Solaris 10 - a Solaris engineer’s personal perspective

Date/Time:
Wednesday 12 October 2005

Speaker:
Gary Pennington of Sun Microsystems Ltd

Description:

Gary worked as a code writer on Solaris and outlined the principles behind the development.

One of the driving principles, he explained, behind the major changes in 10, was the emphasis on improved reliability coupled with unlimited scalability. During the development phase, there were daily updates to the code with the company.

It had to be as fast as, or faster, than any competing operating systems and also recover from non-fatal hardware problems. One nice touch he demonstrated was the error messages, which explained in ‘English’ what had caused the problem, rather than a dump of codes to the screen or to file which have then to be analyzed to give an indication of the cause of the problem.

The operating system was also designed to be secure ‘out of the box’ rather than security setting having to be applied after installation.

He then explained that the O/S was 80% non-hardware specific, allowing systems to be quickly recovered in the event of a major hardware failure when like for like kit was not available or had been commandeered for other business critical systems.

After explaining the principles and goals behind Solaris 10 development, he then went on to talk about some earlier products and explained that product innovation was the rationale for 10.

Some of the headline grabbing elements were the advanced resource management capabilities of Solaris 10, where resources could be aggregated, partitioned or virtual systems could be created on the product.

It offers dynamic limiting of resources, processing power, memory etc. by project, process or task. It also offers dynamic partitioning, where resources are re-allocated as and when needed, as per rules set up on the system.

So if a process or task was not using all the resources allocated to it an any given time and others were using a high percentage, additional resources would be allocated to those without any operator intervention and switched back if and when needed.

With virtualisation, local ‘zones’ could be set up, each acting as a complete operating system in its own right, totally independent of each other, with its own resources. This would ensure that even in the event of a disastrous user error causing the application to fail, other zones running the same or different applications would be totally unaffected.

He then went deeper and in much more technical detail into further aspects of the new system well beyond this writer’s abilities to explain. However the slides used during the presentation are published on the WYBCS website if you were unable to attend the event and you would like to view.

The presentation was followed by a question and answer session and Gary presented some ‘freebies’ kindly donated by Sun Microsystems to some questioners.