back to article Don't believe the IT hype: Ye cannae change the laws of physics

It's fun to be on the receiving end of IT advertising. The vendor's ads start by promising to solve your business problems better than the competition can, and then the superlatives begin to snowball until an answer to global warming and a solution for war in Iraq are both in there among the plug-ins you can buy to make your …

COMMENTS

This topic is closed for new posts.
  1. Velv

    You are of course 100% correct. In most cases the Open Source option performs just as well.

    But when it breaks (and they all do), somebody is going to get kicked! But who?

    If you've paid Oracle et al for that support along with the product, you get to kick them. If you pay for support for the Open Source, you get to kick the support supplier. If you didn't pay for support, time to slide a tea tray down your pants.

    I'm not saying its your fault, I'm saying you're getting the blame. Open Source can be cheaper - but never view it as free.

    1. Anonymous Coward
      Anonymous Coward

      What management is really paying for ...

      ...is someone to blame.

      I've no experience of Oracle support, but my experience of IBM both as a customer and within the organisation was that unless you're a very important/rich customer you waste more time explaining the bug over and over again through various levels of support than it would take to just code around the problem.

      When you finally get acknowledgement that the bug is really a bug, you then have to implement the workaround anyway because you can't afford to wait for it to be fixed.

      Consequently, since I moved to mostly open source stacks I'm more productive. From my experience of Hibernate support a few years ago, at least the developers would tell you to fuck off immediately rather than waste your time.

      1. K

        Re: What management is really paying for ...

        Glad I'm not the only one that found this!

        I found a bug in on a Layer 2-7 Switch for the IBM blade chassis, fortunately I had a contact into the development team who pushed out a fix within 48 hrs, amazing.. but it then took IBM 7 months to make this fix publicly available.

        1. John Smith 19 Gold badge
          Unhappy

          Re: What management is really paying for ...

          "I found a bug in on a Layer 2-7 Switch for the IBM blade chassis, fortunately I had a contact into the development team who pushed out a fix within 48 hrs, amazing.. but it then took IBM 7 months to make this fix publicly available."

          This was presumably for a non contentious fix (IE everybody recognizes it's a bug, no skin off IBM to fix it, no serious logistical problem, no liability to admit to etc).

          Imagine if it was not?

          And BTW I doubt many large software companies are any different in this regard IE they could be exposed to lawsuits.

          1. K

            Re: What management is really paying for ...

            They are Application aware switches (i.e. they load balance etc) and the bug was in the handling SSL sessions and failure to release them, so eventually the switched overloaded and crawled to a stop. Given its with SSL, I'd be very suprised if others didn't experience it. Saying that, these switches are an optional extra, and give the £12k price tag for each one, I doubt many people brought them.

    2. Anonymous Coward
      Anonymous Coward

      Agreed

      Different domain but similar situation. Customer has bought Google search appliances which have worked rocked solid since purchase and are, according the team involved nae problem to configure and update.

      Another department decided in its wisdom that what everyone needed was Lucene, company-wide Google licences already bought and paid for notwithstanding. No idea what the customisations and stuff have costed but I have witnessed the search go down, unnoticed for long periods of time. Now Lucene is a fine search engine and the results are scarily close to what Google produces. But if stories like this are why managers insist on buying brands. Even if they fail it won't reflect poorly on them.

      I happen to use a lot and develop some open source myself and am a big supporter but too many people think too much only about the cost of licences.

    3. Vic

      > Open Source can be cheaper - but never view it as free.

      "free" is irrelevant outside a quick lash-up to see if the product suits your needs.

      "Free" is *far* more important, and is the reason we promote FOSS...

      Vic.

  2. John Miles 1

    Can afford to do focussed solutions with Open Source

    I like the reminder to choose a good algorithm before throwing hardware/systems at it.

    Another problem with expensive commercial solutions is that because they cost so much they have to be justified by applying them to a wide variety of requirements - and then they struggle to fulfill the one you are interested in. Whereas if there is minimal cost in deploying an open source solution it can be focused on addressing a particular requirement => Smaller, simpler, faster and more reliable system.

    Of course, there is the resultant risk/problem of multiple fragmented solutions. But providing they are not too disparate, several separate solutions that reliably do their own job may be preferable to one monster that never quite delivers.

