Solution: IBM uniVerse on Linux

Several years ago, a manager of our company voiced issues that he felt were major obstacles in recommending Linux as a platform on which we might install. Recently, several clients have requested the consideration of Linux as a viable installation option and the email that contained his concerns was forwarded to me, and I was asked to respond to the original concerns. Before responding, I did a little research and found some very telling information at the following sites: The story of the Linux Kernel by Linus Torvalds, its creator, and Microsoft Windows NT Server 4.0 versus Unix which is an article telling the benefits of unix in general over the windows environment. The former explains why Linux is so much faster than any other operating system, including other Unix platforms, and the latter includes specifics dealing with Linux and how cost-efficient it is.


Here are my thoughts on Linux/uniVerse from the perspective of addressing the concerns voiced by management. Since these concerns were stated several years ago, we may find, upon closer inspection and a revisit of the issues, that we can propose a viable and acceptable alternative upon which to run the legacy product.

Issue 1: the dynamic nature of open source and specifically the Linux kernel.
A: All kernels change with each new release. The fact that Linux changes are open source causes them to be made more rapidly. This is positive because fixes are available almost immediately. Another major advantage of the open source kernel is this: If we have an issue that we consider major, we can submit to the open source community our recommendations and affect a change in our favor that would become permanent. Future "tuning" would become unnecessary for that issue. We would never have to be directly involved in developing the kernel. Currently, when we have to make OS level changes, we have to rebuild the kernels of HP, IBM and other platforms. This won't be affected by whether or not the kernel is open source or not, other than those changes we regularly have to make in HP and IBM kernels becoming permanent for future releases. With the Linux kernel, we could adjust the settings we need to, build the kernel, and use it for all our applicable installs. Build the kernel once, distribute it many times. That's not only unfeasible with our other vendors, but impractible and would be disallowed by them. We could also just have HP or IBM or Varmark do it for us, since they now support Linux the same way they support their other OSes. This becomes an argument FOR rather than AGAINST Linux.

Careful consideration, however, has been made by our vendors, as to which distribution of Linux to choose. They have made the choice of RedHat Linux because it has been packaged as a solid, reliable server. Over 80% of all servers installed worldwide during 1999 were Linux servers. The vast majority of those were RedHat servers.

Issue 2: qualifying device drivers.
A: The Linux kernel is a macro-kernel (See http://www.linuxworld.com/linuxworld/lw-1999-03/lw-03-opensources_p.html for the benefits of a macro-kernel), unlike all other "commercial" vendors, such as SCO, HP, IBM, etc. This means that all the drivers for peripheral hardware are NOT part of the kernel. This makes a much faster, more stable and more robust kernel. Hardware devices for which drivers are necessary are added to the kernel as loadable modules, similar to the way Netware uses the NLM. This allows for further streamlining of the kernel, adding to speed and interoperability. Device drivers for peripheral hardware is becoming a higher priority for vendors. To alleviate concerns for device driver issues, it is recommended that hardware be assembled from easily obtainable and readily available inventories. The performance of Linux alleviates the perceived necessity of other operating systems of assembling the computers from leading edge components. Using second generation parts becomes not only desirable, but adds a significant level of trust in hardware that has already been proven "in the field" so to speak.

Issue 3: application tuning and performance, distribution, patches to the OS, etc.
A: Issues dealing with performance are highly in favor of Linux as it has proven to outperform every other OS on which we currently install. The distribution of the OS and patches thereto are done by the open source community, various vendors, such as RedHat, Debian, etc. and once a single source is chosen, we can burn our own CDs for cost and distribute the kernel as we see fit. This is a tremendous advantage because we can address our concerns once, burn that into our distribution master CD and then not have to re-tune every client we sell.

Issue 4: Our company should focus on the continued development and maintenance of our legacy software, rather than diluting it with Linux OS support and maintenance for which there is little to be had in return.
A: We currently do no OS research and maintenance for other platforms and have no cause to believe we would need to do so for Linux. There are more vendors who support Linux than any other OS on which we currently install. In this regard, Linux becomes just another platform we might offer without the added cost of maintaining it. Varmark, HP, IBM and others offer support on a 7X24 basis. We don't do OS support now. We wouldn't have to do it for Linux.

Summary: Linux is virtually free. It has an unlimited user license. It has proven to be incredibly fast, versatile, reliable and predictable. There is potential to serve our clients a product that outperforms every other platform on which we install. It might be considered prudent to offer it as an alternative to NT/2000, HP, IBM or other high-cost, high-maintenance platforms.