The Experts below are selected from a list of 567 Experts worldwide ranked by ideXlab platform
Adnan Yahya - One of the best experts on this subject based on the ideXlab platform.
-
WEBIST (Selected Papers) - Shortest Remaining Response Time Scheduling for Improved Web Server Performance
Lecture Notes in Business Information Processing, 2009Co-Authors: Ahmad Alsadeh, Adnan YahyaAbstract:The Shortest-Remaining-Response-Time (SRRT) policy has been proposed for scheduling static HTTP Requests in web servers to reduce the mean response time. The SRRT prioritizes Requests based on a combination of the current round-trip-time (RTT), TCP congestion window size (cwnd) and the size of what remains of the Requested file. We compare SRRT to Shortest-Remaining-Processing-Time (SRPT) and Processor-Sharing (PS) policies. The SRRT shows the best improvement in the mean response time. SRRT gives an average improvement of about 7.5% over SRPT. This improvement comes at a negligible expense in response time for long Requests. We found that under 100Mbps link, only 1.5% of long Requests have longer response times than under PS. The Longest Request under SRRT has an increase in response time by a factor 1.7 over PS. For 10Mbps link, only 2.4% of Requests are penalized, and SRRT increases the Longest Request time by a factor 2.2 over PS.
Ahmad Alsadeh - One of the best experts on this subject based on the ideXlab platform.
-
WEBIST (Selected Papers) - Shortest Remaining Response Time Scheduling for Improved Web Server Performance
Lecture Notes in Business Information Processing, 2009Co-Authors: Ahmad Alsadeh, Adnan YahyaAbstract:The Shortest-Remaining-Response-Time (SRRT) policy has been proposed for scheduling static HTTP Requests in web servers to reduce the mean response time. The SRRT prioritizes Requests based on a combination of the current round-trip-time (RTT), TCP congestion window size (cwnd) and the size of what remains of the Requested file. We compare SRRT to Shortest-Remaining-Processing-Time (SRPT) and Processor-Sharing (PS) policies. The SRRT shows the best improvement in the mean response time. SRRT gives an average improvement of about 7.5% over SRPT. This improvement comes at a negligible expense in response time for long Requests. We found that under 100Mbps link, only 1.5% of long Requests have longer response times than under PS. The Longest Request under SRRT has an increase in response time by a factor 1.7 over PS. For 10Mbps link, only 2.4% of Requests are penalized, and SRRT increases the Longest Request time by a factor 2.2 over PS.
Kathryn S. Mckinley - One of the best experts on this subject based on the ideXlab platform.
-
MICRO - Exploiting heterogeneity for tail latency and energy efficiency
Proceedings of the 50th Annual IEEE ACM International Symposium on Microarchitecture - MICRO-50 '17, 2017Co-Authors: E. Haque, Sameh Elnikety, Thu D. Nguyen, Ricardo Bianchini, Kathryn S. MckinleyAbstract:Interactive service providers have strict requirements on high-percentile (tail) latency to meet user expectations. If providers meet tail latency targets with less energy, they increase profits, because energy is a significant operating expense. Unfortunately, optimizing tail latency and energy are typically conflicting goals. Our work resolves this conflict by exploiting servers with per-core Dynamic Voltage and Frequency Scaling (DVFS) and Asymmetric Multicore Processors (AMPs). We introduce the Adaptive Slow-to- Fast scheduling framework, which matches the heterogeneity of the workload–a mix of short and long Requests–to the heterogeneity of the hardware– cores running at different speeds. The scheduler prioritizes long Requests to faster cores by exploiting the insight that long Requests reveal themselves. We use control theory to design threshold-based scheduling policies that use individual Request progress, load, competition, and latency targets to optimize performance and energy. We configure our framework to optimize Energy Efficiency for a given Tail Latency (EETL) for both DVFS and AMP. In this framework, each Request self-schedules, starting on a slow core and then migrating itself to faster cores. At high load, when a desired AMP core speed s is not available for a Request but a faster core is, the Longest Request on an s core type migrates early to make room for the other Request. Compared to per-core DVFS systems, EETL for AMPs delivers the same tail latency, reduces energy by 18% to 50%, and improves capacity (throughput) by 32% to 82%. We demonstrate that our framework effectively exploits dynamic DVFS and static AMP heterogeneity to reduce provisioing and operational costs for interactive services. CCS CONCEPTS • Computer systems organization $\rightarrow$ Heterogeneous (hybrid) systems; • Software and its engineering $\rightarrow$ Scheduling;
