Empires built on free code aren't cheap

Five years ago, Joe Kraus declared that it was a "great time to be an entrepreneur." In the midst of dwindling hardware and software costs, among other things, it's never been easier to start and scale a company. Or so the reasoning goes. It's undoubtedly true that startup costs have gone down. Ironically, this comes at the …

[Anonymous Coward]

Also..

" it ends up writing or customizing quite a bit of code.

That's not cheap."

Should also mention if costs of (a) writing completely custom software or (b) licensing software from ibm/microsoft/whatever compared to said customisation of existing open source code.

Having worked in such a startup at it's growth stage I can say that licensing at scale is much much more expensive than customisation and setup of existing open source code.

In fact, this company scaled at a point where it needed to either expand its oracle license or custom built code that would take load off oracle DBs so the license would not need expanding. They custom built the code using standard tools and open source technology because it was cheaper.

Scale works both ways. Licensing in my personal experience turns out more expensive.

[gimbal]

Not Linux's Problem

So many questions...

- Would Facebook's costs really be any different if they were using closed-source platforms?

- To what extents have they had to customize anything in the Linux environment?

They'd still have to write new code, whether they were using closed-source or open-source platforms.

You've got peanut butter in my chocolate sir, and chocolate in my peanut butter. I could've sworn I caught a whiff of propaganda about open-source licensing, somewhere around here.

If I have so misunderstood the intentions and the effects of the author's statements about FaceBook's choice of open-source licensing, "My bad."

[Tom 35]

While the company may not buy as much software...

Maybe they went with free software because they could modify it. It's not like Microsoft have something they could use off the shelf, they would still have to do a lot of coding and it would have to sit on top of the software they paid for since they can't modify it.

[asdf]

missing the point

I fail to understand how using proprietary software will save money on the hardware costs. The best open source software for the most part scales about as well as the expensive stuff. The article implies using proprietary software is cheaper if your business takes off. I will grant growing an online business is not any cheaper these days but I believe this had very little to do with the software.

[Charles 9]

Not the point.

What the article seems to really say is that the software costs are probably not the biggest expense of a startup, especially not now. Proprietary or open-source, it's not as expensive as the INFRASTRUCTURE needed to operate at scale. And, unfortunately, infrastructure goes back to physical assets: servers and space..and electricity. Infrastructure will always be a limited resource, which means it's going to cost good money to acquire and maintain. And electricity will probably not get much cheaper for the near future barring some breakthrough.

[Hein-Pieter van Braam]

Actually...

What he's ACTUALLY saying is that software costs ARE a big expense of a startup but that the cost of the software becomes less important once you start scaling up, a lot because then the infrastructure becomes the highest cost.

On a sidenote: Even then, having the ability to customize the software is probably a big bonus, especially if you an epic number of machines, it would be possible to strip everything you don't need and not have to worry about maintaining those bits.

[Danny 14]

right

this is only looking at WEB outfits. More traditional businesses such as one of my clients that run 3 garages are laughing now they have a LAMP server for their web presence and linux for their EPOS - not an MS system in sight.

Another of my clients run 5 shops that sell dissimilar products but all run from LAMP and linux EPOS again, same situation almost. Keeping off MS has 1) stopped worrying about the stupid and horrific licence models that exist for normal people wanting SQL, ISA VPN (or whatever it is called nowadays) and exchange without worrying about how many times they bring their laptop in from home or their employees VPN'g from home to do work.

[Andy 97]

"Fee fi fo fum"

I smell the blood of PR from Redmond!

[psyq]

Redmond?

Matt Asay is chief operating officer of Ubuntu commercial operation Canonical. With more than a decade spent in open source

Gee... doesn't look like a Redmond PR-droid, does he?

[johnnytruant]

M$PR

Pro-Microsoft PR in an article written by the COO of Canonical (Ubuntu)? An Open Source Initiative board member?

Really? You're sure about that?

[Stuart Jones]

Yes, but...

With people like Google, MS and Amazon offering very scalable hosting solutions, it might just provide a mechanism to scale until the product is monetised sufficiently to break out. In addition a lot of the hard work earlier web companies have done in creating scalable frameworks is now done (e.g. NoSQL etc.)

[amehaye]

Re: Yes, but...

