Jeff Jaffe’s Blog

Usually I use my blog soapbox to inform. Novell has quite a number of exciting technologies, products, and initiatives, and our stakeholders should understand the context of our work.

Today, I also include a plea. One of our most significant projects – the Linux kernel driver project – was launched earlier this year in a grassroots, low-key fashion without much fanfare. It is a significant project which will grow the adoption and utility of Linux. Without any help, it has attracted a community of hundreds of participants. I would like to get thousands. So my purpose is to describe the project and hopefully encourage participation.

Background

We all know the importance of drivers for an operating system. There is a huge ecosystem of devices that are created. Thanks to the of plug-and-play that is available in modern Linux desktops such as Novell’s SUSE Linux Enterprise Desktop 10, it is quite easy for these devices to connect to Linux. Still, there is the problem of creating the device driver.

Many enlightened vendors will simultaneously build drivers for Linux and Windows – particularly most of the larger vendors. Unfortunately, some vendors don’t want to spend twice, or smaller vendors may lack the skills. To make matters worse, some don’t even provide the hardware specifications to facilitate driver creation by others.

Empowerment

Last January, Greg Kroah-Hartman started a focus on this problem. He told people that he would form a team in the community to create free open source drivers. He didn’t ask or get permission – he just did it. Novell takes little credit for this – other than the fact that we stayed out of his way!

In any case, his idea has been extremely popular. He now has a team of over one hundred developers that are engaged in this mission. There have also been some nice press reports – see here and here.

Recently, Greg asked to work on this full time. This was an easy decision for Novell management. After all, one of our overarching goals is to foster the growth of Linux. Few ideas have been as impactful as this. Done by a respected community member, a project respectful of community norms, a project that is abiding by the Linux kernel’s GPL license requirements.

The plea

This is a truly noble effort. Anyone who has the skill and the time should consider joining this effort. It helps all: Linux, users, and the device manufacturers. Also, as compared to other projects – where one’s approach to innovation might run counter to a different perspective (e.g. a design detail of the kernel that might affect other parts of the kernel) – drivers are relatively unencumbered. An innovator can build some drivers, and as long as the process respects basic capabilities, it requires little negotiation. In that sense it should be extremely satisfying.

Often I run into talented young programmers who would like to begin participating in the community, but don’t really know where to start. This would be a good place! An innovator can provide immediate value, demonstrate his or her skill, and learn how to participate in a relatively self-contained environment. Today we have hundreds of participants. Greg will probably shudder when I say that I would like to see thousands. Greg will ask – how will I keep track of all of this? I don’t know. But it is a problem worth having!

The vendor side

Once we have this team of thousands of developers we want to make sure that we are addressing the important driver problems. If you are a vendor, please visit this project and let us know your needs. If you are a user, please let us know what your most pressing Linux driver requirements are, so we can start a dialog with the vendor.

Earlier this year, I provided a preview of the goodies that were coming in Open Enterprise Server 2 – with a focus on virtualization, interoperability, NetWare migration, and training. With today’s announcement of the General Availability of OES2, I won’t repeat these themes, but rather I will drill down on how events of the last several months have intensified the value provided by OES2.

Growth of the Linux ecosystem

The basis of our strategy to move NetWare services, capabilities, and customers to OES has always been the recognition that the Linux ecosystem is huge. Our customers immediately get a shot-in-the-arm in terms of device drivers for new disks and printers, processor enhancements, multiprocessor capabilities, etc.

This advantage has been propelled to new heights through the efforts of one of Novell’s Linux gurus – Greg Kroah-Hartman. Greg is the inspired and inspiring grass-roots leader of the Linux driver project since he started it earlier this year. This project aims to substantially grow the number of free device drivers for Linux. It was widely reported earlier this week (see here, here and here) that only a few months later this project has gained over 100 developers across the industry.

I’ll have more to say about the broad implications of this important project for Linux in a future post. The simple point here is another proof point of our strategy of moving NetWare to a Linux base. With the growth of Linux hardware support, NetWare users using OES can be sure their systems will continue to run successfully on new hardware in the future.

Quality

We have been conscious that in moving NetWare workloads to Linux, we must ensure that existing workloads – many of which have been running for years – work perfectly in the new environment. As a result, we developed a beta program which would not relax until we achieved total satisfaction from our beta customers. On announce date, we are prepared with 100 beta test sites sufficiently satisfied that they are ready to move to the product version in production. We have run over one million test cases, and performance has increased 15 percent over prior releases.

Our desktop to data center strategy continually brings Linux to new environments. We have talked much about our innovative desktop and mission critical server and virtualization support. Now Linux supports the world’s most complete file, print, and storage management infrastructure – with quality.

Virtualization performance stress testing

One important aspect of the quality and performance testing was performance testing of NetWare virtualized over SUSE Linux Enteprise Server using Xen.Recall that Xen is the open source virtualization technology that Novell first brought as a capability within Linux. Novell is also the leader who stress-tests this technology so that it competes with existing proprietary technologies such as VMWare.

OES 2 took this testing to new levels. High intensity file performance required for a classic file server system like NetWare is not trivial to support in a virtualized environment. We spent a tremendous effort to tune, refine, and optimize our solution. This required infrastructure in the base operating system (which was added to SUSE Linux Enterprise Server 10 SP 1 a few months ago), as well as exploitation in OES 2.

The result: for existing NetWare customers – an ability to migrate to OES and get good function, performance, and devices; and for the Xen community – a maturing of Xen virtualization technology.

Focus on cost for the I/T manager

The major focus for OES 2 has been to move to Linux and enjoy the Linux ecosystem. I worry that it might be overlooked that OES 2 has added function – focused on reducing the cost of computing for existing and new NetWare (workgroup) customers.

This is done by allowing customers to reduce their spend on the two most expensive hardware elements: processor and storage.

Processor cost reduction is achieved via consolidation. With 64 bit support, customers move their servers to more cost effective, higher performance systems. Further, they can consolidate multiple servers and multiple applications on a single server through virtualization. Virtualization in turn goes further with 64 bit processors!

Virtualization also helps storage efficiencies. But the greater storage efficiencies come with Novell’s innovative Dynamic Storage Technology which automatically “ages” rarely used storage and migrates it to lower cost devices. Consider the fact that about 80 percent of the data being backed up to expensive storage devices is stale. Think about costs of expensive storage at one-fifth of current capacity.

Summary

OES 2 is the culmination of a several year vision that Novell has had to move NetWare workloads to Linux. The confluence of several factors of the past few months – the Linux driver project, a focused beta program, virtualization stress testing, and Dynamic Storage Technology – has brought this promise to new levels.

Install and enjoy!