Efficient, reliable storage of electronic data is a growing challenge of the digital age. Ceph, an open-source file system developed by computer scientists in the Baskin School of Engineering at the University of California, Santa Cruz, offers an innovative approach to the data storage challenge.
A file system is the software that organizes and manages computer files and provides access to them on a data storage device such as a hard disk (or on an array of devices in a large storage system). Ceph is a distributed network file system, the type that allows multiple clients to access files stored on remote servers. It is designed to manage vast amounts of data with ease, meeting the growing demand for data storage on the scale of petabytes. (A petabyte is one quadrillion bytes, or a million gigabytes--enough data to fill 223,000 DVDs.)
According to a recent study, the amount of digital data created in a year grew by 62 percent from 2008 to 2009, and will increase 44-fold by 2020 (IDC "Digital Universe Study"). The study also reported a growing gap between the amount of digital data being created (nearly 800,000 petabytes in 2009) and the amount of available storage.
"Everybody needs a solution to the problem of how to handle all this data. Our goal with Ceph was to build a scalable system that would allow storage of petabytes and beyond," said Scott Brandt, professor of computer science at UC Santa Cruz and director of the UCSC / Los Alamos Institute for Scalable Scientific Data Management at UCSC's Baskin School of Engineering.
Although Ceph is still in development, it just received a big vote of confidence from Linus Torvalds, who included Ceph in the latest version of the Linux kernel (Linux 2.6.34). Brandt's graduate student Sage Weil did much of the development work on Ceph for his Ph.D. thesis, and he has continued working on it since earning his degree in 2007.
"Having Ceph in the Linux kernel makes it much easier for people to use, so a lot more people will be testing it now and contributing to the project," Weil said.
Ceph is designed for high performance, scalability, and reliability. "When you reach the scale of a storage system with 10,000 disks, failure of individual components becomes the norm, so the file system has to be able to monitor and deal with that. With Ceph, any node can fail and the system heals itself," Weil said.
A Ceph file system can be seamlessly expanded simply by adding new servers to the system. Ceph automatically distributes data across the new servers. All data are replicated across multiple storage devices, and if any device fails, the data are automatically re-replicated to other devices.
Ceph is also designed to dynamically adapt to changing workloads. It can handle scenarios in which thousands of clients suddenly access the same file or write to the same directory.
"A Ceph storage system can be both big and fast, because all the operations are parallelized," Brandt said. "We made it open-source, so the source code is out there, and people are starting to be interested."
Brandt, Weil, and faculty and students in UCSC's Systems Research Lab and Storage Systems Research Center have been working on Ceph for about 8 years. The project was primarily funded by a Department of Energy grant from the Lawrence Livermore, Los Alamos, and Sandia National Laboratories. Data storage is a big concern for the national labs, where the most powerful supercomputers in the world run simulations that generate vast amounts of data.
The project has also received industry support from companies such as Yahoo and NetApp. Weil, who founded a web hosting company (DreamHost) before pursuing his doctorate at UCSC, said companies like his need affordable, scalable storage systems for customer data. But he said his main motivation for continuing to work on Ceph is simply a desire to see the project succeed.
"I want it to be something people can use and not just a research project," Weil said. "It began as pure research, and now it's actually made it into the real world."