  3. This post has been deleted by its author

  4. Anonymous Coward
    Anonymous Coward

    Estate Agents

    I built a DB of UK Estate Agents and did it basically the same way in PHP/MySQL and the result outperformed the PO's closed-source "solution" by enough that it mattered in the late 90's. The problem of how to allow for water (eg, the Bristol Channel) was much harder than the problem of building the initial DB search thanks to the glories of the National Grid and the fact that the curvature of the earth wasn't really an issue on that scale.

    The alternative of using closed source software simply wasn't viable as what we wanted to do was very niche at the time and we couldn't afford to pay the Post Office enough to make them care.

    1. g e

      Re: Estate Agents

      A lot of the time Eastings and Northings are good enough, too, given that they're commonly provided in UK data. Keep your maths in the integer domain, especially if a box search is good enough and you don't need to compute a real radial distance.

    2. Corinne

      Re: Estate Agents

      "The problem of how to allow for water (eg, the Bristol Channel)...."

      Tell me about it! I live right on the Thames Estuary - most of the time I'm after shops in Kent, but get told that my nearest is in Essex. As the crow (or seagull round here) flies it may be closer, but by the time I've got all the way up to the Dartford Cossing (most easterly Thames crossing), paid the hefty toll each way & driven back east it's a LOT further away than it seems.

      1. Michael Wojcik Silver badge

        Re: Estate Agents

        As the crow (or seagull round here) flies it may be closer, but by the time I've got all the way up to the Dartford Cossing (most easterly Thames crossing), paid the hefty toll each way & driven back east it's a LOT further away than it seems.

        Surely the solution is an amphibious car. Or perhaps a dirigible.

        Or you could spread the workload across more resources. Hire a man in a van on the other side of the estuary to go to the shop and then drive to the bank, then another in a boat to take the goods across the water, while you drive to the bank on your side. Parallelism!

        And we can generalize this so it handles all shopping: for anything you want to purchase in Kent, hire someone to drop it off at some point on your bank of the estuary, thus reducing the problem to one already solved.

  5. This post has been deleted by its author

  6. Nick Ryan Silver badge

    Most problems like this are down to using inefficient data formats. Longitude and latitude are good for many purposes but there are other (I'm under NDA on this, just Google them!) methods of storing spatial co-ordinates other than cartesian coordinates and some of these make it much easier to locate the closest N records out of a large dataset.

    It still does often come down to the required level of accuracy, smart filtering such in the boundary box of the example given in the article and other manually prescribed optimisations.

    1. John Smith 19 Gold badge
      Happy

      "Most problems like this are down to using inefficient data formats. Longitude and latitude are good for many purposes but there are other (I'm under NDA on this, just Google them!) methods of storing spatial co-ordinates other than cartesian coordinates and some of these make it much easier to locate the closest N records out of a large dataset."

      Could it begin with a V?

  7. J.G.Harston Silver badge

    If you're only going out to 50 miles, don't bother with great circles, just use pythagoras. Assuming lat/long grids are rectangular works accurately enough for a 50-mile distance up to about 75 degrees from the equator.

  8. Anonymous Coward
    Anonymous Coward

    Just wondering

    Mr. Cartwright, you have heard of spatial indices, have you?

    http://www.cs.bgu.ac.il/~atdb082/wiki.files/paper6.pdf (one of the seminal papers in the field).

    Implemented by most if not all major DBMS these days to more than acceptable levels of performance. If we are talking about a travelling salesman solution (as opposed to straight line distances) that can also be arranged by using the proper tools.

    I am sorry, but your example came across as so badly chosen that I could not follow your point.

    As for your other remark regarding choosing the commercial option « if you want tens of millions of rows and super-fast, concurrent processing », may I point out that this is also a bit of an unfortunate generalisation? There comes a point when maintaining your own (possibly FOSS-based) solution makes more sense that going for a "commercial" (i.e., off-the-shelf) offering, as the likes of Google, Twitter, Facebook, Amazon, etc., etc., can attest.

    1. Don Jefe
      Happy

      Re: Just wondering

      Agree with your point on the not so wonderful example. There are far better ways to deal with that scenario.

      However, you pretty much named all (except finance) the companies that really need a insanely huge database. Most organizations simply can't justify the cost of building out something "perfect" and maintaining it when a COTS solution gets the job done.

