CPA-2008 I/O Projects
- Effective Resource Allocation under Temporal Dependence
- HECFSIO Topics: Next generation I/O
- Keywords: Resource Management
Smirni, Evgenia William and Mary
Temporal dependence within the workload of any computing or networking system
has been widely recognized as a significant factor affecting performance.
More specifically, autocorrelation in flows, is catastrophic for performance.
In a simple single server system, autocorrelation in the arrival
intensities or service demands may result in
user response times that are slower by several orders of magnitude.
In homogeneous clusters where size-based load balancing policies have been proved
optimal for performance, autocorrelation in the arrival intensities of jobs obliterates
any performance benefit of traditional load balancing policies.
In multi-tiered systems, if a service process of any of the tiers is autocorrelated,
then user response times are very high, in spite of the fact that the bottleneck resource
in the system is far from saturation and that the measured throughput and utilizations
in all other tiers are also modest, falsely indicating that the system can sustain higher capacities.
In storage systems, autocorrelation in the arrival or service processes at the disk
level may result in significant user-perceived performance degradation.
This project aims at providing a practical way to characterize and quantify
the performance impacts of autocorrelated flows in systems.
The main focus is on the development of new technologies for resource allocation
that consider autocorrelation as an important characteristic of any stochastic process.
On-line monitoring of autocorrelation provides the necessary information for scheduling
parameterization, making an important step toward the development of autonomic systems.
- HybridStore: An Enterprise-scale Storage System Employing
Solid-state Memory and Hard Disk Drives
- HECFSIO Topics: Next Generation I/O
- Keyworks: Novel storage devices
Urgaonkar, Bhuvan Penn State University
The mechanical movement inherent in the operation of the hard disk poses access speed limits for many workloads and storage systems are consuming increasing amounts of power. Flash memory overcomes some key limitations of the hard disk including faster access to non-sequential data and significantly lower power usage. Encouraged by these advantages offered by flash memory and the recent emergence of high-capacity flash drives, this research will design and evaluate a hybrid system. Named HybridStore, this system will exploit complementary properties of these two media to provide improved performance, service differentiation, and thermal/power behavior in enterprise-scale storage. HybridStore will comprise a dynamic data management solution that will adapt the use of available flash to workload conditions. Techniques for improving performance (e.g., moving non-sequential content to flash, use of flash as a write buffer) will be investigated. The investigators will explore how flash can facilitate improved service differentiation by reducing the variance of access times inherent in the operation of disks. Finally, the the feasibility of selected replication of popular content on disk and flash and diverting more IO traffic to flash during periods of thermal emergencies will be investigated. Power savings resulting from opportunities to slow down disks, without compromising performance, will also be explored. The investigators will implement a Linux-based prototype Direct-Attached Storage HybridStore system that will manage a hard disk drive and a SATA-enabled flash drive attached to the shared IO bus. To explore other hybrid configurations (such as flash on disk or RAID controller), a comprehensive simulator called HybridSim will be implemented. The PIs will enhance the graduate and undergraduate curricula at Penn State with topics related to this research.
