Sun Preps Solaris 10 With 'Predictive' Self Healing

Sun Microsystems has fleshed out an updated release of its Solaris operating system. After deciding upon the structure and nuances that would make up Solaris 10, Sun homed in on supplying its customers with superior value, stability, security, and performance.

To meet its self-imposed checklist, the Solaris product team turned to its technology tool chest which includes: N1 grid containers, a new "predictive" self healing framework, process rights management derived from Trusted Solaris, and a new bottleneck hunting technology dubbed dynamic trace.

As Sun treks down the road toward Solaris 10, developers must buy into the company's agenda to reap the most essential benefits out of the upgrade, and for Sun to capitalize its continued investment in the OS. To ferry this process along, pre-release bits of Solaris with new features are available to developers at regular monthly intervals through Sun's Software Express program.

One of Solaris 10's core design aims is to reduce licensing costs. Sun has turned to N1 Grid containers, which it refers to as the next generation of software partitioning, to allow a single system to be divided up into 4000 different partitions per copy of Solaris. In addition to storage virtualization, Sun has also provided customers the capacity to divvy up processing power in the same way, on a machine basis.

Like mainframes, grid containers function as separate machines inside a single box, dynamically balancing and restraining business processes. This arrangement can provide savings in cost over virtual machines that require multiple operating systems and licenses to be used for each partition.

Although N1 Grid containers consolidate resources into a single copy of Solaris, Tony Iams, a senior analyst for D. H. Brown Associates cautioned that fault management becomes much more of a focal point due to this very fact.

"If something happens to the master operating system, it will bring down all of the containers," Iams told BetaNews.

Solaris Heal Thyself

To combat this threat from within, Sun has devised an extensible "predictive" self-healing framework to probe hardware and software environments and intervene before a system becomes "ill." The self healing functions of Solaris will, for example, copy data from failing memory to another area of memory. Solaris can also determine the correct order in which to restart services, and report back to administrators when there is a problem.

Sun's self healing is set up into three separate layers: Kernel telemetry at the operating system level, a hardware monitoring engine, and a fault manager that sits between the two. The end result is a registry that keeps track of all open services and their dependencies.

To make all of this work, and to ensure an acceptable level of stability in an N1 grid container environment, both Sun and its developers must populate the framework. Sun is busy readying engines to plug into its self healing architecture, but must also convince ISV's to integrate services into the fault manager. Sun is scheduled to release new engines each quarter.

Graham Lovell, Sun's Director of OS Marketing and Systems Software Marketing touted Sun’s self healing framework while saying that its approach differs from rival IBM's, whose implementation Lovell deems as, "bolted on concepts," and "lacking the predictive element and customizability offered by Sun."

In response, an IBM spokesperson told BetaNews, "IBM is unmatched in its research and development of self managing capabilities ($500M a year), and delivering products, services and alliances (i.e. Cisco) that are making this a reality."

Big Blue first applied the concept of autonomic computing to its eServer product line under the guise of project eLiza. Since that time, an IBM spokesperson told BetaNews the company has incorporated over 400 product features into 36 distinct products.

The spokesperson pointed out that IBM sees the full spectrum of autonomic systems as self-configuring, healing, optimizing and protecting.

In the case of Solaris 10's self healing capabilities, D. H. Brown's Iams is firmly in the Sun camp. Iams views Sun's self healing approach as more comprehensive than IBM's due to the fact that Solaris looks at the actual services that are running.

"IBM is a lot more at the hardware level," said Iams.

Locking Down Solaris

To prevent hardware powered by Solaris from falling into the hands of hackers, Sun has borrowed process rights management, or role-based access control, from Trusted Solaris - a military grade version of its operating system. Essentially, administrators can give individual processes the minimal rights needed for a task.

Sun believes that eliminating unnecessary super-user permissions will thwart hackers who are able to penetrate a system and stop them from doing further damage once inside.

Given the security problems that it is facing, "Microsoft would want to have it," a Sun spokesperson told BetaNews.

A free utility for Linux called Sudo (superuser do) offers a similar method to lock down Linux boxes.

Speeding up Solaris

Although security is a primary focus of Solaris 10, better performance is firmly embedded within the brick and mortar of the OS.  Sun claims that users will experience between a three to thirty percent boost in speed.

On the hardware side, Sun has dished out new UltraSPARC IV processors to boost performance, as well as issuing a new line up servers. Other improvements include support for multiple thread CPUs and new instruction sets. In addition, Sun has rewritten its TCP/IP stack and undertook the usual code optimizations that come with a new product release.

What really makes Solaris 10 capable of outperforming its predecessor is a new feature known as Dynamic Tracing.

Dynamic Tracing is a tool to analyze and diagnose bottlenecks that stem from rogue applications or poor system design. Sun designed this feature to work in a production environment under normal conditions.

"There is nothing you can do (with Dynamic Trace) to cause a system to crash," Sun's Lovell told BetaNews.

Although the feature is turned off by default, over 30,000 probes extend deep into the heart of the operating system. A special scripting language can be used to write complex diagnostic procedures.

Sun expects to ship Solaris 10 by year's end. Pricing has not yet been announced.