Are you building an R&D lab or a business | Hacker News Reader

Are you building an R&D lab or a business

markmaunder.com

47 pointsmmaunder14 comments

Wow, from the title I thought that article was going to take the opposite tact (ie, you should be building a business, not an R&D lab).

Very interesting though.

Without investors or a nice endowment it's hard to work the years needed to create a nice "r&d" product. If I launched twitter the first server bill would have put me out of business. There is a balance between making something cool and something "uncool" that people might pay for today.

really? Servers are cheaper than you think, if you don't use the cloud.

I got a quote from one of my vendors the other day... fifteen hundred bucks a month for a full gigabit of traffic and 15 amps of power. (now, I haven't bought from this guy, so I dono... but he is reselling he.net bandwidth, which is fine stuff. I got another quote direct from he.net at a /much/ better data center for just the gigabit, and they wanted two grand a month, before negotiation. Now, for me, power is a big deal and bandwidth is less so... but both are pretty cheap when you get down to it.)

that's enough power to run 9 single-socket 8 core 32GiB ram opterons, give or take. each one of those puppies is about two grand in parts, give or take, before disk. And a gigabit is a ridiculous amount of bandwidth, especially if you mostly push text and thumbnails.

Now, I don't know how much ram/cpu/disk per user twitter uses. But you can push rather a lot of text with that much cpu/ram, and a grand and a half a month is pretty reasonable 'hacker with a job' sustaining money... hell, the 18 grand for the servers isn't any more than you'd pay for a low-end new car.

Now, obviously, working full time and running a startup is /hard/ - but my point is that you don't need VC level money to do this... just about any fully employed hacker has the means to support a ridiculous amount of infrastructure.

> Servers are cheaper than you think, if you don't use the cloud

I'm confused why the cloud is viewed as expensive. For managed hosting on a big name provider perhaps, but there are dozens of virtualization providers who are capable of providing power and bandwidth for unbelievably small amounts of money.

Sure you don't get the uptime, reliability, and support of a big name host, but it's usually good enough. From what I've gathered, twitter's growth has not been severely hampered by the fail whale, especially in the beginning.

Once you've reached critical mass, you can move on to a larger host. I think it makes sense to rent rather than own, especially if you're just starting out.

I priced out some servers and cloud providers. It appears that the "cloud" servers are about 30% more expensive. Though the gap widens when you start building bigger machines (32G RAM, latest processors, etc.) EDIT: I'm looking at 2-16G machines.

On the other hand, in my case the extra capacity would sit idle for a while. (I replace servers every 3 years.) Going with a cloud platform, I would have likely upgraded a few times and taken on less risk.

The cool thing about cloud hosting is that you can always switch to owning your own hardware if it will save you money. It's much harder to go the other way.

Also, once you try to shore up risks and add things like drive snapshots, the costs get much closer. (ie: start using all of the features offered by Amazon AWS.)

how much do you pay for servers? 'cause I'd be out of business if the difference between amortised hardware cost and what i charged you was 30%

I'm one of those virtualization providers... but even at my scale, which is pretty small, I price my services such that worst-case, I pay off my hardware in four months. that seems expensive to me.

Of course, if you need small servers (and I only rent small VPSs) I'm saving you money. even if you get the server for free, (which isn't that difficult) finding someone to host a server with 512MiB of ram for $12/month is going to be really hard.

Compare hiring a NW admin and system admin at ~$8k/mo. each. An MSP can provide you with a lot of servers / powerful servers before you even come close to spending $16k/mo. Choose your MSP carefully though. At a minimum consider the SAS70 report from several before choosing. That will 'cull the herd' of the MSPs who make claims and promises but don't back them up with verification from an independent auditor.

You don't need a full time SysAdmin (much less two) until you have a /lot/ of hardware. Until then, you can get reasonably good contract SysAdmins for $100/hr to do your hardware stuff if you can't handle it. (your average independent hardware guy will contract out closer to $40-$70/hr, if you get him or her direct.)

Really, from what I've seen, most of the "cloud" providers only replace your hardware guy; you still need a SysAdmin (or a developer who can handle SysAdmin type stuff, which is probably what you want unless you are an infrastructure provider.)

It's like anything else; If you know the market and know what you are doing, you can save a lot of money by doing or supervising parts of it yourself. If you don't know what you are doing... well, sometimes it's best to find the person with the best reputation you can find and pay full price. I know I'm that way with accounting and legal services; I overpay because I don't know enough about the field to do or supervise any of the work myself.

But, I think that if you spend significant amounts of money on server infrastructure, it's probably worth your time to learn enough to get a better deal on things like contract labor.

If you would be building the Twitter, had no money and server costs were exploding, I guess it would be pretty easy to get some money somewhere by just showing your visitor stats?

That's correct for an angel round in the valley right now provided the stats are very recent and they're trending upwards i.e. your growth is accelerating.

The reason is because growth is what all valley angels look for. So if you have growth you create (as Cialdini would put it) social scarcity for your stock which creates a feeding frenzy.

This is only in the valley. Seattle (where I'm based) is different. Many startups here focus on generating revenue from early on and there is less growth based demand. The valuations here are also lower. I think the number of investors is the reason. Not enough for a feeding frenzy and less growth-based exit stories to fuel the appetite.

How long do you think the stark regional disparities in funding are going to persist?

Well it's only been 40 years. But financial armageddon is nigh so perhaps a decade from now China will be the new valley and the only thing the US will be good for is a place to base your servers since we're the center of the giant Internet star topology.

And every one of them is focused on generating cash in their business. The road they’ve chosen is a longer, harder road with a lower chance of success but a much higher reward (think Michael Dell, Bill Gates, Larry Ellison) if they succeed.

Wait, so generating cash in your business has a lower chance of success than getting acquired? Maybe it's because I'm not on the west coast, but I know at least 20X more people with cashflow-positive companies than people whose companies have been acquired.

Or maybe he just means they have a lower chance of becoming as successful as Dell and Gates and Ellison, than others have of being acquired. But he seems to imply that if these people don't generate as much cash as those super-companies, that they've somehow failed.

...which reminds me of the thread on here not long ago on when exactly a startup can be considered successful: http://news.ycombinator.com/item?id=1383566

Reply on news.ycombinator.com