Data preservation is critically important to the Arctic Data Center. We recognize that data preservation is difficult, both for technical and non-technical reasons. We have developed this data preservation plan to be explicit about how the Arctic Data Center ensures the long-term preservation of the data entrusted to the repository. Key to this plan is our belief that no single organization can possibly provide sufficient institutional stability to guarantee multi-decadal preservation, and that partnerships among committed archives are necessary for successful data longevity. The guiding principles for our preservation plan follow.
- Preserve the bits.
- The primary mission of the Arctic Data Center is data preservation and data access. High-quality data management is essential to data preservation. All submitted data and metadata are reviewed and edited before acceptance to ensure high-quality data products are available to the research community. Data are managed following best practice for systems administration at UCSB’s North Hall Data Center (NHDC), which complies with a subset of Tier 1 ANSI/TIA Data Center Standards.
- Open science, open standards.
- Wherever possible, we utilize and encourage the use of open standards for representation of data and metadata, and for provisioning of services. Metadata are managed in the open Ecological Metadata Language (EML), and we encourage researchers to provide data using open data formats such as ASCII CSV for tabular data and PNG / JPG for imagery. Open formats support accessibility of the data in the future, even in the face of large software changes. In addition, the repository supports open access via the DataONE REST API for external groups to be able to access all components of the system.
- Replicate data and metadata.
- All metadata and data are replicated at geographically distinct locations including 1) DataONE replication nodes and 2) at the NOAA National Centers for Environmental Information. An archival copy is also made periodically on the Amazon AWS cloud service. Replication is automated, and occurs any time that a change to any file in the system is made. Replication assures that data and metadata remain available even in the case of unplanned local system outages (such as a regional-scale fire or earthquake event), and provides for higher-performance access to data from multiple replica sites.
- Strong versioning.
- Following the Force11 Data Citation guidelines, every version of every object in the system is assigned a unique identifier which is used to track that version of the object and relate it to earlier versions. For data packages, a DataCite DOI identifier is assigned and registered upon publication. All updates to objects are tracked, and old versions of data packages and data objects remain accessible even after an update, ensuring that any citations to the original versions of data can continue to be resolved to exactly the version of the data that was cited. Older versions of data packages are clearly marked, making it easy to navigate to the most recently updated versions, and search systems point users to the most recent version.
- Frequent auditing.
- All data, metadata, and other objects in the system are provided with a checksum that can be used to validate that the contents of the object have not changed over time. The Arctic Data Center participates in the DataONE federation, which audits all objects to ensure that the current copy of the object matches the original authoritative copy. In addition, DataONE checks all replica copies to ensure that they continue to persist and have matching checksums. This periodic auditing ensures that accidental content corruption due to disk, network, and human error are detected and remedied in a timely manner.
In addition to this preservation plan, we recognize that over long time periods spanning many decades, it is extremely difficult to predict and sustain funding for single institutions. Our replication policy ensures high-availability during normal operations, but also provides security should NSF’s investment in data archival wane. Should the main Arctic Data Center fail to be sustained, then the management of the Arctic Data Center will work with our partnering institutions to ensure that the archival replicas that they hold continue to be preserved and available to the scientific community. This will likely mean that NOAA’s National Centers for Environmental Information would become the authoritative holder of the data until a time when continued support from NSF can be obtained to re-establish operations.
UCSB North Hall Data Center
Primary systems are maintained at the North Hall Data Center, which complies with a subset of the Tier 1 ANSI/TIA Data Center Standards. Networking at 10GbE is via redundant connections to the public Internet and Internet2 through the CalREN2 and CENIC networks. Room UPS power backed by an emergency generator is available up to the 162kW capacity of the data center. Primary cooling capacity is derived from the campus chilled water loop. With the campus chilled water loop subject to regional power outages, secondary emergency cooling is from two locally installed chillers with a total 60 Tons of capacity. When NHDC is on emergency power, the emergency chilled water is used for the UPS room, AHU 5 (campus networking), and chilled water distribution to advanced rack cooling technologies. All racks are mounted on zone 4 ISO-Base platforms for seismic protection. The NHDC is subject to the environmental conditions of the campus and the region. Planned outages involving all equipment within NHDC will be uncommon, but occasionally necessary for certain types of maintenance activity. During such outages, data and metadata from the Arctic Data Center will still be available via our replica holdings, but data submissions will be delayed until normal operations are restored.