E. Haque - One of the best experts on this subject based on the ideXlab platform.
-
MICRO - Exploiting heterogeneity for tail latency and energy efficiency
Proceedings of the 50th Annual IEEE ACM International Symposium on Microarchitecture - MICRO-50 '17, 2017Co-Authors: E. Haque, Sameh Elnikety, Thu D. Nguyen, Ricardo Bianchini, Kathryn S. MckinleyAbstract:Interactive service providers have strict requirements on high-percentile (tail) latency to meet user expectations. If providers meet tail latency targets with less energy, they increase profits, because energy is a significant operating expense. Unfortunately, optimizing tail latency and energy are typically conflicting goals. Our work resolves this conflict by exploiting servers with per-core Dynamic Voltage and Frequency Scaling (DVFS) and Asymmetric Multicore Processors (AMPs). We introduce the Adaptive Slow-to- Fast scheduling framework, which matches the heterogeneity of the workload–a mix of short and long Requests–to the heterogeneity of the hardware– cores running at different speeds. The scheduler prioritizes long Requests to faster cores by exploiting the insight that long Requests reveal themselves. We use control theory to design threshold-based scheduling policies that use individual Request progress, load, competition, and latency targets to optimize performance and energy. We configure our framework to optimize Energy Efficiency for a given Tail Latency (EETL) for both DVFS and AMP. In this framework, each Request self-schedules, starting on a slow core and then migrating itself to faster cores. At high load, when a desired AMP core speed s is not available for a Request but a faster core is, the Longest Request on an s core type migrates early to make room for the other Request. Compared to per-core DVFS systems, EETL for AMPs delivers the same tail latency, reduces energy by 18% to 50%, and improves capacity (throughput) by 32% to 82%. We demonstrate that our framework effectively exploits dynamic DVFS and static AMP heterogeneity to reduce provisioing and operational costs for interactive services. CCS CONCEPTS • Computer systems organization $\rightarrow$ Heterogeneous (hybrid) systems; • Software and its engineering $\rightarrow$ Scheduling;
Sameh Elnikety - One of the best experts on this subject based on the ideXlab platform.
-
MICRO - Exploiting heterogeneity for tail latency and energy efficiency
Proceedings of the 50th Annual IEEE ACM International Symposium on Microarchitecture - MICRO-50 '17, 2017Co-Authors: E. Haque, Sameh Elnikety, Thu D. Nguyen, Ricardo Bianchini, Kathryn S. MckinleyAbstract:Interactive service providers have strict requirements on high-percentile (tail) latency to meet user expectations. If providers meet tail latency targets with less energy, they increase profits, because energy is a significant operating expense. Unfortunately, optimizing tail latency and energy are typically conflicting goals. Our work resolves this conflict by exploiting servers with per-core Dynamic Voltage and Frequency Scaling (DVFS) and Asymmetric Multicore Processors (AMPs). We introduce the Adaptive Slow-to- Fast scheduling framework, which matches the heterogeneity of the workload–a mix of short and long Requests–to the heterogeneity of the hardware– cores running at different speeds. The scheduler prioritizes long Requests to faster cores by exploiting the insight that long Requests reveal themselves. We use control theory to design threshold-based scheduling policies that use individual Request progress, load, competition, and latency targets to optimize performance and energy. We configure our framework to optimize Energy Efficiency for a given Tail Latency (EETL) for both DVFS and AMP. In this framework, each Request self-schedules, starting on a slow core and then migrating itself to faster cores. At high load, when a desired AMP core speed s is not available for a Request but a faster core is, the Longest Request on an s core type migrates early to make room for the other Request. Compared to per-core DVFS systems, EETL for AMPs delivers the same tail latency, reduces energy by 18% to 50%, and improves capacity (throughput) by 32% to 82%. We demonstrate that our framework effectively exploits dynamic DVFS and static AMP heterogeneity to reduce provisioing and operational costs for interactive services. CCS CONCEPTS • Computer systems organization $\rightarrow$ Heterogeneous (hybrid) systems; • Software and its engineering $\rightarrow$ Scheduling;