Ah, but you cannot break out of Google or Microsoft solutions. They incur a very tight lock-in. Amazon is the easiest to break out of, which is exactly the reason for its huge success. Developers aren't stupid, they see lock-in for what it is.

[Stuart Jones]

Not necessarily true

If you run Microsoft technology anyway then you can easily move away from Azure assuming you avoid Azure specifics. Azure will also run PHP and will happily run things like NHibernate

Google is more vendor lock-in because afaik you cant host your own GoogleApps

Either way, my point just taking Amazon as an example you can scale up quickly without huge infrastructure investment

[Mike Taylor]

Point of the article

Is that software (whether licenced or open or free) isn't the only cost, that network infrastructure is a very expensive part of building a company.

He's not making points about licence types, nor open vs proprietory scaling.

As the COO of Ubuntu Linux, I seriously doubt he's criticising Linux. Or being a MS shill.

[Doug Glass]

And as COO ...

... you have your company's financial reports to back that claim up. As opposed to others who "know this" or "know that" and have nothing to support their claims other than what's in their mind about how things work.

[Anonymous Coward]

Who reads reports...

... when you can read the latest blogpost from some ubuntu bigwig instead?

[Steven Jones]

Profoundly wrong...

The analysis is surely profoundly wrong. Nobody is going to quibble with the notion that it requires serious amounts of expenditure to become a dominant player in one of the mass-market platforms like Google, Facebook, Flickr and YouTube. However, the fact is that many of these have grown from essentially nothing in just a few years to user bases in the hundreds of millions. That simply would not have been possible without the ability to prototype and develop from an array of open source software running on commodity hardware. Then there is the availability of people familiar with the technology.

Try comparing the costs involved with using proprietary software or developing completely from scratch. Yes, it did eventually take serious money to fund the incredible expansion of these platforms, but investors only came forwards once the principles were established. In that sense it was the open source revolution that was the enabler. The extensive modification that now goes on is just now part of the continual fine tuning and optimisation that inevitably moves these platforms to get economies of scale.

However, there are some problems. Firstly it is evident that the exclusiveness of some of these platforms tends to monopoly. Facebook's success is crowding out all the competing platforms. We know what Google has done in the online advertising market (and is busily extending into other areas). Being able to do this isn't just a matter of the cost of development, it's often just down to issues of interoperability, the way user's organise their lives and simple convenience. Also these monster, but low margin companies, are now in a position to spot innovation and kill or buy it before it becomes a threat, a tactic used to devastating effect by Microsoft.

So the golden age of new mass-market platforms may yet be over, but it will be the old fashioned techniques of using market strength and exclusive technology that will be the cause. What I think is inarguable is that we would not have seen the Web and Internet as it currently exists without the solid base of open source software enabling the innovators to offer new services. What will remain will be a few vast low-margin companies with some much smaller niche players. Cheap software will be the enabler, and the key matter here is technology free of monopoly IPR owners.

[Doug Glass]

Huh?

Re: "What I think is inarguable is that we would not have seen the Web and Internet as it currently exists without the solid base of open source software enabling the innovators to offer new services."

I assume since you make the statement that your conclusion is "inarguable" that you have the cost figures in your posession and can therefore backup your claim by publishing them. Along with the source data and the statements by the underwriting firm's testiment to the data's authenticity and validatity.

I say you don't have that data and supporting documents and that makes your argument, well ...not exactle inarguable but simple bluster.

[Steen Hive]

@Huh?

Maybe you should direct the flame at the point he was making instead of a different point that seems to exist only in your head.

Or are you asserting that it is plausible the Internet and Web as it exists today would be entirely the same as an Internet without, say - the Berkeley TCP/IP stack, or something?

[copsewood]

software isn't cheap

"While ostensibly free, to make projects like Linux work for its purposes, Facebook heavily customizes them. While the company may not buy as much software, it ends up writing or customizing quite a bit of code."

Except where they are using Affero GPL code, Facebook are not under any obligation to contribute back. But if they don't, they'll find maintaining their changes cheaper against a rapidly moving mainstream project (e.g. Linux kernel) if they do contribute back and maintain their improvements and facilities as part of the mainstream. Many companies pay open source specialists to do this work for them e.g. Canonical, Suse or Red Hat, simply because working effectively within the wider development community to obtain the cost benefits of fork avoidance can require significant expertise.

