Slashdot Mirror

← Back to Stories (view on slashdot.org)

A Look At Microsoft's 'Mini Internet' For Testing IE

Posted by Soulskill on from the a-series-of-tiny-tubes dept.

MrSeb writes "With the grandiose bluster that only an aging juggernaut can pull off, Microsoft has detailed the Internet Explorer Performance Lab and its extraordinary efforts to ensure IE9 is competitive and IE10 is the fastest browser in the world. Here are a few bullet points: 128 test computers, 20,000 tests per day, over 850 metrics analyzed, 480GB of runtime data per day, and a granularity of just 100 nanoseconds. The data is reported to 11 server-class (16-core, 16GB of RAM) computers, and the data is stored on a 24-core, 64GB SQL server. The 'mini internet' has content servers, DNS servers, and network emulators (to model various different latencies, throughputs, packet loss)."

28 of 241 comments (clear)

  1. Could use the real internet eh! by Anonymous Coward · · Score: 4, Interesting

    Why not just use the real internet?

    1. Re:Could use the real internet eh! by weszz · · Score: 5, Informative

      They wanted to account for any kind of lag, so by having it all in house and disconnected from even their internal network, they have control over all variables so everything is equal.

      They did this post on their blog yesterday http://blogs.msdn.com/b/b8/

    2. Re:Could use the real internet eh! by greichert · · Score: 5, Insightful

      Because you can not have reproducable results on the real Internet. Only a fake one, where eveyrthing is controlled and reproducable, can be used for testing and making sure some settings do not make the browser slower.

    3. Re:Could use the real internet eh! by ackthpt · · Score: 3, Interesting

      They wanted to account for any kind of lag, so by having it all in house and disconnected from even their internal network, they have control over all variables so everything is equal.

      They did this post on their blog yesterday http://blogs.msdn.com/b/b8/

      They care about it so they created a genuine imitation of the real thing.

      Honestly, I'd go at some of the pages I have to each day, which are ludicrous in their use of content and scripting - web developers just pick up and drop widgets all over the place, never a look toward what impact it has on the page being interpreted or used on the receiving end. I know I've got a bad one when I hear the processor fan kick in for a stinkin' web page!!!

      --

      A feeling of having made the same mistake before: Deja Foobar
    4. Re:Could use the real internet eh! by weszz · · Score: 3, Informative

      they never do say what kind of sites, just that they are sanitized and are real world web pages... but if you are comparing IE to itself or Chrome or FF, whatever site you use should be similar in speeds, except for ones that detect clients and do different things...

      "Content servers are web servers that stand in for the millions of web hosts on the Internet. Each content server hosts real world web pages that have been captured locally. The captured pages go through a process we refer to as sanitization, where we tweak portions of the web content to ensure reproducible determinism. For example, JavaScript Date functions or Math.Random() calls will be replaced with a static value. Additionally, the dynamic URLs created by ad frameworks are locked to the URL that was first used by the framework.

      After sanitization, content is served similarly to static content through an ISAPI filter that maps a hash of the URL to the content, allowing instantaneous lookup. Each web server is a 16-core machine with 16GB of RAM to minimize variability and ensure that content is in memory (no disk access required).

      Content servers can also host dynamic web apps like Outlook Web Access or Office Web Apps. In these cases, the application server and any multi-tier dependencies are hosted on dedicated servers in the lab, just like real world environments."

    5. Re:Could use the real internet eh! by gorzek · · Score: 4, Funny

      Yes, there must be some conspiracy! Microsoft couldn't possibly want to make a good browser! They must have ulterior motives!

      --
      Check out my world simulator thingy.
    6. Re:Could use the real internet eh! by Bill,+Shooter+of+Bul · · Score: 5, Insightful

      You have to figure out what the variables that you have problems with in real world usage, before you can start optimising your product to account for them.
      There has to be iterative cycles of real world, then fake internet testing to really make it work well.
      It would also help if you were able to test your competition alogn the same lines.

      I additionally wonder if they are accouting for all of the different behavious of all of the various webservers out there. If they are only testing agianst iis, well, that's not very good.

