Date: Sun, 15 Apr 2001 23:06:11 -0700 From: "Ted Mittelstaedt" <tedm@toybox.placo.com> To: "Mikhail Kruk" <meshko@cs.brandeis.edu>, <questions@FreeBSD.ORG> Subject: RE: Hotmail FreeBSD-->Win2k migration Message-ID: <003801c0c63b$5644e660$1401a8c0@tedm.placo.com> In-Reply-To: <Pine.LNX.4.33.0104151416450.20540-100000@daedalus.cs.brandeis.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
A couple of comments for you on this. For starters, this really should go into advocacy. Everyone knew that when Microsoft bought Hotmail that they would eventually move as much of it as possible over to Microsoft server products. But, there are several observations that I'll make on this paper: 1) Generally, when people write these papers they start with a list of all the features that they want, and why the new software is going to provide them. They then follow with several paragraphs that explain why the existing "Legacy" software does NOT provide these features. You see, in business in these large enterprise computing environments, most computing decisions that tamper with an existing system are made on a cost basis. In short, unless the existing system is broken (thus failing and losing you money because pissed-off customers are going elsewhere) you would never do a migration like this unless considerable cost justification exists for doing it. In this paper, Microsoft first of all NEVER SAYS that the existing FreeBSD solution _does_not_work_. They never give a shopping list of things that FreeBSD _can_not_do_ and that Windows 2K _can_ do. Instead they ignore this completely - they give reasons for switching, but never come out and say that the existing solution _does_not_ meet those needs. Also, nowhere in this paper is any kind of cost-analysis that shows that they will save any money by doing this. In short, what you have here is a whitepaper on a migration that has no technical reason and no business reason for ever taking place. Clearly, this is an example of the "eat your own dog food" principle in action - rest assured that normal profit-making businesses would never do this migration just to be able to say they are running a different operating system. 2) Note that this paper stated that the actual migration took place from June and July of 2000. However, it does NOT state when the actual _planning_ for the migration commenced. It is reasonable to assume that the planning for the migration actually started in 1997 - the year that Microsoft purchased Hotmail. There were a number of unofficial reports in the press in 1998 by former Hotmail employees that Microsoft attempted to change the site to NT4 and failed. So, your looking at a migration here which has consumed at least 3 YEARS of planning by multiple people - many of them software developers. Nowhere in the whitepaper did I see any costs attached to this. Let me ask you: if 10 software developers that make $250,000.00 a year (I'll assume that because of the visibility of this project that they put very senior people on this project) work for 3 years, how much does that cost? Well, consider that with bennies being 2X the gross salary, I calculate 15 million dollars IN PERSONELL COSTS ALONE. And I don't know if they put just 10 people on this, they probably put a lot more people on it than that. And, note that this is JUST for the migration itself - this cost is OVER AND ABOVE the normal administrative costs of the service. Also, consider that as soon as it became evident that Microsoft intended to gut the FreeBSD section, I would guess that every FreeBSD developer that Hotmail employed that was any good would have bolted. So, on top of that you would have had to press all your Windows developers into service on FreeBSD - so now you have incompetent FreeBSD developers that are running the thing, which I think would have made the migration a lot harder to do. It's also very notable that the operating system that they DID move to - Windows 2000, was being written DURING 2 of these 3 planning years, and furthermore that the SAME company that did the migration, wrote the operating system. The situation is analogous to someone like General Motors deciding to switch from SAP (I assume that they use this) to some other software package, and then writing the entire software package and the operating system that it runs on from scratch. All this proves is 2 things. First is that Windows 2K has probably been enhanced to run Hotmail, and let me say that this is no guarentee that the enhancements to Win2K are going to be applicable to any other businesses computing model. Second, it certainly embodies the adage that if you throw enough money at something, you can probably get it to work. 3) They try a lot to hide this, but they are STILL using UNIX for at least some part of the service. Note these lines buried in the report: "...Scripts to implement the new version of the site code or other software components are pushed out to each machine (through RDIST)...." "...RDIST was the logical choice at Hotmail because the storage machines are still running UNIX and RDIST was already in use..." Also, note the front page of this: "...The original builders of the application created a two-tier architecture built around various UNIX systems. FreeBSD, a UNIX-like system similar to the Linux operating system, was used to run the front-end Web servers that handled login, Microsoft OutlookŪ Express, and Web-based content delivery tasks. The current network of more than 5,000 servers is organized into about a dozen clusters; each consisting of front-end servers linked to data storage machines..." So, in short, the "Data storage machines" are STILL running UNIX. The "front-end Web servers" WERE running FreeBSD and NOW are running Windows 2K. Hmmm... I wonder what that UNIX is that's running on the data storage systems, is it, could it be, SOLARIS? (In your best Church Lady voice) So, what we have here is a facinating migration that took place from FreeBSD to Windows 2K. Facinating not because of what it says - ie: all the technical details - but because of what it OMITS. The fact that the migration was completed at all is certainly a feather in Microsoft's cap - but all it proves is that yes, it's possible to use Windows 2K to run the front end services for Hotmail. So, now we all have proof that Windows 2K can now do something that FreeBSD was already doing, rather successfully I'd say (consider how successful that Hotmail got to running FreeBSD) In short, Microsoft has been struggling to _live_up_to_ FreeBSD with Windows 2K. So, now they have proven that they can do it - but they have NOT proven that Win2K is _better_. Also, they have NOT proven that this migration has any applicability to any other large FreeBSD or UNIX installation. I'll leave you with one final thought to consider. If FreeBSD is so worthless that Microsoft had to upgrade Hotmail to Windows 2K, then why hasn't Microsoft released all of the CGI and management scripts that were used under FreeBSD to run Hotmail? It would seem that if the existing FreeBSD solution was so poor, then they would want to show the world how much of a hack that the FreeBSD setup was. But, don't hold your breath waiting for this because it will never happen - the reason is that the FreeBSD installation was so GOOD that Microsoft is scared to death of releasing any of that stuff. Ted Mittelstaedt tedm@toybox.placo.com Author of: The FreeBSD Corporate Networker's Guide Book website: http://www.freebsd-corp-net-guide.com >-----Original Message----- >From: owner-freebsd-questions@FreeBSD.ORG >[mailto:owner-freebsd-questions@FreeBSD.ORG]On Behalf Of Mikhail Kruk >Sent: Sunday, April 15, 2001 11:19 AM >To: questions@FreeBSD.ORG >Subject: Hotmail FreeBSD-->Win2k migration > > >Here is the link: >http://www.microsoft.com/technet/migration/hotmail/default.asp > >IMHO it's worth reading. I'd like to hear what people think... ideally >something from people who work/used to work at Hotmail. >Some parts of their description make sense, but others don't. For one I >can't believe their stories about how it's much easier to develop and >debug software on Windows. Of course they probably assigned Windows >developers on the project. >Also here is some math: >While running FreeBSD Hotmail had 3,600 servers. Now it has more than >5,000. It's been less than a year since the transition. Before >transition size of the cluster was 300, after transition -- 400 (of >course it's more of a design issue than a performance indicator, but >still). > >So what everybody thinks? In particular I don't have any experience with >CGI and multithreaded alternatives so I'd like to hear opinions of >experienced people about that. > >[aslo posted to c.u.b.f.m, please cc: to me if replying, I'm not >subscribed] > > >To Unsubscribe: send mail to majordomo@FreeBSD.org >with "unsubscribe freebsd-questions" in the body of the message > To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?003801c0c63b$5644e660$1401a8c0>