Because software isn't cheap, whatever improves reuse is worth looking at, and this is the real strength of the open-source development model.

[Destroy All Monsters]

So going to epic scale is expensive?

Uh, yes. Also, water is wet, Sky is blue, You need to be at the right place at the right time (preferably while the economic bubble is still inflating) etc. etc.

"average seed deal size has actually climbed to $5.4m, up from $4.0m in the fourth quarter of 2009"

Just means that money is still cheap. Better hurry though, the next stage of the economic crash can't be far off. And this time, there will be dead people hanging from trees.

[Anonymous Coward]

meh

hmm let's test the economies of scale of FOSS and Windows across 50000 servers.

50000 x £1000 (notional cost for 1 M$ server license) = £50,000,000

50000 x £0 (cost of FOSS) = £0

or did I miss something?

[The Original Steve]

meh to you too

You forgot training, pay disparity between FOSS and Windows admins, additional overheads for documentation any customisations (on a Google scale you'd need a large department just for that), product support (or pay SERIOUS cash for your own Red-hat like super geeks to maintain your code and JUST your code) etc.

It's called economies of scale. Red Hat can offer RHEL at £2k per server per year. The support part eats a LOT of that up.

Someone who really knows a heavily customised Linux platform - so custom it's nearily impossible to trace back where it even forked from - and I'm talking someone who is going to HAVE to be the last line of defence - will not be working for under £80k a year. And that's one person - he'll need a team. Architects, documenters, testers, security pro's, network programmers etc.

So no - don't kid yourself. FOSS in a real enterprise environment is no cheaper or expensive than Windows or any other non-FOSS product.

And FYI - do your research. My Windows 2008 R2 licences are coming in at £290. That includes free support 24/7 direct from the developers, training on the platform, TechNet, MSDN, (shit load of training/development environments), and of course I'm saving about 30% for every tech I employee as they don't need to modify of a load of text files just to configure a web server. (e.g. Windows tech's are cheaper than FOSS).

Your argument doesn't stack up.

ide out will want a much higher annual fee. I wouldn't be happy being the last line of defence with no vendor support for a Linux distribution for

[Sooty]

don't forget

There are other issues at play here. All a company really cares about is getting something to work, if that means that they have to get all Microsoft kit, then they will. If it means paying a bit more... then they will!

We had this a few years ago, for a mainly inhouse development, we originally went for the best solution all around, including hardware and software from a number of different vendors, but after several months of effort, we just couldn't get some of the IBM and Microsoft kit to play nicely together, so in the end, after exhausting all options, we rejected the contract with IBM and went all Microsoft!

In the end, it didn't matter which was sticking to the standards, or which was cheaper, or who was providing the source code, it only mattered what worked for us in the time we had!

[Danny 14]

not true

Training is irrelevent as this is from startup hence the same. Plus your 2008 licences seem expensive. Mine cost £29 base with £105 support (phoenix software) - granted I only have 5 servers so maybe it scales upwards? . My MSDN is a flat £270 for the year

I also imagine that at 50000 servers they will have many more "points" to play with, thus get a slighly better deal than us mere plebians.

[The Original Steve]

@Danny 14 / Not True

Um, yeah it is true, and you cannot own a legal licence of Windows Server for £29.

Maybe for Windows 7 Starter OEM, but not for Windows Server.

The £290 is Windows Server Enterprise via a Select Plus Agreement. We have 260 servers.

Let me know your LAR / licence reseller and I'll be happy to buy Windows for £29! (FYI - MSDN is per person. A SA addition to your licence allows you MSDN and TechNet for free - but remember it's only for training, development and testing!)

And training is relevant. A start-up means what? Having one guy who knows some PHP is NOT the same as what I described. A home-taught PHP dev turning into a custom kernel maintainer will require a LOT of cost. (either time or money)

[Anonymous Coward]

Yes you missed something.

1. of you are having 50,000 licenses turst me, you won't be buying them @£1000 each.

Next.

50,000 servers =

several datacentres + routers (multilayer) + power, plus rent + swtiches + bandwidth + staff +,+,+,+.+

"£50,000,000" is peanuts compared to this.

