I attended the USENIX Annual Technical Conference (ATC) this year in San Antonio, Texas for the O'Reilly Network. The USENIX ATC offers attendees an interesting mix of papers and talks by academia, well-known industry professionals, and researchers working for companies across the world.
The conference offers three days of advanced tutorials, but I only arrived for the keynote by Neal Stephenson (author of Cryptonomicon and also the near-famous "In the Beginning Was the Command Line"), three days of technical sessions, Birds of a Feather (BOF) talks, and Works In Progress (WIPs). Technical sessions are only an hour and a half long and typically provide a bird's-eye view of the technology being discussed. BOFs are much less formal, and usually involve the audience working closely with the speaker to discuss cool new events and happenings (I even attended a Microsoft BOF -- more on that later). Finally, WIPs allow presenters to discuss the status of projects that are very much still in the works.
My first full day at the USENIX ATC was Thursday, when I made a visit to "Samba — Ins and Outs, LDAP" by Gerald Carter of the Samba Team. Carter, who seems to be something of a geek, as you may find when you visit his site, is the author of LDAP System Administration by O'Reilly & Associates as well as the co-author of Sams' Teach Yourself SAMBA in 24 Hours, with Richard Sharpe.
If you haven't been keeping up with the open source Samba project, the aim
of which is to provide a Windows-compatible file, print, and authentication service,
hold on to your seat. As Samba users are already well aware, Samba 3.0 is
slated to support Active Directory. According to Carter, the support has reached
stable levels, and is basically ready. Indeed, if your system has Kerberos
and LDAP support already installed when
compiling Samba, the
configure script will detect them and build
in Active Directory support for you. At that point, you just need to do some
configuration and you are ready to go.
Later that day I attended a very interesting guru session by Andrew Hume (yes, he does have a big smile like that in real life) of AT&T Labs (which seems to be doing some very interesting work, according to their research areas page), titled "Legacy Systems/Big Data/Freenix Clusters." Hume is currently working on a large Linux cluster used by AT&T as a computational resource and may be moving to one of the BSDs in the future. If you haven't noticed the power and low expense of open-source-based clusters, then you need to take a better look at the market. There are a lot of vendors pushing Linux-based clusters, and it's quite easy to brew your own custom solutions. Seeing AT&T using these solutions demonstrates that big business has really taken a hold on open source solutions as a real cost saver.
Interested in doing a Linux-based cluster? You can build your own Beowulf if you want, quite easily, by following the guidelines at the Beowulf homepage or by purchasing a pre-built system from one of the many vendors such as Linux Networx. If you do build a Linux cluster for fun, let me know. I am always interested in HPC on a low budget.
The last session that I attended Thursday was on n-Place Rsync. The talk was by David Rasch and Randal Burns of Johns Hopkins University. (Unfortunately, I was unable to find any links to this project. If someone knows where n-Place Rsync is hiding, let me know and I'll update the article.) n-Place Rsync was developed as a thesis project with the goal of allowing rsync-like capability for devices with extremely low amounts of storage. Specifically, Rasch and Burns hoped to address file updates on mobile and wireless devices. The speakers addressed the problem of file updates on these kinds of devices by doing in-place replacement where temporary storage was not needed. It's not clear whether n-Place Rsync has been released into the open source community, but the idea seems quite useful. I hope to see more about it, and projects based on it, in the near future at SourceForge.
As mentioned earlier I did make a showing at the Microsoft BOF. What did Microsoft have to say at a conference dominated by those embracing open source? Well, migration, of course! Microsoft sent people that work closely with their Microsoft Windows Services for UNIX product to discuss how Microsoft is trying to address the needs of UNIX administrators. In all honesty, the BOF was very well done. (Whether I say that because there were plentiful amounts of beer and wine will not be discussed.)
The Microsoft representatives were quite open about asking the audience what problems they were having when working with Windows systems in their networks and how Microsoft could address those problems. They also spoke in depth about Microsoft Windows Services for UNIX, which they say provides an almost complete development and execution environment for UNIX software and services. Indeed, it would appear that the software provides an environment as complete as Cygwin, but with closer integration into Windows.
For the uninformed, Cygwin provides a set of libraries and binaries that run on top of Windows to give you near UNIX- like capabilities and compatibility. For example, you can run a version of XFree86 on your Windows machine, if you desired.
As far as open source is concerned, the speakers were more than happy to mention that they had GPL'd software included with Microsoft Windows Services for UNIX and that the source code was available at microsoft.com. They did not develop the software, of course, but have included it with their product to build a more complete environment. Is Microsoft seeing the light? Doubtful. It would appear that they simply see that UNIX people are using certain software on a regular basis, and so they had better support it too. This is simply good marketing.
On a related note, I spoke with both presenters personally after the BOF and found one obvious characteristic: both had very heavy UNIX backgrounds. The shared experiences included working with a well-known UNIX operating system company, developing the initial product that later was vacuumed up into the Microsoft Windows Services for UNIX product from Microsoft, and even working with Red Hat. What's the lesson to take from this? Microsoft is hiring the best that they can find to make sure they can compete with UNIX head-to-head. Whether they will succeed is best left to future historians.
I spent much of the next morning attending the Annual Meeting of the USENIX Board. USENIX is obviously having a hard debate on how to better market itself to those that would best benefit from membership, and much of the meeting revolved around better ways to do it. It's good to see that a diverse group such as USENIX continues to try and revitalize its image and work better to become a part of the larger research and industry communities.
Later Friday night, I attended an absolutely fascinating Work-In-Progress by the guys from LinuxBIOS. If you haven't been tracking what LinuxBIOS is doing, then now is the time. LinuxBIOS replaces the traditional BIOS, allowing a computer to boot quickly and offer resources as a node to cluster. The LinuxBIOS guys were talking about times of 12 seconds from system start to hand-off to the operating system. What's so cool about LinuxBIOS is that the project itself is open source, even though they work closely with commercial interests. LinuxBIOS is showing that the combination of open source and commercial support can indeed work. I think we will be hearing more from them in the future.
The last sessions that I attended were on Saturday, and focused mostly on the business side of web hosting and system administration. In the web hosting guru session, the needs of a web hosting environment were picked apart in some level of detail. Notably, many hosting environments rely extensively on open source software. (Surprise, surprise!) Solaris continues to have a very heavy presence with ISPs and web hosting services, but Linux and certainly FreeBSD continue to push into this market. The primary reason seems to be cost. Linux and FreeBSD are known for their stability, so in many cases the only differentiating factor between them and Solaris or other UNIX systems are purchase, licensing, and operating costs. Linux and FreeBSD win here, hands down. Solaris continues to make a difference when large amounts of IO are concerned, but even there, FreeBSD can really compete.
What exactly did I really learn from this conference? First of all, research is as strong as ever and there is a lot coming out of the USENIX and open source communities. Also, Samba is really starting to break significant ground with Active Directory, and it looks like they are ready to really push into the Active Directory networks that are springing up left and right.
As a parting thought, if you are pushing open source in your network, you need to be ready to convince management why that solution is the best one, because Microsoft is not making any pretense about just being another player in the high-end market. Instead, they are really pushing hard to show that they have the software and solutions to solve that problem for you, and with their Microsoft Windows Services for UNIX, they are trying to compete directly with your UNIX servers to support your own UNIX users. Are you ready to take them on?
Dustin Puryear is a consultant providing expertise in managing and integrating UNIX and Windows systems and services, with a strong focus on open source.
Return to the Linux DevCenter.
Copyright © 2009 O'Reilly Media, Inc.