    2. Steve Knox

      Re: Just wondering

      Spatial indices was my first thought as well. But then I read the rest of his analogy without prejudice.

      Turns out that the analogy is still apt, because his entire point is that this problem isn't necessarily solved any "better" with spatial indices. Instead, you've just shuffled the work around to different points in time and storage (i.e, the spatial indices have to be created and maintained [time], and stored in the database [storage]).

      Now, if your particular system is amenable to slower inserts and more storage space in return for faster queries, spatial indices are definitely an option. But if your inserts are time-constrained, or your storage budget has disappeared (or you can't afford the licensing fees for a DBMS with spatial indices), then you have to roll your own.

    3. Anonymous Coward
      Anonymous Coward

      Re: Just wondering

      No kidding. And you don't need a special name like "spatial index" or a paper; find the lat/lon of points 5 miles west and north of the center; query for places with that range of lat/lons. Filter the result set if you really need it to be a circle. But honestly, this is not geodetics; you just want places close in a general sense. (What you *really* want is places within 5 minutes by driving/taxi/public transit, and that is harder.)

      Often in IT, "The laws of physics" are just "The limits of management's imagination."

  9. Anonymous Coward
    Anonymous Coward

    SQL Server 2008+ has spatial indexes and data types. Trying to roll your own seems pointless.

    1. Lee D Silver badge

      How much is SQL Server + CAL's for a "variety of travel companies" (what size, how many have to do this query, do they even USE a Windows database at all)? How many times are you going to do this? How often? Over how large a dataset? For what purpose (A one-off mailing? A regular five-minute security check? Every database entry?) What's the impact of turning on something that someone (maybe not even a professional DB admin but just an IT admin) isn't familiar with? Do all the versions and databases in use across the organisation support it? How do you test that SQL Server 2008 would benefit you if you don't have that software anywhere else? How much will it save you over the home-spun solution? And where does it mention that they have (or could afford) that version of SQL Server, or even a Windows desktop?

      Just what, precisely, leads you to the conclusion that your suggestion would always be better for everyone? This is kind of the part of the article. The "bodge" works everywhere, works for everyone, and doesn't need extra licences, support, staff, skills, etc.

      Blinkered "nobody can do anything other than the Microsoft way" IT is one of my bugbears.

      1. BlueGreen

        let me try

        > How much is SQL Server + CAL's for a "variety of travel companies"

        about £25,000 (or maybe it's $25,000) per physical socket (regardless of cores), so for some, relatively cheap

        > How many times are you going to do this? How often? Over how large a dataset? For what purpose (A one-off mailing? A regular five-minute security check? Every database entry?)

        that's a business decision.

        > What's the impact of turning on something that someone (maybe not even a professional DB admin but just an IT admin) isn't familiar with?

        Same as any other index. None. You don't want them, don't create them, and if they're there but unneeded, they have not impact except space - and if you do need them, they are there.

        > Do all the versions and databases in use across the organisation support it?

        being a db, it's a centralised point of access. No-one else needs to run it, just access it. Can be done through a web browser and be totally platform independent if you want.

        > How do you test that SQL Server 2008 would benefit you if you don't have that software anywhere else?

        Not sure what you're saying, but you can get an eval license for next to nothing.

        > How much will it save you over the home-spun solution?

        That's again a business decision. In our case, the difference betwen great pain and no pain (see my comment below to Gio Ciampa about db sizes)

        > And where does it mention that they have (or could afford) that version of SQL Server, or even a Windows desktop?

        then use postgres, as others have said it has geo indexes and it's free (disclaimer - I've not used it yet)

        > This is kind of the part of the article. The "bodge" works everywhere, works for everyone, and doesn't need extra licences, support, staff, skills, etc.

        bloody hell! no it doesn't. It's a tradeoff; a business decision (as influenced by technical factors). It can be very, very expensive to do it the 'bodge' way; far more than an off the peg package. That is an astonishingly arrogant attitude and potentially damaging to an employer.

        > Blinkered "nobody can do anything other than the Microsoft way" IT is one of my bugbears.

        The blinkering is yours, if you're unaware of other commercial and free options.