      --
      Well.. maybe. Or Maybe not. But Definitely not sort of.
    7. Re:Could use the real internet eh! by thatskinnyguy · · Score: 3, Insightful

      Because that would be somewhat unscientific. A lab setting is controllable and you would be able to trigger and know where latency is coming from and how to correct certain behaviors in software. With the real internet, it's anyone's guess where lag is coming from.

      --
      The game.
    8. Re:Could use the real internet eh! by gorzek · · Score: 4, Insightful

      So, what's going to make anyone use a Microsoft browser? They've been losing market share for ages because their browsers suck. How do they get people back? Make a good browser. Your argument might mean something if Microsoft sold IE as a standalone product, but they don't. It costs nothing (in terms of cash coming straight out of your pocket) to switch browsers, and users are notoriously not fond of switching. Since you can't count on your competitors' products to be lousy, you can only compete by making yours better. The browser market is about as Darwinistic as a software market could be.

      --
      Check out my world simulator thingy.
    9. Re:Could use the real internet eh! by smooth+wombat · · Score: 4, Insightful

      web developers just pick up and drop widgets all over the place,

      Rule #2 of IT that should never be broken: Never let a web designer design your web page.

      Giving free reign to a web designer to design a web site is like giving a two year-old a Faberge egg.

      --
      We will bankrupt ourselves in the vain search for absolute security. -- Dwight D. Eisenhower
    10. Re:Could use the real internet eh! by Bengie · · Score: 3, Insightful

      Probably an evolving tick-tock setup between real and test.

      Step1. Performance Test App against real world
      Step2. Document Real world issues
      Step3. Create test environment to run your issue cases against
      Step4. Optimize app against test environment
      Step5. Goto Step1, adding any new cases

      Steps 1 and 2 can happen independent of 3 and 4. Step5 is just to make the logic seem serial.

  2. And still... by jcreus · · Score: 4, Insightful

    Beaten by Chrome and Firefox.

    1. Re:And still... by Anonymous Coward · · Score: 5, Funny

      The only thing Firefox does fast anymore is update.

    2. Re:And still... by thedonger · · Score: 5, Insightful

      And when all we care about is the fastest browser - in nanoseconds! - will we begin to forget the truly important criteria for choosing a browser?

      Or better still, by the time IE is on par with Chrome the actual browser will be irrelevant because mobile platforms - in which IE has little share - will do to traditional computers what Cromagnons did to Neanderthals. The next generation will use integrated devices, unaware they were using a browser, and with little or no need for even a choice.

      --
      Help fight poverty: Punch a poor person.
    3. Re:And still... by mcrbids · · Score: 3, Insightful

      Firefox hasn't become any slower. What's happened is that everything else become so much faster. There used to be a "pregnant pause" when entering a domain, I remember when 30 seconds to load a page was the norm, over 56k.

      Now, I expect the results from a google search to appear as I type, interactively. This isn't just an improvement, it's a whole 'nother animal at that level of performance.

      --
      I have no problem with your religion until you decide it's reason to deprive others of the truth.
  3. But will it run Linux? by erroneus · · Score: 4, Interesting

    I couldn't resist. But with all the work and effort and resources going into this, how is it that operations a tiny fraction of this can generate fast, reliable and standards complaint browsers better than MSIE?

    Microsoft, the problem isn't that you're not spending enough money. It's that you're not doing it right.

    1. Re:But will it run Linux? by roc97007 · · Score: 3, Insightful

      I couldn't resist. But with all the work and effort and resources going into this, how is it that operations a tiny fraction of this can generate fast, reliable and standards complaint browsers better than MSIE?

      Microsoft, the problem isn't that you're not spending enough money. It's that you're not doing it right.

      I'm not familiar with IE 8 and 9, but in the past the issue was many years and revisions of code reuse and accumulation of cruft, an insane amount of backwards compatibility and some poor initial design choices combined to make each new version bigger, slower and buggier. I imagine this combines to make developing and especially testing any new release a massive undertaking.

      To be fair, I would argue that Firefox is starting that downward spiral now. Each new version is slower and has a bigger footprint. Personally, I've switched to Chrome, but when Chrome inevitably starts lagging, I'll be on the lookout for the next completely new browser. Not merely because it's new, but because it's less likely to have years of bad decisions weighing it down.

      The problem in Microsoft's case is that they seem incapable of dumping what they have and doing a complete rewrite. There may be marketing reasons for this, granted, but if they made a clean break it'd be better for them in the long run.

