Do startups use Open Source?

I was very much of the impression that startups these days, because they want to keep spending to a minimum, would be more likely to use Open Source tools to develop their applications. The likes of MySQL instead of Microsoft SQL Server, for instance.

This view was re-inforced by an interview I did with Salim Ismail for the it@cork pre-conference podcast series where he said all his startups used open source software.

However, after a chat with Microsoft’s Rob Burke on his blog, now I’m not so sure!

In my comment, I said Microsoft’s SQL Server should support other platforms and in this way, startups would be more likely to use it (i.e. if they didn’t have to splash out for a Windows license). Rob’s answer surprised me though, he said:

Our group at Microsoft Ireland can, quite literally, not adequately keep up with the demand we get from local startups (and larger ISVs) who see the value of the platform for the data tier and want to find the best on-ramp. You may have noticed – we’re hiring two more evangelists! 🙂

So startups in Ireland are choosing Microsoft SQL Server in droves? Why? The latest version of MySQL has stored procedures, triggers and views. It is platform independent, has a very strong support community and runs some of the better known sites on the web like Craigs List,, Digg, Flickr, and Wikipedia, to name but a few.

If you chose SQL Server, you are locked into the Windows platform and although there are free versions of SQL Server to start out with, a fully licenced version to run a web site will cost you tens of thousands of Euros/dollars.

Why would any startup choose SQL Server? What am I missing?

27 thoughts on “Do startups use Open Source?”

  1. We use a combo of both in work. Mainly down to the fact that we use different applications. I’d be more a fan of mySQL than SQL Server as Server really does have an “enterprise” feel about it that can be a bit intimidating. MySQL can do a hell of a lot and still not be intimidating to manage.

  2. It seems hard to think of! Brand new companies flocking to buy something for several thousand dollars that they could get for free… Wow.

  3. Coincidentally I asked a very similar question earlier this morning on the LouderVoice blog. We are going entirely the LAMP route (with P=Python in this case) and I wondered if/why any startup would go the MS route. Total licence cost for us for DevTools+DB+Framework+OS = 0 for a live system. What are the equivalent MS numbers?

    I have no real issue with “lock-in” as once you commit to an approach, open source or not, the cost of switching is prohibitive. However if the mantra of “developers, developers, developers” is to ring true in a low cost Web 2.0 world then MS will have to bite the bullet and make a huge swathe of stuff available for free and that includes clusterable versions of MSSQL and Windows 2003 maybe up to a certain traffic/transaction level after which you have to pay. Free Dev Tools and dev versions of OS/DB won’t cut it any more.

  4. The licence cost isn’t, and shouldn’t be, the only factor in choosing your technology set.

    You’ve got to factor in your own existing expertise and/or any codebase you might have.

    You also have to assess whether you’ll be able to source developers experienced with the tecnologies you want to use in the future – (for example) Rails might seem like a good idea now, but will you be able to easily source top-drawer Rails developers as and when you need them?

    Ultimately if commercial technologies fit better, you’d be an idiot to ignore them simply to avoid upfront licence costs.

  5. I suspect yer man’s talking out of his arse, to be honest.

    A group which “quite literally can’t keep up with demand” is a group which

    a) doesn’t fill a person with confidence about its ability to do anything else for me later and

    b) *demonstrably* does *not* need more “evangelists” to ramp up demand.

    That said, if there are new companies using MS stuff, it’s possibly because everyone in Ireland who uses anything else has either emigrated or started her own company.

    Try doing a job search on any Irish board for ‘PHP’ outside of Dublin and you’ll see what I mean. This morning’s result for Co. Cork: one technical writer and one misclassified ASP job. *Every other listing in Munster* is some muppet at a recruitment agency copy-and-pasting the same Microsoft shopping list he uses on everything else.

    The IT job market here is a train wreck. Unless, of course, you have “Five years’ AJAX experience and 10 years of .NET”.

