Ian Murdock on OpenSolaris… And Beyond
I paid great attention to Ian Murdock’s talk at LugRadio Live USA 2008, as he’s an important person at Sun in terms of the open source community, and I’m community facing as well. It was also the first time I got to meet Ian (after his talk), and we hit off a conversation really quickly. I look forward to working alongside Ian more… Now to the talk notes.
A Bit About Ian
- Linux user, developer, and advocate since 1993
- Founder of Debian, co-founder of Progeny
- Joined Sun in March 2007, Chief OS strategist (launched Project Indiana), and now since February 2008 is VP Developer and Community Marketing
What’s a Linux guy doing at Sun?
- When people say they want Linux… they don’t actually mean that they want Linux. You don’t only want the kernel, but you want the userspace as well. You really want a distribution, not Linux itself. Ian is more of an open source guy. So create a business model behind OpenSolaris?
- Solaris ships all userland utilities you’d expect from Linux. GNU Utilities vs. UNIX Utilities. Where the real differentiation is, is Linux Kernel vs. Solaris Kernel.
- Its the kernel where you drive change for the OS. Look at ZFS, Dtrace, containers, etc.
- Linux has multiple configurable platforms with expanding ecosystems. OpenSolaris, is an integrated operating system, with binary compatibility.
- Where Sun wants to go with OpenSolaris? Have binary compatibility in the core platform, and have the expanding ecosystem in it.
- Project Indiana has Solaris innovation with the “distribution” model of the Linux world, i.e. have package repositories. Also, to close the “familiarity gap” so that there is a GNU userland, a package management system, etc. And of course, focus on the unique Solaris capabilities – upgrade rollback via ZFS snapshots, AMP stack with integrated DTrace probes (MySQL in OpenSolaris Developer Preview 2 already has this capability), binary compatibility, etc.
- Remember, people deploy what’s accessible to them. This is a whole new deployment model. This is how Linux got into the Enterprise. This is the reasoning behind OpenSolaris… Solaris itself is great in the Enterprise, but lacks a user community. Follow the Linux (and largely, Microsoft) model…
- One Solaris Platform, two delivery mechanisms – Solaris: enterprise class, support, long release cycle (3-5 years); OpenSolaris: latest Solaris innovations, short release cycle (6 months), much easier to use, network economy, support from Sun also
- Adoption-Led Market is what its at now. Users become customers. Give users something of value. That’s the basic business model at Sun, around open source.
- Volume drives value! New consumers/communities on the network will drive infrastructure demand. Reach the developer community through massive volume, get it in use as many places as possible, then sell software/storage/servers/services/etc.
- Just spent $1 billion on MySQL. Databases are often deployed on other middleware, say identity management. Sun sells identity management software.
- Developer platforms of choice are moving up the stack. Traditional Sun platforms are increasingly invisible infrastructure. New platforms remove barriers to entry and hide complexity so developers can focus on getting to market quickly.
- Remember, no one “owns” the web. The Web 2.0 world is like open source in the 90s.
- Hot new platforms are silos – you write your application to Google APIs or Amazon APIs and there’s only one place to deploy it. Sun is a leader in open standards for 25 years, how do you take this leadership into the new platform world? One needs the ability to deploy on a number of different platforms, and take your application from one to another. Sun strategy?
- Where do down the stack technologies, like operating systems, fit? Do they still matter? Yes.
- Solaris 10 is free today (you just have to register to get it)
On the packaging system
- OpenSolaris package system is new. Needed to build own for a few reasons, despite studying apt/yum/conary.
1) platform is very different from Linux; use ZFS and you get rollback, and this is Solaris specific
2) zones/containers, single shared kernel, multiple application environments, which is intelligent for resource sharing, and the package system is fundamentally aware of zones
3) an opportunity to innovate; Sun’s all about innovation. Package managers haven’t changed much in over 10 years (apt, yum)
4) the notion of customisation, i.e. have a version control like facility for patches, etc. so that changes can be reconciled when you rebase with upstream, is something the new package management system can use. This is kind of like Conary?
5) All customers tend to customise, even (especially) in Enterprises. IPS helps them (refer to 4)
Questions & Answers
- OpenSolaris and Solaris, how is it in sync? Codebase is in sync, generally. Solaris 10 is one codebase, and there’s a development version (i.e. the next release). There is a Solaris “train” and an OpenSolaris train now. How will these two trains come together? Solaris Next based on OpenSolaris? OpenSolaris with Enterprise support? Questions that are being asked now, and there’s no real decision yet. Wait till 2-3 OpenSolaris releases first, before finding out more.
- What is Sun doing to make OpenSolaris work better in a virtualized environment? I.e. working better in Xen, QEMU, VMWare, etc. Linux naturally has better device support than OpenSolaris, so virtualization is very important. Mentions the innoTek purchase, so that VirtualBox could be used for bundle delivery.
- Is Linux and OpenSolaris in par, security wise? Solaris has a slight edge at this point… They’re fairly comparable, but there’s more maturity in the Solaris product.
- OpenSolaris on PowerPC – unknown status, there’s been some work at Sun Labs to port it to PPC, you can boot it and do some basic things, but unsure if its ready for users or not.