      --
      Oliver's law of assumed responsibility: If you're seen fixing it, you will be blamed for breaking it.
  4. 1/10,000 of a millisecond by tepples · · Score: 4, Funny

    It means that each millisecond of ping is divided into over 9000 parts.

  5. Re:Granularity of 100 nanoseconds by ackthpt · · Score: 4, Funny

    Granularity of 100 nanoseconds: What does that mean?

    That's as small as they could get the bits, pounding on them with Steve's chair.

    --

    A feeling of having made the same mistake before: Deja Foobar
  6. Re:IE Crap by Krojack · · Score: 5, Funny

    And still unable to correctly implement CSS2 and HTML4

    Fixed that for ya.

  7. HHGttG by Verdatum · · Score: 3, Funny

    Wasn't this a plot-point in the Hitch-Hiker's Guide to the Galaxy series? Having an artificial universe on-site so they could go exploring but still be able to come back for long lunches...

  8. Re:IE Crap by Anonymous Coward · · Score: 5, Insightful

    Correctly according to whom? Neither CSS3 nor HTML5 are completed standards and various portions remain in flux. Of the more mature bits IE9 and IE10 implement quite a bit of it and do so quite comprehensively.

    MS is also one of very few organizations that is very actively involved with the W3C Test Suite by submitting test cases for each portion of the standard under various circumstances to demonstrate correct behaviors. What Google and Mozilla do instead is slap together a partial implementation and call it a day. More than once has their implementations been found to be not only incomplete but also incorrect.

    Stop relying on scores given by non-authoritative tests demonstrating exceedingly limited and selective interpretations of non-standardized functionality. Oh, and HTML5 Video does NOT specify a codec, in fact it was designed to handle many simultaneous codecs, including h.264, which is explicitly referenced in the draft.

  9. Mini Internet? by Smask · · Score: 4, Funny

    Does that mean they have only porn sites with midget porn? And a mini 4chan, populated with toddlers?

  10. Re:Was /. been bought or what? by gorzek · · Score: 4, Funny

    What we really need is another Bitcoin story!

    --
    Check out my world simulator thingy.
  11. Re:Granularity of 100 nanoseconds by X0563511 · · Score: 3, Insightful

    It means the runtime data they have collected is stored in a 100ns interval.

    --
    For large sets, this will be our guide even unto death, for the LORD will work for each type of data it is applied to...
  12. Firefox isn't slow at all. by sakdoctor · · Score: 4, Insightful

    While slashdot mocks the computer industry marketing for describing computers using a single metric, you seem to be quite happy with that when it comes to browser performance.

    An example: Chrome (v8 engine) has this reputation for amazing speed, but IE9 absolutely grinds Chrome into the dust when it comes to simply repositioning elements on screen; something which today's web apps spend a lot of their time doing. You can feel it too if you know what you're looking for. I don't follow IEs development as closely as Chrome or Firefox, but IE must be hardware accelerating these translations.

    I fully expect Google to focus on performance cases which help their specific apps. Again, a conflict of interest, akin to Microsoft pre-caching masses of junk, so that Office can appear to start up much faster than the competition.

  13. Why not? by kiwimate · · Score: 5, Insightful

    MSDN blogs are often very technically detailed, written by people who know this stuff from the inside, and if it's about a topic that's of general computing interest then it seems that's a good thing. And the blog in question is chock full of some really good detailed stuff about how they're doing performance testing, reasons why the lab is architected the way it is, detailed graphics on how they measure performance, how they analyze it...on and on.

    Frankly, this seems more akin to old Slashdot than a lot of the nonsense we see here today. (That story the other day about a girl sent home from school because her lunch wasn't healthy, and then quickly called into question over what happened? Really? What was that topic even doing on Slashdot in the first place?) Whatever you think about Microsoft, having this extremely detailed look into how one of the world's biggest software vendors (or are they the biggest now?) goes about performance testing, and how they ensure consistent results, should be really, really interesting to anyone involved in IT.

  14. Re:Granularity of 100 nanoseconds by Renegrade · · Score: 4, Informative

    No, it means that Windows has a 100ns granularity on it's timestamps.

    http://msdn.microsoft.com/en-us/library/windows/desktop/ms724284(v=vs.85).aspx