Ask HN: Is it economically possible to clone Google Search on the public cloud?

I guess I'm looking for a simple yes or no in terms of yes, the costs of the public cloud would not preclude such a service from being ( even marginally ) profitable. Or no, the public cloud is priced in such a way that a business alike to Google Search could never exist atop it without heavy losses.

But I realize the reality may not be so simple. Hopefully I've constrained this enough with assumptions that the question is answerable and clear. And I don't know enough about the economics or infrastructure, or usage, to be able to even begin to calculate this reasonably myself, nor to have a feel for it. My feeling is "it would be close." which could be totally off the mark.

I guess my intention from asking this question is to expand my intuition for the "lay of the land" of cloud computing economics in 2017, so that I can better judge the type of and feasibility of businesses based on this model.

Finally, I'm not sure if this is the best place to ask and apologize if I've annoyed anyone here by asking such a ( possibly construed as ) bold question. It's not a super important question, it just popped into my head. I checked out reddit but didn't see any cloud forums that were active. And this place is probably better than asking on SE, because the answers, even if a bit irate or confrontational sometimes, are probably more diverse and have a chance of being more informed and in depth, in my experience.

Considering that the profit margins of Google search are pretty high, yes, I think it's possible to run it economically, though maybe not competitvely. You would also be limited to a 90-95% product, because a lot of optimizations are not possible in the public cloud vs. your own data centers.

Well, not the biggest competitor, but DuckDuckGo runs mostly on amazon AWS: http://highscalability.com/blog/2013/1/28/duckduckgo-archite...

Yes, maybe something like DuckDuckGo. So I looked and the CEO says they are profitable ( does not say by how much ) and Crunchbase says they only took 3M all this time.

And yes their search results give a qualitative idea of what a "replica of Google" might look like.

I guess that answers it. It would be nice to know their usage numbers and AWS bill.

Their usage numbers from 2017-01 were around 13 million direct searches[1]. I guess that's "only" 150 QPS. Looking at some AWS pricing information and stories, maybe that is affordable ( a couple hundred to couple thousand USD per day ). But I really don't know enough about it to say for sure.

Maybe I have no idea what I'm talking about, but I'm skeptical that a search engine with enough users to be well known and profitable, can only take 3M. Do their ad revenues really cover their staff and AWS costs? This could be completely wrong, but isn't DuckDuckGo subsidised by some other partners?

[1]: https://en.wikipedia.org/wiki/DuckDuckGo

> Currently we're just making money on affiliate sales to amazon and ebay. We may do an ad or two, but in a minimal way and in a way that protects privacy.

https://www.quora.com/What-is-DuckDuckGos-business-model

actually from their own website https://duck.co/help/company/advertising-and-affiliates:

DuckDuckGo generates revenue in two ways, while upholding our privacy policy:

1 Advertising

2 Affiliate revenue

Ok that is interesting. I guess those affiliate revenues are lucrative.

However, and not that this is regarding the original question, but taking this further, I question how far a search company can scale if its revenue is affiliate deals.

In some sense, they're captive to the pricing power exerted by their affiliates. Whereas a more numerous market of diverse advertisers have a lot less bargaining power to set prices, because their individual contributions are smaller and there's more of them.

So it seems that's one reason why, if your main revenue stream is affiliate deals, that could be a bottleneck on your growth at some point.

Well, sure, it's likely that DDG is on the extreme edge of outcomes.

They're not likely paying AWS list price.

Also likely, they are not optimizing for a narrow scope (higher efficiency)

DuckDuckGo is not a very good comparison, since they don't do their own web indexing. Instead they aggregate the web search results from various sources like Bing, Yahoo and Yandex [1].

[1] https://duck.co/help/results/sources

One of the sources listed their is DuckDuckBot, which is them doing their own web indexing.

But yes, they do augment it enormously with 3rd-party input. It's unlikely to be their primary source.

It depends on what you mean by "economically" — if you mean cheaply, then no.

Exactly as I said, like revenues outweigh costs.

But you're right that economically can mean cheaply. I ought to change it to "economically feasible" since that phrase has more connotation of revenues outweigh costs. Unfortunately I can't edit it now tho.

You mention cost, but you don't mention quality. 'Clone of PageRank' seems the largest cost in terms of quality.

I did but looking at it again it isn't that clear: "replicating basic functionality", "some semblance of relevant search results". You're right that emphasizing that more would be clearer.

So since the edit horizon for the post has expired I'll add it here.

Regarding quality criterion: replicating PageRank is possible ( link graph betweeness centrality ), and AFAIK, it's not this algorithm, although it is useful that yields Google's current edge, it's their proprietary embellishments. What I intended for this question is it is a replication of PageRank algorithm, which is still AFAIK, very useful. We can also assume that we add any low cost improvements to it that help to fight spam or otherwise improve results. Qualitatively, people looking at these search results would say, "well they're not quite as good as Google, but they are good, there's little spam, they're aware of my search history context." So basically, a lightweight version of Google Search.