OpenSolaris in the Lab

By Jason Brooks  |  Posted 2009-01-30 Print this article Print

I tested OpenSolaris 2008.11 on a Lenovo Thinkpad T60 with 3GB of RAM and an ATI X1300 graphics adapter. Support for suspend-to-RAM-aka sleep mode-is a new feature of OpenSolaris 2008.11, but one that's available on a limited range of notebook hardware, which apparently did not include my test machine. Also new in OpenSolaris is support for 3D acceleration and Compiz graphical eye candy on systems with Nvidia graphics adapters, but not on ATI systems, such as my test machine.

The system supported my wired Ethernet adapter, an Intel 82573L, without issue, and for the most part, my Intel 3945ABG wireless adapter also worked well. However, I did experience some wireless issues with my somewhat aged home access point-after periods of connectivity of varying lengths, OpenSolaris would lose its link, requiring me to restart the system's NWAM (Network Auto Magic) daemon and connect again. I've experienced no such issues with the same hardware and access point with Linux and Windows.

As I mentioned earlier, OpenSolaris ships with most of the same applications that grace the Linux desktop, with the notable exception of the Tomboy note taking tool (one of my favorite Linux applications) or any other applications based on Mono, the open-source implementation of Microsoft's .NET framework. This is too bad, because a fair amount of the desktop application innovation occurring on Linux is Mono-based.

I found the Firefox 3 installation that ships by default with OpenSolaris noticeably pokey in its performance-typing status updates into Twitter, for instance, was practically too sluggish to stand. Early in my testing, I installed a beta version of Firefox 3.1, which delivered acceptable performance. Initially, I also missed Mozilla Prism, the project's site-specific Web browser that I use for keeping my GMail isolated from the rest of my Web browsing for performance and security reasons. However, during my tests, Sun produced a build of Prism that suited my needs.

My experiences with Firefox and Prism on OpenSolaris offer a window into the current state of software packaging and availability on the platform, which is marked by a mix of new and old software management systems, unpackaged tarball binaries, and applications for which no Solaris version exists.

The poorly-performing version of Firefox that I started out with came from the OpenSolaris IPS repositories, which work more or less like the repository systems for Ubuntu or Fedora. I could configure my system with multiple software repositories, such as those for stable and development packages, or for contributed packages offered through the main OpenSolaris project and through the volunteer Blastwave packaging project.

I could access pre-packaged software either from the command line or through a graphical client that resembles Ubuntu's Synaptic, but I couldn't browse through available packages in all of my configured repositories at once, the way I can on Ubuntu or Fedora. Rather, when I searched for packages, I had to visit each of my configured repositories to see if the software I sought was available.

I would have preferred to stick exclusively to software in the system's IPS repositories, but the faster Firefox 3.1 build was not available through these channels, but rather in a contributed builds folder on Mozilla's FTP site.

What's more, the package was not in IPS format, but in Sun's older SRV4 format, which OpenSolaris still supports. As far as I could gather, SRV4 packages are unseen by OpenSolaris' IPS system, and multiple package formats on a single machine lead, in my experience, to management headaches down the road. The Mozilla Prism build that I mentioned came in yet another format, a plain tarball that I extracted and ran from my home directory.

I expect this packaging fragmentation to improve over time-the current state of Solaris software management has improved dramatically in just the past year-but a tougher challenge to overcome involves applications that haven't been ported to run on Solaris at all. As I write this, the piece of software that comes most quickly to mind is Google Gears, the offline Web app-enabling software that Google has just embraced for its GMail service.

As I mentioned earlier, it's possible to run Linux software under the Solaris kernel using the system's branded containers functionality, but the feature is no picnic to configure for graphical applications, and involves: creating a tarball of an existing Linux system; configuring a branded container, complete with a static IP address that's bonded to a specific network adapter (an unwelcome proposition for notebook users who switch frequently between different wired and wireless networks); and configuring your Linux container and your host system to allow forwarding X over ssh.

I would like to see the OpenSolaris project streamline the process of installing and using Linux applications within a branded container, perhaps by making a bare-bones, relatively up-to-date installation of CentOS or Ubuntu through the OpenSolaris repositories for the purpose, and by turning the platform's newly-introduced Crossbow network virtualization bits to the task of providing the Linux containers with the network connectivity they require as transparently as possible.

Executive Editor Jason Brooks can be reached at

As Editor in Chief of eWEEK Labs, Jason Brooks manages the Labs team and is responsible for eWEEK's print edition. Brooks joined eWEEK in 1999, and has covered wireless networking, office productivity suites, mobile devices, Windows, virtualization, and desktops and notebooks. JasonÔÇÖs coverage is currently focused on Linux and Unix operating systems, open-source software and licensing, cloud computing and Software as a Service. Follow Jason on Twitter at jasonbrooks, or reach him by email at

Submit a Comment

Loading Comments...
Manage your Newsletters: Login   Register My Newsletters

Rocket Fuel