        (and BTW this may sound very upbeat about mssql2008 but it's not. I don't like it any more)

    2. Anonymous Coward
      Anonymous Coward

      « SQL Server 2008+ has spatial indexes and data types. Trying to roll your own seems pointless. »

      Same as most mainstream RDBMS these days, commercial (Oracle, SQL Server), open source (Postgres [had them for decades], MySQL and derivatives), and public domain (SQLite). There are even standard format representations (WKT, WKB) to provide some degree of interoperability across DB systems.

  10. Morten

    No problem with DB2 ...

    Use DB2 Spacial Extender.. http://www-01.ibm.com/software/data/spatial/db2spatial/

    :)

  11. Phil O'Sophical Silver badge
    Pint

    WIne

    Just to add to the pedantic nitpicking, allowing for fixed costs such as the bottle, shipping and duty, a £9.00 bottle of wine probably contains £4 worth of wine, and a £90 bottle likely contains £80 worth of wine, so the comparison is 20x, not 10x. It's not linear.

    Probably not relevant at that level, but it does explain why a £10 bottle is much more than twice as good as a £5 one, and your guests will notice that.

    </pedant>

    1. Anonymous Coward
      Anonymous Coward

      Re: WIne

      At a dinner party about 20 years ago there were two bottles labelled "A" and "B". They were both judged by the guests as tasting pretty good. The "expert" also correctly identified one as being a "good" example of Saint-Julien. Some people felt that the other wine was marginally better. That one was actually a Bulgarian native grape.

      Both wines had been bought at our local up market wine shop. The Saint-Julien cost £27 - the Bulgarian was £2.50 .

      At that time Eastern European wines were obviously sold at lower prices than their quality deserved - however the price gap was incredible.

      I've never paid that sort of money again for the cachet of a French region.

    2. Anonymous Coward
      Anonymous Coward

      Re: WIne

      <pedant>your markup is badly formed as you have omitted the start tag</pedant>

      1. Phil O'Sophical Silver badge

        Re: WIne

        I'm in pedant mode by default...

  12. Gio Ciampa

    "The best you can do is find a cunning way to crunch the numbers faster - by disregarding bits of data that you know aren't relevant, for instance, or by using a parallel-processing system, or by using a heuristic algorithm that gives a slightly less (but still acceptably) accurate solution with a significantly smaller amount of work."

    Or..

    You could use a different database paradigm entirely - drop the row/column stuff and use sparse arrays where the coordinates are the subscripts, and the search (assuming the longitude/latitude restrictions are still being used) becomes a matter of:

    (apologies for the '.'s - quickest way I could think of to get some manner of indentation)

    set long=startLong

    for {

    ....set long=$next(destinationList(long))

    ....if long>endLong quit

    ....set lang=startLang

    ....for {

    ........set lang=$next(destinationList(long,lang))

    ........if lang>endLang quit

    ........if $$howfar(long,lang,userLong,userLang)<maxDistance {

    ............set resultList(distance)=destinationList(long,lang)

    ........}

    ....}

    }

    then just go through the resultList array, picking out as many as you need... self indexing, simpler, quicker, and a whole long easier to debug...

    1. Gio Ciampa
      FAIL

      oops... should have set distance to the result of the $$howfar() function... ah well... the joys of knocking out bits of code in 2 minutes flat...

    2. BlueGreen

      @Gio Ciampa

      I've half a billion entries in the db I'm babysitting. How well would your method deal with that?

      1. Gio Ciampa

        Re: @Gio Ciampa

        I'll admit the code was written to fit the hotel example, so is quite specific to that. You can play "spot the language" if you like - the code would work as written (bar the dumb error as mentioned...)

        My point was that the rigid adherence to the row/column table paradigm isn't necessariy the best way to go - best tool for the job and all that...

    3. Hungry Sean
      Holmes

      your suggestion implies a sort on the longitude coordinate (i.e. it serves as an index into your query) and also does a bounding on the longitude. You'll find that the method described in the article is substantially similar though more efficient. I'd suggest you read a decent background article about DBs-- might be useful to you someday :-).

  13. Lee D Silver badge

    What's good enough for you?

    And the rule that so many people miss is:

    What's good enough for you?

    The answer varies wildly and in the same way that a county of libraries and schools don't all need expensive Cisco switches on every site, you often don't need too much complexity, or top-end software or huge expensive products to get a job done.