  6. in terms of open source, we’ve not only gone the php/mysql route, but are also using in place of having to pay for ms office. we are using other non-open source software, but in general have a preference for open source at this stage, so long as it is competitive on a performance level.

  7. I’m not surprised that startups are still using anything-but-open-source. It’s what pointy haired bosses know about. It’s what they hear in the adverts, it’s familiar. Microsoft software runs their desktops and laptops so it follows that the same company should provide the server software. “Nobody got fired for buying ..”

    Enlightened bosses will buy into Free Software, but there are many more who are happy to shell out year after year for infrastructure software that’s available for free elsewhere.

  8. It’s not always as simple as that Donncha.

    Taking Microsoft Office, many businesses will be running custom Access applications, VBA scripts, etc. In those cases, it may prove more expensive in the short- to medium-term to move away from MS Office, rather than staying with it. Microsoft are not exactly forcing you to upgrade to a newer version of Office either – Office 97 works fine on Windows 98, 2000, and XP.

    There’s also the simple truth that while free desktops such as Gnome might be getting better, Windows is still superior at this time (at least for general use).

    Is anyone here suggesting free/open solutions even factoring the cost of re-training in? A typical business will consist of more than just the bearded gnomes…

  9. The initial cost isn’t the whole cost. Training, familiarity etc all go towards the Total Cost of Ownership.

    Assume for a minute you start up a new company and it reasonably IT related. Let’s say oh I don’t know emmmmmmm a data centre.

    And your going to have staff. So your initial reaction as a reasonably techie type person is open source the lot. Management software, monitoring software, network analysis software, open office for everyone running on some form of Linux.

    Then bit by bit your staff increases. Your receptionist/typist can ALT key her way through MS Word with her eyes closed but now she’s back to basics figuring out how to tabulate a page.

    Meanwhile your data centre has grown, supporting the open source free network tools is getting time consuming, it’d be nice to be able to ring someone and say “Get this sorted out or else!�

    Your list of clients has grown exponentially and you need to automate as many processes as possible but you discover that automating your apps is a brave new world because only a handful of people in the world have used this app in anger.

    Your techies are using a variety of email/Scheduling clients (checkout Google’s internal mail client policy to see this taken to madness) and scheduling a meeting automatically is almost impossible.

    In the midst of this you now have more non techy gurus working for you than gurus. Slowly but surely the cost of maintaining the various free software applications far exceeds the purchase cost.

    As your company has expanded more and more of the decisions on software purchases aren’t being made by you but managers. Managers who wonder what happens if I go down the route of a new application that is free but may not work/be supported/Used by anyone else in the world. On the other hand if they purchase application B that is used worldwide/supported/known to work (Kind of) which option do you think will give the manager the most job security?

    And now you are staring down the barrel of migrating data, apps and users to commercial software. Fun fun fun.

  10. Purchase is not use. I have the utmost respect for Robert but I think he falls into a regular Microsoft trap of measuring the number of ISV packs (ActionPack and Empower Licenses) they are distributing in Ireland.

    Now these are very good deals where you get a whole pack of Microsoft business applications (ActionPack) or ISV applications (Empower) at a knock down price (around 350 euro). We are an ActionPack subscriber but we develop in Python and our database in Postgres.

    I suspect if you scratched the surface of a lot of ISVs you’d find the same story. The economics are inescapable especially if you conside that the minute you move to web hosting the database of choice is MySQL.

    And of course its free, works with Ruby, Rails, Python, Java, PHP etc. etc. with zero effort on the developers part.

  11. Multiple attempts to post a reply here are failing Tom. Either I get a blank screen, a “duplicate detected” message or nothing happens. Do comments have a length limitation?

  12. Sixth attempt to post this reply:

    I think the conversation has extended in scope beyond Tom’s original point where he was really talking about implementation technologies for web application startups.

    I’m very pro Open Source and our main server is a Fedora box, I have Kubuntu on an old laptop but I run XP on my desktop and main laptop along with Office. I tried the pure Openoffice route (and still do a bit for internal docs) but as long as you are exchanging documents with others, the lack of 100% compatibility will always be a problem. I got tired of constantly reformatting docs and spreadsheets due to the small glitches. PDF only works in a read-only sharing scenario.