THAT IS THE POINT OF THE ARTICLE.

[heyrick]

Aaaaargh! "monetize"...

Most annoying damn word of the decade...

.

As for this article. Um... What is it trying to say, exactly? That setting up and maintaining an epic-sized data centre is expensive? Well, duh.

Given this, I'm not even sure if the first few paragraphs are even particularly relevant. A thousand ultra-fast server machines is a thousand ultra-fast server machines. I can actually see MASSIVE benefits to open source solutions which will be of great assistance in helping to reduce costs. Consider, firstly, a thousand ultra-fast machines with a thousand licences for a proprietory operating system. Expense that doesn't need to be. Consider reliability and ease of deployment of said system.

It is expensive to customise free code? Well, firstly it wins on price because you CAN customise the code. I'm sure some big companies have code-share deals with Microsoft for the ability to customise things, and I'm sure it comes with a ton of NDAs and a suitably large price tag. Or is the article wishing to assert that modifying existing working code is pricier than designing something new from scratch, along with off of the groundwork that has to be done (and is, mostly, already present if basing upon an existing solution).

[small question: if facebook modified GPL code, would the licence not imply they give back to the OSS community?]

So, after all of this, I'm confused as to the point this article is trying to make; other than getting popular and running massive server farms is a pricey thing to do, especially in a culture where Free is Glorious and there's little concern given to how things like Google actually work. But that's not an article, it's all pretty self-evident, really.

[frank ly]

Use by friends and employees is not redistribution.

" ... if facebook modified GPL code, would the licence not imply they give back to the OSS community? "

If you modify GPL code and redistribute it, then you have to give the source code, including your modifications.

Using modified GPL code for your own purposes is not redistribution of the code. A company or organisation can install modified GPL code on thousands of computers and use them for their own activities, provided those computers belong to them. What they must not do is give the executable to anyone and say, "here, this is cool, try playing with it at home." (in that case, they must give the source code as well).

[D. M]

The point is

Software cost isn't the biggest spending.

However, it is still better if you can save some money on software and licensing. They are not CHEAP.

[Anonymous Coward]

Problems with business model, not scale

For our sales and support site rack hosting costs us £40 a month. The server itself costs us slightly less when amortized over its expected life. Keeping it up and running from a technical perspective is perhaps four hours a month. New content is mostly generated anyway for more general marketing and support functions but still costs something to get into a form for the web site and physically get it up there. Even in you include the online advertising the site probably costs in the region of £300 a month.

For that we get around 5000 visits (not hits) per month. Around 80% of those seem to be frivolous visits that are never going to lead anywhere - casual browsers and the techo-wankers. 15% are support related and that saves us money compared to supporting them in person. 5% are genuinely prospective sales and around 1% ultimately lead to a sale although that may take several visits.

The result: around ten sales per month. That is admittedly a poor conversion rate but we're talking about high value stuff here: on average just over £20000 in takings for our £300 per month. Our server hovers at a pitiful 1.5% average load. We could scale to £1 million per month still with a single server and only a small increase in bandwidth costs. That's an unbelievably good deal. If we _were_ getting £1 million a month in online sales then we'd be more than happy to spend the extra to scale the web site.

The figures speak for themselves because the business model is right - it is clear from the very outset where the money is coming from. It is also true that the web site is not the primary business but the central fact is that visits are profitable on average. If that is true the costs of scaling are irrelevant to the extra income being generated. If it is not then losses are going to scale with visits.

If you have a business model where you are pushing out expensive content for a niche market free and hoping to claw back the cost with the pittance you get from online advertising that isn't going to work. The reason for that is economic rather than technical - technology is not going to come to the rescue of a fundamentally flawed business model.

AC since while I don't have a problem going through our figures no doubt somebody considers them commercially sensitive.

[Joe Montana]

Scale...

If you have requirements which aren't addressed by readily available software, be it free or commercial you will have to write it yourself... The fact that open source code already goes much of the way for free, as opposed to having to write your own from scratch, or to use commercial code that may not even provide the ability to customise it (or charge a huge extra for the privilege), still makes the open offering much cheaper.