    I get asked by people who work in IT (but not schools) why the schools I've worked for don't "just run Exchange", for example. They could. But they don't need it. So they don't have it. Their email is covered. Their calendaring is non-existent or very specialist. The other features they wouldn't understand, let alone use. And by the time you set it all up, what you've actually "saved" (in terms of some poor sod not having to make up a calendar or appointments once in a blue moon) is actually chicken-feed next to the overall cost of implementation.

    There are some things where you can just say "Why?" and do it yourself quicker or cheaper than the original solutions because you just don't know if it will work, be used, have value, etc. Sometimes, yes, they expand and you find yourself having to migrate to a bigger product anyway, but a lot of the time the system deployed is perfectly adequate or nobody even uses it and thus it's cheaper to keep it running and/or throw it in the bin.

    I put a Linux server into some of my schools - the initial reason often differs but I've put them in for everything from web-caching (overloaded external lines due to no decent on-site caching), web-filtering, or even just things like internal web-serving without having to deploy IIS.

    Sometimes they weren't sure about it, but the reason it was put in varied and I clearly stated "this is a test" - a free test to see if the feature they demand is what they want, works how they want, can be done more cheaply than existing products, will be used, etc. They are free to say it's inadequate if they can show it is (and, hell, then I'll be the first to rip it out and replace it with something else because that means PROGRESS).

    And in more schools that not, years later the "free test" server (usually some old office machine or spare server) is still running and doing the job it was deployed for AND MORE. Because it ends up being just that useful to warrant its continued existence but not unsuitable enough to warrant changing it for a piece of hardware dedicated to that purpose with support, etc. Hell, in one school I was given the job of making their "network go faster" and in the first week tripled their Internet speeds (measured by the Head using Speedtest, because he was obsessed with it) in the middle of the day just by some transparent proxy/caching. If the box broke, or there was a problem, I demonstrated how taking the "in" Ethernet and the "out" Ethernet out of the machine, throwing the machine away, and joining them with a connector made everyone else carry on working exactly how they always had (and, thus, we could just throw it away if it was a problem).

    When I upgraded the machine three years later (without a single day of downtime), the connector was still in the same "dust shadow" on the top of the machine. It was even "good enough" to deploy more services and more things on it - some of them got used, some of them didn't, some of them were useful, some of them weren't, some of them were better than a commercial equivalent, some weren't.

    Anyone who just spends a fortune on something, deploys it, and doesn't do small scale tests to see if they even NEEDED it first isn't really doing their job. The best way is to build an in-house, temporary solution and see how far you get. If you can't get the results that you DEFINITELY get with a better product, buy the product - you've proved it's worth having. If you get results similar to it or better, why bother to change from in-house? You've proved your own solution is better.

    "Good enough". It's said like it's a bad thing, if you say it wrong, but it means the exact opposite. I wish my bank account were "good enough" for the things I wanted to do. My current school has a Mediawiki Intranet. It's "good enough" that it hasn't changed in years, and we haven't bothered to trial any alternatives. It has a IT ticket system. It's "good enough" that it hasn't changed in years, and we haven't bothered to trial any alternatives. It runs all sorts of things that we would have had to buy, but a quick test of something in-house, custom-made or just open-source etc. was often "good enough" that we don't bother to go any further.

    Sure, there are advantages depending on your scale and business to having outside support, or being able to hire people instantly familiar with the products you're using, etc. But for most stuff, an in-house or even a cheaper solution is "good enough", so why spend money on something that won't provide tangible benefits over that "good enough" solution? Even if you have to throw it out when the guy leaves and put in something else because you can't replicate his solution, surely that's saved you years of licensing and support that you would otherwise have had to pay for? And if his solution wasn't "good enough" while he was there, you should have been replacing it/him anyway.

    Most small businesses do not need Oracle, Cisco or even Microsoft. They just don't. You can point out flaws, support issues or whatever else but the fact of the matter is if they are happy and the system is good enough for them, why not use it? Why spend money you don't have to? If they have to take contingency and staffing skills into account, they will, it still doesn't mean it's not better to have some solution for X amount of years before they determine they NEED something more. It's just saved them X years of support, licensing, maintenance, training, etc.