    Despite building our Web App using LAMP, I’ll be happily buying Vista when it comes out, it looks gorgeous and I was impressed with the RC running on VMWare.

    Recent release of Ubuntu and Fedora have been riddled with desktop usability bugs which are not getting fixed from release to release. I’m very close to putting Win2k back on the old laptop due to the Kubuntu problems and I generally only use Fedora for server functionality.

  13. I work at a (US) startup where we spun up the entire corporate infrastructure with a completely opensource stack (linux/openbsd, lamp portal stack, sso with openldap + kerberos, etc) with off the shelf intel hardware. we eventually wound up migrating pieces of the web side to java/j2ee, but again with opensource components (spring, tomcat, hibernate, etc.) The only thing we spent money on was the networking infrastructure (cisco/netscreen), although in retrospect we could have probably used openbsd and pf. All in all, i think it cost somewhere around $20k-30k to get the entire infrastructure up in over the course of 4-5 months. It has been up for 4 years now with minimal issues, outages, etc. we’re supporting internal and external portals with about 1000 users, processing somewhere around 5-10k emails a day, managing a few hundred mailing lists, etc. so the environment handles a fair deal of traffic.

    management made a (non-technical/political) decision last year to switch platforms to .NET/MS. We’ve probably spent, in the last 12 months, somewhere around $300-500k (if not more) in hardware/licensing/support. we have weekly issues with sharepoint, had to retrofit large amounts of ldap code to talk to AD because of subtle microsoft incompatibilties, have to maintain 2 different development frameworks (.NET 1.0 because sharepoint doesnt support .NET 2.0), etc. its essentially a nightmare.

    morale of the story: we’re currently switching platforms back to OSS and Java.

  14. Our group at Microsoft Ireland can, quite literally, not adequately keep up with the demand we get from local startups

    …doesn’t mean that more people are using MySQL over MSSQL. In fact from that qualitative information alone you can’t even derive whether MSSQL has significant traction in the market place.

    Clearly there are many other factors at play here – such as how many employees the MS DB Ireland group have, how many customers their support, which business sectors, etc.

    Also go to any trade show and you’ll see competing vendors all claim they have the leading product – with varying stats to ‘proove’ it.

    If you want to test your hypothesis why not carry out a vote or something on your blog instead?

  15. This is classic FUD. Had this been a rational discussion, he probably would have told you “We have sold in excess of…” or “we sell this many licenses…”

    If you consider how foolish the statement is: ‘We can hardly keep up with the demand’… they are just selling licenses! How hard is that?

  16. I just posted a reply and it seems to have been lost into the ether when I hit submit… sorry folks, really frustrating but have to go now and will respond later…

  17. OK here goes — John H, with respect, please see the job postings here, under “Development”:

    And to your further point about being concerned about our sustained engagement, I think you misunderstand the purpose and goals of the team I’m a part of (Developer and Platform Evangelism Group), and how we fit into the larger Irish subsidiary (where we have groups that perform deeper and longer-term engagement with our partners), and of course Microsoft as a whole.

    However, Ben M, you are also correct, my role, and consequently what I was referrihng to in the comment of the previous post, is not just focused on the data tier. Being able to work more deeply with startups and ISVs across all the Microsoft developer and platform technologies is a major factor in why we are hiring new members of the team.

    Back to databases: Here at Tech Ed we have seen both Oracle and SQL well represented on the conference floor, and as you point out, every crowd has varying stats to ‘prove’ their points. For those who are only comparing SQL against open source alternatives, keep in mind that there are actually quite a few database vendors who believe (with good reason IMHO) that they are adding value in many business scenarios – including startup scenarios.

    I am very interested to hear your thoughts on why or why not you wouldn’t pay for a database. I would think that the data tier is perhaps where it’s most obvious that total cost of ownership, reliability, scalability, availability and (importantly) manageability are all critical.