Also having to modify code isn't about scaling... If you needed to write or modify custom code changes are you did much of that when starting up, and once your modifications are made to open source there are no additional software costs to run on any number of systems. Commercial software would typically charge you per install or even per cpu, so if you have thousands of new users and need 10 more servers thats not only 10 more sets of hardware, dc space and power but now 10 sets of software licenses on top.

Whichever way you look at it, operating a big internet site with proprietary software is pretty much impractical... The only major player who does that is microsoft, and the usual disadvantages don't apply to them since they have the source code and don't have to pay for licenses.

[Gareth]

You mean setting up a giant company isn't free yet??

Echoing the sentiment of above; why is it news that setting up a company that provides service to a notable percentage of the world's population is still expensive? Surely the cost of doing that has still come way, way down compared to what it would have cost 20, 10 or even 5 years ago.

Also, the level at which low-level customization of infrastructure is required keeps on rising - the percentage of companies that even need to do this is falling. I'm working with enterprise clients with regional markets for their sites and the most we've had to do for performance is to tune Apache and memcache.

The bar continues to lower. Maybe the cutting edge giants are not the place it should lower to - those guys will always need more than they can get off the shelf. Is it even a realistic goal to want to provide everything for them in a stock install?

[henchan]

outside the big box

There is no immutable law which says that web application architecture must comprise exactly one (logical) server and so many clients. Certainly, it tends to be that way at present, but the scaling costs discussed here might come to be seen by a new generation of entrepreneurs as the economic driver for an alternative architecture.

Wave was designed to be federated. P2P file-sharing is another case in point. Just one massively successful implementation of a distributed web app design (which admittedly has not occurred yet - unless you count the web itself as an application) would send a signal into the marketplace that those big-box investments in centralised data-centres were not so smart after all.

[heyrick]

Servers and clients

Re: There is no immutable law which says that web application architecture must comprise exactly one (logical) server and so many clients.

I always find it odd when some (news, etc) sites get way over the top numbers of visitors due to an external event, and this is reported as "crashing the server". Can't the server accept the connections it can accept, and when it reaches capacity just *not* accept any more connections? Or redirect people to a simple non-parsed non-PHP response saying "We're busy, try again later"?

[Catweazle]

Easy to gain popularity?

"Unfortunately, it's often easier to gain popularity on the web than it is to actually monetize that popularity..."

Easier to gain popularity than to set up a profitable business?

Is this on the authority of Prof. Trewlaney or based on statistical figures?

I'd like to see some numbers on sites succeeding in being popular and sites succeeding in being profitable.

[Eddie Johnson]

Quite a Lopsided Field You Chose

"Facebook has to run all that open-source software somewhere, and does so in its own custom data center and leased space within others' data centers. Twitter just announced its first data center. Google? Its data centers are legion."

So you're comparing the start-up to 3 companies who are pretty much at their pinnacle? Not to mention 3 companies which are all ad based so they need to churn huge terabytes of traffic to make a tiny margin? How about comparing to some actual companies that generate actual, useful products? (I'll concede Google's product does have some actual value but not the other 2.)

Or is your point that worthless, nonproductive companies catering to the latest fad are the future of venture capital and the internet? You may be right with that, I've never been impressed with what I saw the VC dollars chasing. They seem more interested in longshots than useful products.

[Henry Wertz 1]

Internet, and the article

@Doug Glass, this really is pretty well inarguable, not because of simple costs like you are implying. But, historically, early implementations of ARPANet software were ALL open source, as it was a research project that numerous people at different locations could and did improve. The OSes were sometimes commercial but the user got access to the source. This was important, as it allowed anybody with enough programming knowledge to make improvements to the TCP/IP stack, either in the form of actual code, or in the form of seeing how it operates and generating RFCs (Requests For Comment) suggesting improvements.

What did commercial companies come up with? X.25, which was complex and strictly connection-oriented. , Thee were also proprietary protocols (DECNet, SNA, and so on). Not saying they were bad but they were not interoperable except through specialized gateways (which were not automated -- one had to connect to the gateway, then to the next machine they wanted to go to -- or in the case of E-Mail specify the gateway manually.) Later on? OSI -- this was "protocol by committee", and frankly didn't work out at all. Many commercial TCP/IP implementations (including the one in Windows) use the BSD code (BSD licensing doesn't require any release of code, just a BSD license file somewhere or other.) Either X.25 or OSI would not provide anything similar to our current internet.

