Is Yahoo! moving from FreeBSD?
tedm at toybox.placo.com
Thu Feb 24 09:43:39 GMT 2005
> -----Original Message-----
> From: owner-freebsd-questions at freebsd.org
> [mailto:owner-freebsd-questions at freebsd.org]On Behalf Of Freminlins
> Sent: Wednesday, February 23, 2005 7:42 AM
> To: Jorn Argelo
> Cc: questions at freebsd.org
> Subject: Re: Is Yahoo! moving from FreeBSD?
> On Wed, 23 Feb 2005 16:36:36 +0100, Jorn Argelo
> <jorn at wcborstel.nl> wrote:
> > I don't think that they would. That'll be a massive
> migration involving lots
> > and lots of costs. They have to pay for RedHat Enterprise
> too. The only reason
> > I can think off is that they want support.Perhaps I missed a
> part, but I don't
> > see the word FreeBSD in that article.
> Although it doesn't state FreeBSD, I understand that Yahoo! runs stuff
> on FreeBSD.
> > Besides, the point of the article is not regarding a
> migration of Yahoo, but
> > Linux and IT in general. It has nothing to do with Yahoo or
> FreeBSD. I think
> > that the author of the article is simply mistaking.
> I'n not sure I agree with that. The author stated "But in December,
> Yahoo started to port ... to Red Hat Enterprise Linux 4.0" That would
> suggest that Yahoo! is moving to Linux.
> I am very interested in this as I have for several years used the
> argument "we use the same OS as Yahoo!". We're not going to migrate to
> Linux if Yahoo! does.
The speaker cited in the article is a "Mason Ng, Yahoo's director of
However I found a "Kevin Timmons, director, engineering operations,
when I googled this up.
I've found numerous references to him in conjunction with Yahoo.
Interestingly, I also found Mason Ng listed as well - here:
as an Oracle Group Product manager. But, very few other references to
name. As it's a rather unique name, I would suspect that the Mason Ng in
the article and the one listed as an Oracle employee last year are one
and the same person.
I would presume that if the article in Computerworld was in fact an
accurate quote, that what has happened is that Yahoo has decided to
move to using Oracle as a backend database, and they stole an Oracle
manager away from Oracle to oversee the move. As Linux is the only open
source OS that Oracle ships on, and Yahoo obviously wants to keep using
source, it gets elected as the platform.
Keep in mind that in 2004, Yahoo stopped using google results, see here:
This represented a return to their roots. Back in 1995 when Yahoo first
got going, it was only their own links on a cobbled-together software
database. Then later on in 1998 they switched
to the Inktomi database, see here:
Obviously then sometime later than that, when the search
market started these large monster OEM search providers, Yahoo started
using Google, as did may other search engines. Then once the users
noticing that they were getting the same results no matter what search
engine was being used, sites like Yahoo realized they better
their products and so they went back to developing their own database.
Yahoo still uses FreeBSD according to this FAQ:
And I would assume that they will probably only use Red Hat where they
want to field Oracle.
Note also that Yahoo is moving to Red Hat Linux according to the article.
NOT moving to "Linux" Red Hat is not free, it is a commercial server
product just like Microsoft's operating systems. Fedora - that is free.
But they aren't moving to Fedora, they are moving to Red Hat.
I think that there is enough circumstantial evidence of what is really
going on for educated guesses to be drawn. Doubtless further googling
will reveal more info about Yahoo to anyone interested. But clearly,
Yahoo has correctly realized that it's database is the real valuable
part of the company, and they have decided to get out of the
database software business. Oracle is the obvious choice as it's
designed for large scale operations, exactly what Yahoo is running.
But if your going to run Oracle, you won't get any support from Oracle
unless you run it on a supported operating system.
I also seriously question that Yahoo's database was ever in the past
on FreeBSD. I suspect their model was an identical model to Hotmail's -
a small core of strange database servers (Yahoo originally ran indy's,
see http://www.sun.com/950523/yahoostory.html for their very first
surrounded by a bunch of cheap Pentiums running
FreeBSD as front end servers. Now they are moving to replace that core
with an Oracle core. But there's no indication that they are going to
replace the shell of FreeBSD servers with RedHat.
Quite obviously Yahoo regards the internal workings of their technology
trade secrets, that is why there's not a lot of documentation out on the
web on it.
Now, I will address one other point, the "we use the same OS as Yahoo!"
First of all, 99.9999% of business customers that are out there are NOT
anywhere near the size of Yahoo and never will be. I am quite sure that
anyone who owns a business dreams of growing to be big and rich. But
how many small business are there in the world compared to large ones?
A business must be run realistically. It is a huge waste of money to go
spend a pile of it on a big "large company" software just in the
your going to grow into it. It is also a waste of money to run out and
a Sun cluster just in the assumption your going to grow into it.
FreeBSD does not have some of the things - such as distributed management
of hundreds to thousands of FreeBSD servers over a large enterprise -
are a requirement for big companies. So, if you have a company even if
are the most pro-FreeBSD company in the world, if they happen to grow to
that size, no matter how much they like FreeBSD they can't use it. Yahoo
is presumably finding this out - they want to use Oracle, so that brings
all the baggage of having to run a platform that Oracle tells you to run
so they will support you.
But here is my main objection to the "we use the same OS as yahoo"
Small companies are not big companies, and big company solutions seldom
are a good fit for them. FreeBSD is an excellent small to medium sized
organization solution. So is Slackware, Debian, Fedora, and a host of
Linuxes. A big company solution is Solaris, or Red Hat (at least, that
what Red Hat is rapidly progressing to) You would not use a small to
medium sized company solution where you need a big company solution - and
you would also not use a big company solution where a small to medium
sized company solution is needed. That doesen't mean of course that it's
impossible to do it - you can for example use Solaris for a small company
server - but the effort required to go against the grain is much higher.
Solaris for example comes with no compiler and you must compile by hand
all the applications you need, and often you must recompile the complier
just before you can even start doing that. It takes days - whereas the
FreeBSD ports system takes a few hours for the largest and most complex
More information about the freebsd-questions