    Of course, on the other hand, can think of LOTS of projects I wouldn’t pay for a database for — but for those, I use SQL Server Express, since I love its tight integration with Visual Studio Express, and the fact I could always scale up to full SQL if any of my ideas were (miraculously) to catch on…

  18. Hopefully we can avoid getting into the usual Slashdot pis**ng contest about MS and focus on the specifics of using SQL Server in a startup.

    As I said over at Rob’s blog, I think SQL Server 2000 is a great product and in an enterprise environment I always pushed it to customers instead of Oracle which I found incredibly difficult to manage. We did have developers who always wanted to go the MySQL route but the reality is that 95% of Fortune 100 companies would not countenance it for mission critical apps. It doesn’t matter if they were right or wrong, they signed the cheques.

    But now I’m in a startup, I can look at MySQL and Apache and then look at Flickr and and then read Cal Henderson’s book and know that I am making a good business decision to go that route.

    Rather than vague generalities, I would love to see the pricing on a co-lo setup using Windows 2003 + SQL Server 2005 + IIS for a simple front-end/back-end two box solution and then look at the same for Fedora + MySQL + Apache.

    I’d be perfectly happy to use MS technologies in what we are building but until they address the core pricing issue for early stage startups, I think they have a big problem. I don’t mean cheap or feature limited, I mean enterprise-level and free. Not in some sort of happy-clappy all software should be free sense but for solid long-term business and mindshare reasons.

    Could they maybe start a partner programme where a start-up gets everything for free until they are cash-flow positive? Sure they’ll lose out on licensing fees but they’ll acquire thousands of more developers who will default to using MS technologies in every new project they encounter.

    And for the successful ones, the upside for Microsoft is huge. I wonder how much MySpace are handing over each month? Imagine if MS had given a ton of free licences to YouTube in 2005? The second Google acquired them, they would have owed MS a pretty pot of money.

  19. Well, first off, comparing MySQL and MS SQL server is really not comparing like with like. PostgreSQL and MsSQL might be closer to the mark. For many things, MySQL is featureful enough, of course, and it’s definitely fast.

    I would suspect a lot of the startups using MS SQL server aren’t particularly high-tech startups. A lot of people starting in the industry may have experience with MS SQL from MSCE and college (do any colleges use it?) and the cost of the license may be as nothing compared to the cost of getting someone who’s any good at an alternative database.

  20. I’d say what Microsoft claims is probably right.

    Why is this? Well, basically, MS is great at developer education, or at least, it is better than anybody else. There is something of a clear system for development which suits the needs of SMEs, at least to some extent. The OSS platforms just don’t have that.

    There are loads of developers who do Microsoft stuff, not so many who do the OSS stuff. At the end of the day, developers are better going with what they know.

    Personally, if it was my company, I would go OSS or maybe OS-X, but I know that loads of people wouldn’t.

  21. Rob, I think you are falling into the classic (old-school?) Microsoft mind-set trap which compares products based on features. As long as MySQL is taking business from MSSQL then they are directly comparable from a market perspective. The “but we have a better product” approach has killed many many big companies over the years.

    I do take your point about non-high-tech startups tho.

  22. No, no, I’m not saying that MS SQL is ‘better’ than MySQL, simply that it is a different type of product. There are product areas which MySQL is just completely unsuited to due to lack or immaturity of features (especially transactions, views, stored procedures, etc.), and there are areas to which it’s more suited than MS SQL or Oracle (speed-critical, embedded, etc.). Certainly, there is a large overlap area, but they’re not interchangeable.

  23. Actually, I am saying that MS SQL is a better product in most scenarios but I’m also saying that it is irrelevant in many cases. And the challenge for Microsoft is to address the challenge of MySQL on different terms than just features.

  24. It isn’t horrific for large databases with lots of queries, either; just ask Google/YouTube/Facebook etc. etc. etc.

    MSSQL used to have a massive advantage over it, but that has definitely dropped off in recent years.

Comments are closed.