STICS: SCSI-to-IP cache for storage area networks

Abstract

7Data storage plays an essential role in today's fast-growing data-intensive network services. New standards and products emerge very rapidly for networked data storage. Given the mature Internet infrastructure, the overwhelming preference among the IT community 9 recently is using IP for storage networking because of economy and convenience. iSCSI is one of the most recent standards that allow SCSI protocols to be carried out over IP networks. However, there are many disparities between SCSI and IP in terms of protocols, speeds, 11 bandwidths, data unit sizes, and design considerations that prevent fast and efficient deployment of storage area network (SAN) over IP. This paper introduces SCSI-to-IP cache storage (STICS), a novel storage architecture that couples reliable and high-speed data caching 13 with low-overhead conversion between SCSI and IP protocols. A STICS block consists of one or several storage devices and an intelligent processing unit with CPU and RAM. The storage devices are used to cache and store data while the intelligent processing unit carries 15 out the caching algorithm, protocol conversion, and self-management functions. Through the efficient caching algorithm and localization of certain unnecessary protocol overheads, STICS can significantly improve performance, reliability, and scalability over current iSCSI 17 systems. Furthermore, STICS can be used as a basic plug-and-play building block for data storage over IP. Analogous to "cache memory" invented several decades ago for bridging the speed gap between CPU and memory, STICS is the first-ever "cache storage" for bridging 19 the gap between SCSI and IP making it possible to build an efficient SAN over IP. We have implemented software STICS prototype on Linux operating system. Numerical results using popular benchmarks such as vxbench, IOzone, PostMark, and EMC's trace have shown 21 a dramatic performance gain over the current iSCSI implementation.

SPEK: a storage performance evaluation kernel module for block level storage systems

“…It enables clients to discover and access SCSI devices directly via the mature TCP/IP technology and existing Ethernet infrastructures. Previous work on iSCSI mainly concentrated on performance evaluation and potential improvements [15], [24], [31], [32]. By using our SPEK, we have evaluated the degraded performance of a popular iSCSI implementation and observed that the performance of iSCSI degrades dramatically in case of faults, but it can rapidly recover after such faults are removed or corrected.…”

“…We have recently proposed a new storage architecture: SCSI-To-IP Cache Storage, or STICS for short [15]. The purpose of STICS is to bridge the disparities between SCSI and IP so that efficient SAN systems can be built over the Internet.…”

“…In this way, it acts as a storage filter to discard a fraction of the data that would otherwise move across the Internet, reducing the bottleneck problem imposed by limited Internet bandwidth. More information about STICS can be found in [15].…”

“…We found that the iSCSI target is saturated at 29.007 MB/s using two test clients. Since it is a software iSCSI implementation, the TCP/IP and iSCSI protocol overheads [15] are the main reason why the target saturates rapidly. The CPU utilization of the iSCSI target is consistently larger than 90% when using two test clients and approaches 100% when using three test clients.…”

“…By generating configurable workloads and injecting configurable faults, users can grab the dynamic changes of potentially compromised performance and therefore quantitatively evaluate system performance of a measured SAN. To demonstrate how SPEK works, we have performed several tests on direct-attached storage systems including single disks and disk arrays and two storage area network systems: an iSCSI-based [14] SAN and a STICS-based [15] SAN.…”