    It's a disease rife within the upper echelons of IT where people who deal with millions of desktops tell people who have a few hundred what they "should" be doing. You can normally spot them by the way they have no consideration of buying extra licenses or vast systems. Some things just don't scale. And a bodge in an SQL script to save hours of database time for a one-off (or even only occasional) query? Yeah, I'll take that bodge, please. If it works adequately, I have no need to do anything else. If it doesn't, well I've cut out some of the work in the meantime and will have to do EXACTLY what I would have needed to do anyway - go find a better solution. Even if it only works for a week - that's an extra week I can spend researching and testing alternatives.

    Guess what? I object to paying for an SQL Server license and CAL's just to run a piece of access control software that stores a few hundred entries a day of an ID card number and door number, or something fires up the school boilers at 6am each day. EVEN if I already have an SQL Server doing other things. And especially when they have the option to use a MySQL database or whatever instead (currently have one piece of access control software that uses FireBird as an internal SQL server if you don't have one of your own - works perfectly and with the right software you can query it like it was any other remote database using standard SQL).

    I have several MySQL instances for various things, I have cron scripts giving my scanners a helping hand to archive scanned documents (the scanners don't have such functionality, so I have them save scans to a server, which copies them to the right destination and archives them for my purposes too - sure we have one huge expensive photocopier that's capable of doing that itself too, but it's easier to tie it into the same system that even the featureless scanners can benefit from), I have fax-to-email gateways run by scripts on a Linux server (when I could just buy a proper fax add-on for the telephony system we have, but that's expensive and a USB-56k fax modem was much cheaper and more powerful, even with the software setup), I have all sorts of similar time-saving, money-saving and effort-saving tricks running.

    Good enough for myself, my employers and my purposes might not be good enough for you. But that's not your decision. It might have flaws, but they might not affect my usage. And it may not be ideal, but that depends on how much money and in-house support rather than journeys through phone menus contributes to your vision of "ideal".

    And, at worst, I have to do what you'd like me to anyway. If it buys me a year of having to do that, that's a year spent on other things (both intellectually and financially) - maybe I spend the money saved on the "proper" solution for something much more vital and expensive, for instance!

    The only question is "What is good enough?".

  14. Bill Mew

    Only so much that can be gained by refining one technology until the next revolution occurs

    I agree with your assertion that a given technology can only be refined so far - propeller driven aircraft were refined and refined with ever diminishing returns, until of course the jet engine revolutionised what was possible.

    Your article provides a great explanation of the "law of Physics" limitations of relational databases, but the laws can be changed with a genuinely new technology.

    The "cognitive systems era" and IBM's Watson is just such a new technology and Watson is entering its first wave of commercialization - to quote El Reg

    http://www.theregister.co.uk/2013/03/01/ibm_ceo_rometty_big_data/page2.html

    "IBM is looking ahead to a "cognitive systems era" in which machine learning and multi-modal decision support systems help banks and investors make better financial decisions, help doctors make better diagnoses, help lawyers make their arguments (rather than just root around like a search engine in briefs for precedence), and help companies do telemarketing by adding computer-assisted persuasion to their call centers."

    This 'Jet engine" may only be entering its first wave of commercialization, but the revolution is coming and the "law of Physics" limitations of relational databases will surely be overcome.

    Bill

    1. Anonymous Coward
      Anonymous Coward

      Re: Only so much that can be gained by refining one technology until the next revolution occurs

      'Your article provides a great explanation of the "law of Physics" limitations of relational databases, but the laws can be changed with a genuinely new technology.'

      Ye cannae' change the Laws of Physics! that's why they're laws, and not just guidelines!

      The jet engine didn't change a single law of physics, and works using exactly the same laws of thermodynamics and fluid dynamics as a propeller engine. What can change is our understanding of the laws of physics, which might enable us to exploit them better; but limits are part of understanding.

      This might sound pedantic, but there's so much woolly thinking about science in current society that even educated people seem to be clueless about "the laws of physics" and what they mean.

  15. Mick Sheppard

    Its all about support and disaster recovery

    The performance difference between different database engines is so slight that a poorly designed schema can destroy the benefit of your expensive Oracle or DB2 installation. Where they shine however is when things go wrong. From the simple recovery of lost database files due to file corruption, or user error, to reliable replication to DR sites, to real technical support.

    Its fine to invest the time and effort in FOSS database solutions but ultimately the business needs to accept the risk of not having anyone to batphone when its all gone wrong and business continuity is threatened. Its this that keeps Oracle, IBM, and MS up there are DB vendors.

