Considering Resource Demand Misalignments To Reduce Resource Over-Provisioning in Cloud Datacenters

Transcription

1 Considering Resource Demand Misalignments To Reduce Resource Over-Provisioning in Cloud Datacenters Liuhua Chen Dept. of Electrical and Computer Eng. Clemson University, USA Haiying Shen Dept. of Computer Science University of Virginia, USA

2 Cloud Computing Cloud computing: large groups of remote servers networked to allow centralized data storage and online access to computer services or resources 2

3 Cloud Providers 3

4 Cloud Customers 4

5 Research Problem and Goal Virtual machines (VMs) Physical machines (PMs) 5

6 Motivation VM VM VM VM VM VM VM VM VM Virtual machine (VM) VM PM VM VM PM VM VM PM Physical machine (PM) Over-loaded PMs low QoS SLO violation penalty Under-loaded PMs resource waste high system cost Problem: reduce over-loaded and under-loaded PMs Goal: high QoS, high resource efficiency, high profit 6

7 Initial Complementary VM Consolidation Previous work (CompVM): L. Chen and H. Shen, Consolidating Complementary VMs with Spatial/Temporalawareness in Cloud Datacenters, Proc. of the 33rd Annual IEEE International Conference on Computer Communications (INFOCOM'14), Toronto, Canada, 2014 VM VM VM VM VM VM VM VM PM PM PM PM How to achieve? 7

8 Memory utilization (%) Initial Complementary VM Consolidation Motivation Utilization (%) CompVM: load balance in the long term consolidate complementary VMs Capacity Resource waste Demand Under-loaded PMs VM VM 1 Spatial Low CPU high MEM VM 2 High CPU low MEM 0 CPU utilization (%) Temporal VM 3 VM 1 VM 2 0 Time T Capacity VM 1 Demand Over-loaded PMs Capacity VM 2 VM 1 Demand No under/over-loaded PMs Patterns? 8

9 Initial Complementary VM Consolidation VM Utilization Pattern Measurement: MapReduce jobs: TeraSort, TestDFSIO read/write cluster trace, trace Periodic resource utilization patterns exist in many VMs running the same short-term job a long-term job TestDFSIO read TestDFSIO read Google cluster trace 9

10 Initial Complementary VM Consolidation Utilization Pattern Detection CPU utilization (%) trace 1 trace 2 pattern Time (hr) u f 1 f 2 f 3 10

11 Initial Complementary VM Consolidation VM Allocation Method Memory utilization (%) Utilization (%) Detect resource utilization pattern of the VM pattern VM7 Choose a PM consolidate complementary VMs VM6 100 VM3 Spatial VM5 For each VM1 PM PM: VM VM2 PM 2 VM4 PM check if it has High CPU PM1 PM2 low MEM PM3 sufficient capacity VM 1 for the VM VM7 Low CPU high MEM Choose a PM 0 CPU utilization (%) 100 VM7 Choose a PM Select the PM: Memory with the least resource waste VM6 remaining resource VM3 VM6 VM5 after allocating VM1 PM the VM3 VM2 PM VM5 VM4 PM VM VM1 PM1 VM2 PM2 VM4 PM3 PM1 PM2 PM3 E j : ratio of aggregated resource demand M j : distance between the average resource demand vector and the capacity vector 0 Time T Temporal VM 1 VM 3 VM2

12 Reducing Prediction Error pattern 12

13 VM Consolidation Reduce Provisioned Resource Pulse deviation yields a pattern with a pulse width larger than the actual pulse width Resource over-provisioning Not revealed or studied before Pulse derived pattern deviation derived pattern 13

14 VM Consolidation Trace Study Pulse deviations are common Google Cluster trace 100 jobs, tasks PlanetLab trace 1000 jobs, 4695 tasks MapReduce benchmarks on a HPC cluster (Wordcount, Grep, Terasort, TestDFSIO and PiEstimator) Average task execution time is around 100 minutes/seconds 14

15 VM Consolidation Trace Study Resource efficiency: demand/capacity Even using CompVM, the resource efficiency still needs to improve Google Cluster trace PlanetLab trace 100 jobs, 1550 tasks 1000 jobs, 4695 tasks 15

16 VM Consolidation Pattern Refinement Methodology Pattern refinement methods Lowering cap : lower each value in the pattern by c high -c low Reducing pulse width Optimal base provisioning c high c low unused resource t 0 1 t 2 t 3 Time 16

17 VM Consolidation Pattern Refinement Methodology Pattern refinement methods Lowering cap Reducing pulse width : postpone the pulse from t 1 to t 3 Optimal base provisioning 17

18 VM Consolidation Pattern Refinement Methodology Pattern refinement methods Lowering cap Reducing pulse width Optimal base provisioning refine pattern based on optimal b value that maximizes resource efficiency b 1 b 2 0 Time 18

19 VM Consolidation Pattern Refinement Methodology Pattern refinement methods Lowering cap Reducing pulse width Optimal base provisioning Risk violating SLOs 19

20 VM Consolidation Performance Evaluation Google Cluster trace Improve by 40% Improve by 10%, 70% Cluster experiment 2000 VMs, CloudSim, Palmetto HPC cluster, Traces: NAS Parallel Benchmark, cluster Pattern refinement yields higher resource efficiency without compromising VM performance by handling pulse deviations! 20

21 Conclusion Trace study Pulse deviations are common Even using CompVM, the resource efficiency still needs to improve Pattern refinement methods Lowering cap Reducing pulse width Optimal base provisioning Experiments Higher resource efficiency without compromising VM performance 21

22 Future Work Consider other factors (e.g., SLOs) in VM consolidation Consider VM migration 22

23 Thank you! Questions & Comments? Haiying Shen Associate Professor Department of Computer Science University of Virginia 23