Nacun.com

Story Details

Maintaining a family tree is an important tradition in Chinese culture. Nacun.com is a web site that enables families and friends to share and record detailed information for future generations including names, dates, images, and video. Nacun.com initially built its systems on Java applications that ran on the Linux operating system. However, numerous issues prompted Nacun.com to re-evaluate its system architecture.

Whenever a system was shut down illegally or accidentally powered down, critical file system tables were overwritten which made it impossible for the system to boot back up. Although administrators could use operating system tools to check and repair the file system, the restoration process was very slow, sometimes taking many hours — and was not always successful. In addition, users could not access the web site during a restoration.

Nacun.com's IT personnel faced several other challenges. Because of the extremely dynamic nature of the site, administrators struggled to accurately predict storage requirements. As a result, users sometimes encountered insufficient storage levels and were unable to post new content. Additionally, because IT personnel were struggling to meet the rapid increase in service volume, adding new features to the web site and performing extensive upgrade systems tests was not viable. In 2008, Nacun.com participated in several Web 2.0 promotions organized by Sun. Impressed by the openness and stability of Sun technologies, and features in the Solaris 10 Operating System, Nacun.com decided to migrate its existing Linux based systems to the Solaris 10 OS running Sun Fire X2100 M2 servers.

As a result of its migration from Linux to Solaris, Nacun.com's system availability was greatly improved and system administration was simplified. Additionally, the ZFS feature in Solaris 10 helped the organization address storage capacity issues. ZFS enables administrators to assign a pool of storage devices to support a system. Consequently, administrators no longer have to predict how much disk space needs to be added to each server. “With ZFS, our system administrator can be completely freed from the manual and complicated calculations for determining storage capacity,” says Tony Lee, CEO of Nacun.com. “ZFS automatically makes adjustments and scales itself dynamically to meet storage requirements.”

Nacun.com has also taken advantage of Solaris Containers, which have further streamlined system administration and improved system throughput. Solaris Containers allow administrators to create multiple partitions — or containers — on a single system. Each container operates as a standalone machine, running its own copy of Solaris 10 OS and applications. “We redeployed the mail, web, database, and application servers in individual Solaris Containers,” explains Lee. “After testing the new environments, we discovered that the operational efficiency of servers increased by 15 percent compared to our original architecture. System maintenance was also more convenient because the applications in each container can be upgraded and maintained independently.” As a result, system upgrades are less complicated and can be completed more quickly.

For each container, administrators can assign a priority status and allocate system resources. Nacun.com can also create dedicated resource pools — that include processing power, memory, and disk — to support specific containers. By taking advantage of resource allocation and resource pools, Nacun.com can avoid conflicts that can occur when a system process consumes too much processing power, memory, or disk space. “We can establish multiple resource pools to satisfy the needs of containers,” says Lee. “Different resource pools can take advantage of different CPUs, memory, and other resources.”

Nacun.com also uses DTrace, a dynamic monitoring tool in Solaris 10 OS, to instantaneously troubleshoot issues. Not only can DTrace quickly locate the cause of a system issue, but also can automatically fix many issues. Recently, DTrace was able to find a serious scalability and memory leakage problem with a key service application on Nacun.com. “DTrace enabled us to fix the application problems directly in the production environment in less than a day,” explains Lee. “As a result, our web site throughput increased 10 times, and system availability was greatly improved.” Developers at Nacun.com also use DTrace to analyze and optimize the performance of applications, helping developers more quickly identify and fix code issues. As a result, Nacun.com has been able to reduce test cycles, improve the quality of code, and ultimately cut the cost of technical support.