, etcetc. Now, don't get me wrong they're likely doing better than i could - it's a very difficult problem to solve. But, something must be fundamentally wrong if a site who's growth has been steady and not "overnight", it just seems like it's always been very very slow. By overnight i just mean, it's growth has appeared predictable; Unlike the poor sites that get HN/Reddit hug of death. edit: Not sure why i was downvoted, but i'll elaborate: From a user perspective, Google never goes down. It's a big event when that happens. I've never seen Amazon down (again, personally!). I've never seen most of the big players down. Yet i've seen Reddit down more times than i can count. It's a meme on the site. Regardless of how difficult the problem domain is, the end user doesn't care. All i'm saying is that the end user experience is not good enough, in my opinion. I don't think that is an unfair statement. So if this caching is what is powering the current state of things, what can be done better? From a technical perspective, is this the best that can be achieved? I surely hope not. Would you argue it's the best we can hope for?
  - brokenmachine 10y ago
    I don't know why you're getting downvoted either. Anecdotally, I also get way more errors/dropped connections/random slowness on Reddit than on any of the other massive sites. Plus the mobile site is beyond terrible, unusable on my phone. I'm sure it's not an easy problem to solve, but they should be able to afford some hosting with all the reddit gold that seems to get thrown around.
    - user5994461 10y ago
      The other sites make tons of money. Dunno about reddit. Do they even have a revenue model except some ads that 60% of their user block?
mozumder 10y ago
If you build it yourself it would cost about $1.5k-2k per instance.
mason55 10y ago
That's the cost of 0.5-1 FTE. To come up with a better solution it would have to be enough cheaper to reduce the engineering hours of design and build and then take zero extra people to maintain. Feels unlikely.

by eunoia 10y ago

degenerate 10y ago

For a text link site that lets you vote submissions up and down? Yes that's a lot. I'm sure a lot of the decisions at reddit corp were along the lines of, "well, we can't go back and change that now, so let's do whatever we need to fix it and move forward" AKA lots of RAM and caching all over the place.

mwpmaybe 10y ago
If a company is grossing $20MM/year and stands to improve that by more than 0.75% by having a faster site, then yes, it makes sense to spend 0.75% of your revenue to do so.
flashman 10y ago
One of my favourite breeds of internet comment is "I'm sure five competent engineers could run Reddit at its current scale."
praneshp 10y ago
If you think that's all reddit is (text link, with up and down), you have to go use it a little bit. Subreddits, comments, self-hosted images, etc disagree.

user5994461 10y ago

Hard to tell. Don't know how much volume they have. Going from a 50th site in the world in a niche topic (stack overflow) to a 27th worldwide that target all audience. It's entirely possible that it grows the traffic by 100x. Anyway, who cares about money? 100k a year is peanuts. At this scale, if you have a solution that works well, you're good. Not even worth thinking about it.

notheguyouthink 10y ago
I mean, doesn't feel like it's working well - fwiw. The number of times i have seen reddit is down pages, ranging from maintenance, to oopsies, to slow loads and failed requests... i just can't count, they seem part of the normal reddit flow. They're better now then they were ages ago, but it still happens. Combine that with slow load times, etcetc. Now, don't get me wrong they're likely doing better than i could - it's a very difficult problem to solve. But, something must be fundamentally wrong if a site who's growth has been steady and not "overnight", it just seems like it's always been very very slow. By overnight i just mean, it's growth has appeared predictable; Unlike the poor sites that get HN/Reddit hug of death. edit: Not sure why i was downvoted, but i'll elaborate: From a user perspective, Google never goes down. It's a big event when that happens. I've never seen Amazon down (again, personally!). I've never seen most of the big players down. Yet i've seen Reddit down more times than i can count. It's a meme on the site. Regardless of how difficult the problem domain is, the end user doesn't care. All i'm saying is that the end user experience is not good enough, in my opinion. I don't think that is an unfair statement. So if this caching is what is powering the current state of things, what can be done better? From a technical perspective, is this the best that can be achieved? I surely hope not. Would you argue it's the best we can hope for?
- brokenmachine 10y ago
  I don't know why you're getting downvoted either. Anecdotally, I also get way more errors/dropped connections/random slowness on Reddit than on any of the other massive sites. Plus the mobile site is beyond terrible, unusable on my phone. I'm sure it's not an easy problem to solve, but they should be able to afford some hosting with all the reddit gold that seems to get thrown around.
  - user5994461 10y ago
    The other sites make tons of money. Dunno about reddit. Do they even have a revenue model except some ads that 60% of their user block?

mozumder 10y ago

If you build it yourself it would cost about $1.5k-2k per instance.

mason55 10y ago

That's the cost of 0.5-1 FTE. To come up with a better solution it would have to be enough cheaper to reduce the engineering hours of design and build and then take zero extra people to maintain. Feels unlikely.