Storage Solutions
Description
This initiative will explore advances in storage systems and protocols and how they may be used to improve ITS' ability to access, back up, restore, and manage data. Fibre channel and IP-based Storage Area Networks (SANs) are solutions for many of the same kinds of issues, but are at different price/performance points. ET will investigate under which conditions it makes sense to implement IP SANs rather than fibre channel SANs. Other areas of interest are hierarchical storage management and storage used for high performance computing applications.
The primary intent of a SAN is to support data communications between computers and storage elements and among storage elements. Its requirements differ from general-purpose data networks and are commonly designed around proprietary architectures that require dedicated fibre. Closely related to SANs are IP storage and network-attached storage (NAS). The IP storage IETF working group is developing standards for encapsulating storage protocols, such as SCSI (e.g. iSCSI) and Fibre Channel, for transport over IP-based networks (FCIP). This may lead to an open systems SAN architecture. NAS is just as the name implies. NAS devices may be attached to a SAN or LAN. There also has been a convergence of SAN and NAS, blurring some of the distinction between the two. As areas of interest, ET is currently investigating iSCSI and FCIP; please refer to the storage working notes for more information.
SAN and NAS have applications to enterprise file systems and high performance disk storage. Some products provide a common name space and data management functionality desired in a an enterprise file system but still lack much of the other functionality. Integrating with enterprise-wide security is often an afterthought as well. Any system or application with large data needs such as high performance computing clusters, digital libraries, and distributed file backup systems can benefit from this technology.
Some questions ET would like to answer through investigations are: As the performance and number of nodes in a high performance computing cluster increase, do any products scale without becoming a bottleneck in the cluster? Is it feasible to operate SANs over a shared connectivity infrastructure (particularly, Penn State's Integrated Backbone)? If so, what accommodations are required? If not, given the cost and scalability problems of dedicated inter-building fibre, what are the alternatives? There often is talk of improving performance by increasing network bandwidth, but where are the real bottlenecks (e.g., spinning storage media access, bus speed)?
Publications
N/A
Resources
The Storage Networking Industry Association (SNIA)
IETF IP Storage Charter
Byte and Switch - The Global Site For Storage Networking
TechWorld Storage Resources
Storage Management Topics
Primary ET Contact
Mike Burns burns@psu.edu
Collaborators
N/A
Expected Deliverable(s)
Report of investigations and evaluations.
Level of ET Involvement
Investigation and evaluation of products.
Initiative Start Date
May 2003