Will Nginx Be to Apache What Chrome is to Firefox?

The Netcraft Web Server Survey, which appears each month, is usually viewed as offering the spectacle of a two-player fight between the open source Apache and Microsoft's IIS. Actually, that's giving Microsoft too much credit, since it's never really been a fight: IIS has occasionally tried to claw its way closer to Apache's market share, failed dismally, and then started sinking back again. But there's another story in these graphs.

On Open Enterprise blog.

The Chinese (Web Servers) Are Coming

The monthly release of the Netcraft survey is always good, since it generally shows the continuing dominance of Apache in the Web server field. But this month has something new and vaguely frightening:

In the February 2009 survey we received responses from 215,675,903 sites. This reflects a phenomenal monthly gain of more than 30 million sites, bringing the total up by more than 16%.

This majority of this month's growth is down to the appearance of 20 million Chinese sites served by QZHTTP. This web server is used by QQ to serve millions of Qzone sites beneath the qq.com domain.

QQ is already well known for providing the most widely used instant messenger client in China, but this month's inclusion of the Qzone blogging service instantly makes the company the largest blog site provider in the survey, surpassing the likes of Windows Live Spaces, Blogger and MySpace.

Got that? QQ's server QZHTTP just put on 20 million sites in the survey - enough seriously to dent both Apache and IIS (and making the latter look suddenly vulnerable to losing its second place).

Does this represent the dawn of a new (Web server) era?

What makes this all slightly troubling is that I don't know anything about QZHTTP: I presume it's not open souce, since I can't find any links to its code. But can anyone give me any more details, please? (Via @codfather.)

Follow me on Twitter @glynmoody

The Empire (No Longer) Strikes Back

One of the most worrying moments in recent open source history was when it became clear that Microsoft was determined to wrench away Apache's crown as top Web server. This began in early 2006, and was soon showing dramatic results, as the April 2006 Netcraft survey commented:

This month's survey brings one of the largest one-month swings in the history of the web server market, as Microsoft gains 4.7 percent share while Apache loses 5.9 percent. The shift is driven by changes at domain registrar Go Daddy, which has just migrated more than 3.5 million hostnames from Linux to Windows. Go Daddy, which had been the world's largest Linux host, is now the world's largest Windows Server 2003 host, as measured by hostnames. The company said it will shift a total of 4.4 million hostnames to Windows Server 2003.

This was a staggering shift, and I feared it might presage a real effort by Microsoft to achieve a major PR win. Things reached their nadir in September 2007:

Apache gains over 3 million hostnames, and around 0.9 million active sites this month. But this is not enough to prevent its market share declining closer to the 50% mark, as Microsoft also gained over 3 million hostnames (a large part of which come from MySpace and Live Spaces, both of which use its Internet Information Server.

At that time, the gap between Apache and Microsoft's IIS was just 15%, down, from 50% just a couple of years earlier.

But since then, Apache has gradually pulled ahead; today the gap is around 18% - still far smaller than it once was, but increasing. I feel that the danger has passed, not least because Microsoft has realised that it was fighting yesterday's battles.

Tomorrow's fight will be about owning the cloud, and the main threat there is not so much Apache, as customised versions of open source software, of the kind employed by Google for its vast server farms: in the latest Netscape survey, Google has around 5% of the Web server market. It's still open vs. closed, but not as we know it.

The crucial point is that Microsoft failed to displace Apache, despite its almost limitless resources. This is the crucial lesson for the future, more important than any particular percentage market share: that Microsoft's attacks can - and have been - beaten off.

Apache on the Up

Not much, mind you, but given that I commented when it was on the way down, it's only fair to point out moves in the other direction:

Apache continues its recovery after steep falls in share over the last eighteen months and is back over 50%. Its share had been negatively affected over that period by the increasing number of blog sites in the survey on large providers like Microsoft and Google, using their own server software. But it is also benefiting from growth at other blog providers like multiply.

After Netcraft

For over a decade, it has been a point of faith that Apache is not only a better Web server than Microsoft's IIS, but that this is demonstrable: the Netcraft survey of public Web servers shows that Apache has been consistently ahead.

Alas, for a variety of reasons - not least Microsoft's determination to reduce the gap, whatever the cost - Apache's lead is falling. So it's good to have this new survey that re-asserts Apache's superiority, and adds a nice extra twist:

Linux websites have better uptime and load faster than Windows-based websites. Research by WatchMouse, a website monitoring company, also shows that web server platform Apache outperforms the Microsoft IIS platform. Therefore, having a Linux website and an Apache webserver platform offers the best choice for professional web pages.

I feel another Microsoft-funded piece of research on its way....

Apache Starts to Patch the Holes

The latest Netcraft survey shows that Apache has pulled back some of the ground it lost to Microsoft's Web server last month. There have been some pretty massive swings recently, as the oscillations in the graph show: these are largely due to switches in the hosting sector, which can often involve millions of Internet names at a stroke. For example, Go Daddy moved over 1.6 million hostnames from Apache to Microsoft's IIS platform in June.

These new gains for Apache are important, because it suggests that Microsoft's relentless campaign to "convince" hosting companies to switch to its products (and who wouldn't love to be a fly on the wall for those conversations?) may finally have run out of steam. It will be interesting to see what happens next month.

Apache Now Leader in Secure Web Servers Too

One of the statistics most often trotted out to demonstrate open source's rise and reach is Apache's total dominance of the public Web server sector (I should know, I've done it often enough myself). This has always stuck in Microsoft's craw, and their standard response is "Well, it doesn't really count since it's mostly mickey-mouse Web servers, whereas we are the tops for grown-up secure Web servers" (not their phraseology, but you get my drift).

The news that Apache is now the leading secure Web server as well as the leading Web server overall rather blows this story out of the water. It also means that all that hard work Microsoft has been doing converting domain registrars in a desperate attempt to boost its market share - that is, gaining share among the mickey-mouse Web servers it so pooh-poohed before - was a complete waste of time and money.

Open Data in the Age of Exponential Science

There's a very interesting article in this week's Nature, as part of its 2020 Computing Special (which miraculously is freely available even to non-subscribers), written by Alexander Szalay and Jim Gray.

I had the pleasure of interviewing Gray a couple of years back. He's a Grand Old Man of the computing world, with a hugely impressive curriculum vitae; he's also a thoroughly charming interviewee with some extremely interesting ideas. For example:

I believe that Alan Turing was right and that eventually machines will be sentient. And I think that's probably going to happen in this century. There's much concern that that might work out badly; I actually am optimistic about it.

The Nature article is entitled "Science in an exponential world", and it considers some of the approaching problems that the vast scaling up of Net-based, collaborative scientific endeavour is likely to bring us in the years to come. Here's one key point:

A collaboration involving hundreds of Internet-connected scientists raises questions about standards for data sharing. Too much effort is wasted on converting from one proprietary data format to another. Standards are essential at several levels: in formatting, so that data written by one group can be easily read and understood by others; in semantics, so that a term used by one group can be translated (often automatically) by another without its meaning being distorted; and in workflows, so that analysis steps can be executed across the Internet and reproduced by others at a later date.

The same considerations apply to all open data in the age of exponential science: without common standards that allow data from different groups, gathered at different times and in varying circumstance, to be brought together meaningfully in all sorts of new ways, the openness is moot.