------------------------------------

As for the actual article -- it's all true. But Matt Assay, a COO at Canonical -- really? You could have put a *somewhat* better spin on things. Here it is -- it's all true, open source is not going to make running a large-scale operation suddenly cost-free, because hardware is expensive. Good sys admins, programmers, etc. are also expensive. However, there are several advantages --- 1) Since the source *is* available, modifications can be made to both customize software in other ways, and to find optimizations, even if site-specific, to speed up the code. A large, popular site will still need large amounts of hardware, but this will reduce the amount of hardware needed compared to just having some closed code that *can't* be customized in any way. 2) Licensing costs. I really don't think I need to add to this, the larger the number of machines, the more licenses the operator would presumably need. Not having to pay for this is money in the bank.

[Anonymous Coward]

Err...

Windows doesn't use BSD code for the IP stack - it did, very briefly in the days of WinNT 3.5 (or was it 3.51, I can't remember) it certainly hasn't used BSD code since NT4.

The code for Windows is available under NDA to many companies. You can modify Windows at pretty much any level, for example I used to work for a company who had a custom gina, wholly written in house to suit our requirements. However in a recent project I worked on, which was installing a few RHEL clusters, we wanted to change some parts of RHEL and were told that we wouldn't be supported if we did, which was a bit of a show stopper.

In my experience the licencing costs of Windows (and other closed source/commercial systems) don't outweigh the cost of support for systems like RHEL. You have to have support, unless you are confident that you can fix any problems in-house.

[Jay Jaffa]

What was this article trying to tell us?

What is this guy trying to say, what's the message? Building out and managing huge server farms is expensive? No shit sherlock - what's it go to do with Open V Closed software?

Muppet Magic.

[Simon R. Bone]

you're forgetting one thing

whatever you're spending on hardware/software you generally need to spend 10-100x more on marketing to make it a success.

[Anonymous Coward]

What can ubuntu do for you?

It is not clear from this article what financial benefits accrue from using open source software, with business models being equal. M$ server, desktop and Office licenses, Oracle Cisco and SAP have real upfront costs... there can be custom coding and administration required there as well. The list goes on. Are there open source alternatives available to fully start-up, expand/scale and maintain an online enterprise, requiring no license fees. It appears not yet anyhow.

[Humu]

Business as usual

You are saying: "Open software is initially cheap. Later you shall probably make some modifications. That will cost a lot!"

So what's the point!

Proprietary software is initially expensive. Later you need to make some modifications. That will cost even more!!!

---

Seen that, done that - Cheers

[Tom 7]

Not in my world

I've helped several small companies with their IT. Most Linux server setups seem to run to two or three servers replacing 10 ort so MS servers. If the customer has some IT knowledge (NOT MS Certification tripe) in house its often possible to set them up correctly from the start - if not they often choose the easy/cheap way of doing things which makes expansion at a alter date almost impossible. This seems to be true of almost all MS setups - once you get past a certain point (20-50 office installs) it is almost impossible to move on to real enterprise setup as the engrained practices are too important to break. OK this happens in FLOSS setups too but at least the decision is made by the customer when FLOSS is available - MS seem to enjoy putting gastric bands on small companies.

Most companies still use IT as a typewriter/filing cabinet replacement. With FLOSS your data can be free as service to the customer is what makes the money for the consultant - not commission on licenses.

[Eddie Edwards]

What's a startup?

"If startup costs are lower, apparently VCs and startups didn't get the memo."

Maybe that's because startups with low costs don't need VCs.

[DrXym]

Not sure I understand the point of the article

Even free software costs money in hardware & personel to run. It doesn't alter the fact that if its the best tool for the job (as it often is) then companies are saving themselves money and grief.

[Anonymous Coward]

Dear Mr Shuttleworth

Are you sure you hired the right guy for the job? This man came from the very same 'Open Source' company that was so full of hype that it hurt. Alfresco was basically a bag of enterprise BS. This man is putting out very confusing messages and probably hurting your target market with exceptionally subjective 'factual' editorial pieces like this. The enterprise Linux market does not need this. You'd be better off spending the money on another personal space flight.