Project Live* Presents a Huge Business Opportunity

By Al Riske

13.June.08 - For Olaf Manczak, it starts with the numbers. He looks around and sees there are thousands of mainframes, millions of servers, and billions of consumer devices.

Then it becomes a question of complexity. To operate a datacenter you have to be highly educated and well trained. Consumer devices, on the other hand, are easy to use.

This is true, Manczak notes, even though consumer devices -- mobile phones, wireless routers, TV set-top boxes, and more -- contain embedded computers running some pretty sophisticated software. In fact, the operating system inside is often the same as (or derived from) what you'd find in many servers.

"Yet, with all these consumer gadgets one never has to go through the tedious process of installing software, creating configuration files, or applying updates," he says. "At most, one has to upgrade the firmware."

After much thought, he realized that a new vision was needed to overcome the barriers between the datacenter and easy software deployment.

"This new approach bridges the simplicity and ease-of-use of the firmware model with the modularity and customizability necessary in multi-purpose, multi-vendor systems," he says.

With datacenters growing about 15 percent each year (and with the trend toward virtualization there are now multiple software installations per physical server), the old way of software management is reaching its limits. So, solving this scalability problem represents a significant business opportunity for Live* and Sun Services, Manczak says.

Potential customers seem to agree. When he shares his vision, Manczak sees two contrasting responses:

A few are resigned to the status quo. They accept a reality that includes millions of dollars and hundreds of man-hours spent on software deployments, using both commercial and ad-hoc automation tools.

Many others are ready for a better solution, especially those running massive high-performance computing clusters. "It would solve all of my rollout problems!" said one member of the HPC community.

"In all the consumer devices, what drives the design is ease of use and reproducibility -- every device behaves in the same way," Manczak says. "What drives the design of servers is that people need to be able to customize them -- the actual use of the server is defined by the end users, not by the vendor. Further, datacenters depend on multi-vendor solutions, so we must support heterogeneous environments."

The objective of Live* is to bridge this gap between the firmware model that drives consumer devices -- a proven, easy-to-use software model for deployment and maintenance -- and datacenter servers, where additional requirements are critical.

"What we propose is something that has all the benefits of firmware. It combines plug-and-play images of software -- which gives you instant and modular deployment -- with a systematic method for applying configuration settings. As a result, deployments can be performed without modifying the original software as we do today on datacenter servers," he says.

"We're taking an iterative approach; our current focus is on very large horizontally scalable environments where people deploy hundreds of thousands of almost identical machines with almost identical software. In such a context the only difference is the configuration of each machine. This is typical in the world of HPC or virtual machine-based cloud computing."

Manczak has experienced the weaknesses of this approach.

"This works great for the initial deployment, but falls apart when the administrator goes to apply updates," he says. "The process is certainly nowhere near as clean as the firmware model of software upgrades."

Live* supports the firmware model because it allows software to be stored on separate images, and for these images to be unified as a single file system to the user. Want to upgrade to a new version of an OS or database? No problem. Just replace the appropriate image with an updated version. In this way, the time-consuming process of repeatedly deploying new software is avoided altogether. Further, custom configuration changes made by the user are still present after an upgrade, because they are stored separately.

It is this method of building a virtual namespace out of both software and configuration settings that sets Live* apart from existing approaches.

Ninety-nine percent of files exist in a unique namespace: the files of a database, for example, are completely different from the files of an operating system. So it's a relatively easy matter to overlay the two.

"That's what we do in name-space unification," he says. "We don't have to understand what's inside these files; we just concatenate the name spaces. But there is this other one percent, yet that one percent is critical. It is structured data. You need to understand the syntax to merge this data."

So Manczak is taking a page from biology.

Project Live* provides a sort of DNA, a universal data structure that everybody can read.

"The thing we did was to think about configuration the same way as it works in biology -- that there is this kind of DNA. Every cell only finds a few relevant genes there, but every cell can read the whole structure. It just uses particular enzymes to extract what it needs," he explains.

"We use what we call generators, which are implanted in the [software] images so they can extract from this structure whatever they need to generate the configuration file. But the configuration structure is non-redundant. It's fully normalized. As with DNA, there are no two genes that describe the same feature, so it's very much the same concept."

"I think Sun has a unique opportunity to create a community of vendors who could support this model. The customer comes to us with a list of the software it needs, and we supply the images that are certified to work together. There is an incredible business opportunity for anybody who verifies, or can teach others to verify, compatibility. Sun has incredible competency in this area. We have the knowledge and ability to physically verify that things are correct, and we can proactively teach people how to do it right, showing them what are the rules, the policies, to ensure you're not breaking compatibility."