Hacker News new | past | comments | ask | show | jobs | submit login

http://www.crunchbase.com/company/groupon

Funding Total $1.14B

    Series D, 1/11            $950M
    Digital Sky Technologies
    Morgan Stanley Venture Partners
    Fidelity Ventures
    Andreessen Horowitz
    Battery Ventures
    Greylock Partners
    Kleiner Perkins Caufield & Byers
    Maverick Capital
    Silver Lake Partners
    Technology Crossover Ventures
The only thing more ridiculous is Twitter.

Funding Total $1.16B

How much hardware would it take to clone Twitter, if done efficiently?

A year ago they were at 1,000 tweets per second and 12,000 queries per second. http://engineering.twitter.com/2010/10/twitters-new-search-a...

The Steve Jobs news set a new record of 10,000 TPS. TPS Report: http://techland.time.com/2011/10/07/twitter-breaks-tweets-pe...

As of 7 months ago, they were at a billion tweets per week: http://blog.twitter.com/2011/03/numbers.html

If the average message is half the maximum length, then that's around 70 gigs/week naïvely (not counting multibyte characters or metadata).

Around 3.6 terabytes per year. Older tweets are probably almost never accessed so they can stay on mechanical HDs, and stuff that IS needed can get temporarily promoted to SSD storage which can handle all that random seeking. And older tweets can probably be nicely compressed.

Search is the only real challenge, but they "solve" that by not indexing most tweets. A subset of recent tweets, not indexed in real time, pushing out older ones.




I've never written a "twitter clone" but I would suspect the real problem isn't pure throughput but rather the mixture of having to n-way route messages to the proper recipients combined with having to spread the load of that across potentially many systems. I do not think doing that is as trivial as it may seem.

However, I'm sure you're right in that someone could reimplement what twitter does more efficiently than they do it if they paid proper attention to all the mistakes twitter made along the way. Not sure it would be worth the bother though, because twitter's real win is having their company name define the action in much the way googling has come to define web search. You can't overcome that by being marginally better.. or, as it happens, even by being marginally better and giving away tons of free money (just ask MS/Bing).


Apparently up to 4B tweets monthly now: http://blog.twitter.com/2011/09/share-photo-via-text-message...

Sending messages to recipients and load-balancing don't faze me. SMS gateway costs do.

I could have cloned it, web-only, missing out on what I assume to be the main userbase.

Of course now with smartphones, web-only is increasingly feasible.

Snappier performance, longer messages, and actually indexing them might make it worth a shot. Except it wouldn't make any money either (I refuse to inflict more advertisements upon the unhappy world).




Applications are open for YC Summer 2019

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: