[jacquesm]

> Person scaling with Azure here. It's pretty awesome and easy.

And it's pretty expensive.

Now, if your competitor is on an open stack and on real computers they will eventually run rings around you because they get more control and lower costs. The initial boost you're experiencing will turn into a straightjacket with a hefty pricetag over the longer term.

[gadders]

Stack Overflow seem to be doing OK: http://highscalability.com/blog/2014/7/21/stackoverflow-upda...

I think there is a case for an open stack, but I think unless you pick a real dog of a platform good people can make it work well. It comes down to a "can I get good people for stack X?" as much as "Does stack X work or cost too much?"

[jacquesm]

Just like in the case of Trello if you already have the experience you can probably save money if you use what you already know how to use. In almost all other cases open source stacks are cheaper.

Stackoverflow and Trello are always held up as the shining examples of how the Microsoft stack is able to hold its own. That's great, I even know people that insist on running their web servers on Apple hardware. Whatever floats your boat. But if you're running a competitive business, if you're going to be using a lot of CPUs and if you will eventually (or already) be faced with lean and mean competition then you are probably better off on an open stack.

Try to imagine Google, Amazon, Ebay, DropBox or AirBnB on the Microsoft platform. And note that two of the above are re-selling their linux based platforms to other users at a profit.

Btw stackoverflow uses plenty of CentOS, I guess licensing from RedHat was too expensive?

[CHY872]

Big names mean nothing. You are not Google, they have very different requirements to you and their scale is much larger than your startup will ever be.

Microsoft aren't stupid - the costs they charge for Azure are largely competitive, and the benefits it provides are tangible.

Picking what technology will run your business is something that really needs to be done on a case by case basis, based on your knowledge and your business. Certainly I don't think it's desirable to make a permanent choice when you are still a startup.

Of course you'd use CentOS, unless you had need of the specific features that licensing it gave you.

[jacquesm]

> Picking what technology will run your business is something that really needs to be done on a case by case basis, based on your knowledge and your business.

We're definitely in agreement there. The main criterium is: use what you know how to use. So if you're comfortable using the MS stack then go for it, otherwise, probably better to avoid it.

[pdpi]

> The main criterium is: use what you know how to use. So if you're comfortable using the MS stack then go for it, otherwise, probably better to avoid it.

So, would it be fair to say "So if you're comfortable using the Linux stack then go for it, otherwise, probably better to avoid it", or was that your parting jab against MS?

[jacquesm]

No, since Linux & FreeBSD are license free they have intrinsic advantages over Microsoft.

[pessimizer]

I read it as saying that the benefits of not basing your business on a proprietary platform are lower than the benefits of going with a stack that you already have some expertise at. You may read differently.

[gecko]

Trello is based on Node and Mongo on Ubuntu; there's no Windows in it. Kiln and FogBugz are based on C# and SQL Server on Windows and Java and Python Debian. Stack is based on C# and SQL Server on Windows and Redis and infrastructure on CentOS. This isn't about saving money or not; its about using the right tool for the right job, factoring in cost of tooling and cost of development. I think you're oversimplifying in your analysis

[ingenieros]

Even Apple was running iCloud off AWS and Azure before they ramped up their own data centers. I don't think you can get any more competitive than Apple in the tech world.

[enigmo]

eBay's frontend is mostly Java on Windows. Many thousands of servers running Windows. Their search grid was on Solaris but has migrated to Linux in the last year or so (2013?).

[scott_karana]

StackOverflow isn't running on Azure: if anything, that article should make it obvious how much they're saving by not using cloud, with OS being unrelated.

[jaegerpicker]

You do realize that Azure has about as much to do with windows as AWS, right? I use Azure to run a completely open source stack based on node, postgres, and python. I also have a long term windows server in AWS. Azure is every bit as open as any other "cloud" based stack

[skc]

> And it's pretty expensive.

Nice blanket statement there. How can you claim it's expensive when you have no idea what or how he uses it?

[jacquesm]

Because I have fairly extensive insight into a very large number of companies and can see their license bills as well as their cloud bills if those are applicable. Compared to that doing a startup on an open source stack using dedicated machines eventually turns out much cheaper. It's like getting hooked on drugs.

The most cost effective way as far as I can see is to hire a sysadmin on a freelance basis until you need one full time, own your own hardware (or lease it by the month until you can afford your own hardware) and pay flat rate for bandwidth. Anything else will sooner or later come to haunt you and then you will need to migrate to some new platform. At that point in time you will learn the true value of the words 'lock-in'.

[UK-AL]

The biggest problem for a startup is getting out there, and getting transaction and revenue. Not penny pinching over what servers your going to use. You want to use as much pre-made stuff to give you the biggest head start as possible.

When you take off, you can raise money to pay people to move you over anyway.

[jacquesm]

Choosing your tech stack with care is one of the more important choices. You're locking yourself in for years to come and mistakes can be very costly to fix. Getting transactions and revenues are obviously also important, the whole trouble with doing a start-up is that you have to do so many things right.

[notwedtm]

If YC has shown us anything, it's that companies can successfully pivot at nearly any point of their lifecycle and still be successful. It's also shown us that they can fail at pivoting and have it be nothing to do with the stack they chose.

[Spearchucker]

If you're worried about lock-in then design for portability. Portability means using open standards rather than open source. The neat thing is that with a well-designed site or service you can switch vendors easily. Cloud is a commodity - who cares who the provider is?

[onion2k]

Now, if your competitor is on an open stack and on real computers they will eventually run rings around you because they get more control and lower costs.

Obviously it depends on what you're actually doing, but for a straightforward SaaS app (a few servers, a db server, a queue and a load balancer in a few datacentres around the world) one good sysadmin/devops guy costs far more than a set of AWS instances that essentially scale themselves.

Having dedicated people necessary for that 'control' is beyond the budget of the majority of startups.

[jacquesm]

You could of course simply hire freelance sysadmin/devops guys. Initially those are not full-time jobs anyway.

The choice is definitely not between going for a cloud platform versus hiring a bunch of people full time. There are other options.

[gommm]

Exactly plus at scale AWS doesn't do everything either, a sysadmin or at least consultants are still needed to navigate the gotchas and help with how to better provision the stack. AWS instances don't just scale themselves.

There is one specific case when I recommend AWS instead of dedicated servers and it's for customers who have widely varying traffic with predictable peaks. In that case having the flexibility afforded by cloud providers to increase the number of instances temporarily to deal with the peak makes sense.

[jacquesm]

The only good use-case I know for AWS and it's ilk is if you need a 10,000 node cluster for a few hours to do some heavily compute intensive work which does not require a very large amount of data to be imported and exported afterwards. This is a pretty limited number of use cases but for those situations it absolutely rocks. Anything else I'd run the numbers very carefully.

[turnip1979]

I don't think that is correct. There are thousands of startups on AWS (an IAAS cloud) that don't run 10K node clusters. While they cost may be high, it is pay as you go. I have a rack full of computers at home and I still have some things served by AWS. The PAAS model (e.g. Heroku, Bluemix) is becoming popular today but those are often hosted in IAAS clouds.

[wfjackson]

>Now, if your competitor is on an open stack and on real computers they will eventually run rings around you because they get more control and lower costs. The initial boost you're experiencing will turn into a straightjacket with a hefty pricetag over the longer term.

From another post by you: >Because I have fairly extensive insight into a very large number of companies and can see their license bills as well as their cloud bills if those are applicable.

Since you claim to have a lot of info on this can you give any, even one, real world example of you just said happening?

Otherwise I am going to call BS on it.

Perhaps you're starting something that is going to take down NewEgg.com ? You're clearly missing the forest for trees here. At work we run a large mix of Windows Servers, SQL servers, Linux, Drupal, MySQL, Moodle etc. running fairly traffic heavy and top ranking public health sites and all said our licensing and hosting costs are about 2% of our annual budget.

[jacquesm]

Are you actually asking me to disclose customer information and taking my (obvious) refusal to do so as calling BS? Interesting. Look, I have no skin in the game, if you're happy forking over tons of money for stuff you don't strictly speaking need then more power to you.

All I note is that MS thinks YC is important enough to designate a person to sway them to the MS stack and SAAS products, this is a tried and true strategy (get them while they're young) and it will likely cost you dearly in the long term if you aren't able to oversee the long term disadvantages of such a move. If you have a long history of using microsoft products and switching to open operating systems and stack components would mean lost time due to re-training then by all means stick to what you know.

It's funny how in one subthread here people are arguing that hosting and bandwidth are the major expenses for any start-up and now it's hosting and costs are 2%. In the end every situation is different and every situation has a different cost analysis for the workload envisioned. Seeing the guts of many companies has shown me that if you're doing something that requires large numbers of expensive licenses or metered bandwidth / cycles / storage then you're probably going to regret that choice at a later point.

[notwedtm]

Except you're not forced into using Microsoft SOFTWARE when using Azure. Like many others have pointed out, you are free to use Linux variants as well.