    1. Vic

      Re: Its all about support and disaster recovery

      > ultimately the business needs to accept the risk of not having anyone to batphone

      No it doesn't.

      You want someone to phone? Support contracts are readily available. 4-hour response with 24-hour coverage? Not a problem.

      What you won't get is guaranteed SLAs for zero cash. So you run the numbers - with FOSS products, I can undercut the big commercial players all the time. I just had to ramp up one of my quotes because I was two full orders of magnitude cheaper than HP, and I figured they wouldn't take me seriously if I quoted what would make me just a healthy profit.

      Vic.

  16. Charlie Clark Silver badge

    Spatial indices on MySQL?

    Must have missed it. PostGIS is the way to go if you want to do this sort of thing with open source. I'm getting increasingly sick of comparisons of MySQL (in whichever flavour) with real RDBMS's. The hacks and kludges are just too painful to endure for anything other than toy projects.

    By all means get a support contract and pay for DBAs and programmers who know their shit.

  17. Anonymous Coward
    Anonymous Coward

    Reminds me of an analysis tool I wrote. It outperformed market leaders in both speed and functionality - and could be enhanced, or bug-fixed, in a matter of hours.

    However several people in the Company wouldn't use it - even though it was free. They bought market leader commercial products with lesser capabilities - and bug-fixing only in a future release.

    The reason they bought the commercial product? It looked good on their CV - and they would tell their boss it was as good as such a thing could be. That was not strictly true. It was as good as any of the commercial products - but could not handle the complex problems that were grist to the mill for the in-house tool.

  18. All names Taken
    Paris Hilton

    Que?

    "But is a £90 bottle of wine 10 times nicer"

    Doesn't really matter.

    Perhaps a £9 bottle is a day to day purchase a bit like teabags or instant coffee but £90 is a statement of a different kind, no?

  19. IanW

    Simples!

    Geographic co-ordinates (both flat surface and data points on a sphere) are data types in MongoDB, and you can fire indexed queries at them to boot. They are even adding various line and polygon related functions useful in GIS apps in the next (2.4) release.

    If anyone wants to learn MongoDB, there are more free 7 week, 10 hours/week online courses being offered shortly. Keep an eye on www.10gen.com or see the training pages at http://education.10gen.com/

  20. Sil

    TCO

    It goes both ways, open-source doesn't change the laws of physics either.

    Free software always sounds good, and makes sense often but still one has to see the big picture, including maintenance, support, ease of integration, training, specialist availability as well a programming environment, e.g. can programmers get more work done faster because of Superior dev ide and goodies.

    So in many cases also commercial software does make lots of sense.

    Also there are many different price points in both open-source and commercial offerings it's not a binary decision. I have no idea what's a Fairtrade Malbec is but to pursue your wine analogy, what's great with wine is that there are many choices in between a $90 bottle and a $1 bottle, and after a bit of (pleasant) research you can find the cheapest one that is still deemed enjoyable or highly enjoyable by all your guests.

    1. Christian Berger

      Re: TCO

      Well true, the licensing cost is only a tiny fraction of the TCO. However with open source solutions there is the great thing that you can have multiple sources of support for the same product. So you can choose the one you prefer most. Plus you might even be able to find employees involved in the development.

  21. Christian Berger

    Sometimes SQL-servers are not the right solution

    A few years ago I was involved in some project where large series of measurements had to be displayed in a web app. We are talking about 30 million data points here. The first solution the web designer came up with, was something based on MySQL. It was far to slow to be of any use. Getting a subset of the data took many seconds. Then I took my time to write a specialized piece of software working on simple files. It provided the same data in fractions of a second.

    1. Vic

      Re: Sometimes SQL-servers are not the right solution

      A few years ago I was involved in some project where large series of measurements had to be displayed in a web app. We are talking about 30 million data points here.

      A few years ago, I was involved in something similar (albeit with rather fewer points).

      My first go at improving the code was to remove the Javascript XML parser he had running on the client machines. It really wasn't very quick when you didn't have a super-dooper computer...

      Vic.

    2. Anonymous Coward
      Anonymous Coward

      Re: Sometimes SQL-servers are not the right solution

      totally - on a different topic look at the projects wanting to store files in records!

This topic is closed for new posts.

Other stories you might like