Economics-inspired Resource and Energy Management for Cloud - PowerPoint PPT Presentation

Keynote on ESaaSA – CLOSER 2015 Economics-inspired Resource and Energy Management for Cloud Environments Luís Veiga INESC-ID Lisboa Instituto Superior Técnico Universidade de Lisboa May 2015

A day in the Clouds Services such as storage, e-mail SaaS (e.g. Gmail), Office (e.g. Office 360), Finance (e.g. FinancialForge) Finance Industry Service models High-level language VMs such as Users the JVM which power platforms PaaS (e.g. Jelastic, Heroku , …) and Software Middleware (e.g. BigMemory, Suppliers Apache Hadoop) System-level VMs able to run complete software stacks IaaS Research (e.g. EC2, LunaCloud , …) Center Deployments Private Community Public 2

Main challenges  In general…  Providers want to maximize clients’ satisfaction while minimizing operational expenditure  But, some defend the infant cloud market is an oligopoly [ 1 ] and not fully passing the benefits to the client  PaaS  Large-scale simulations, e-Science applications, increasingly depend on manage language runtimes (e.g. JVM, CLR)  Resource allocation tailored to the applications, taking into account the effective progress of the workload  IaaS  In public, but mostly in community an private clouds, all-or-nothing resource allocation is not flexible enough  A multi-level SLA agreement could foster competition and enlarge the market  Energy and environmental footprint become prime concerns 3

A glimpse into recent work SaaS Multi-threaded application Datacenter Applications Finance Distributed shared objects space wide Deployment Industry Resize based on effective Checkpoint at Resource Interface progress and resource usage the application level Distributor Software PaaS Suppliers Resize based on SLA negotiated with the client Datacenter-wide VM VM provisioner Deployment Research and Broker Center Interface IaaS 4

Layered view of the researched topics Economics-Inspired Resource Management Models Study about High-level Partial Utility Yield-based (QoE) and Return On Adaptability in Virtual Models and Cost Model Investement (RoI) Machines Classifications CCPE IEEE TCC CSSE Distributed Object Heap and Policies for Distributed ARM 2012 Small, Distributed Datacenters for Workload Distribution based on Resource SAC 2013 Architecture Infrastructure-as-a-Service CloudCom 2013 CloudCom 2013 Utilization and Efficiency CCGrid & DOA 2012 CloudCom 2014 CoopIS 2011 Workload Sys-VM Scheduling Resource Management In the JVM distribution Allocation and mechanisms Heap Progress Scheduling Partial Utility- Checkpoint / JSR-284 grow/shrink monitor Mechanisms driven algorithms restore matrices framework IaaS topic PaaS topic 5

Outline  Introduction  A study about «adaptability in virtual machines»  PaaS  Models, Mechanisms, Evaluation  IaaS  Models, Mechanisms, Evaluation  Energy and Community Clouds  Models, Mechanisms, Evaluation  Publications, Conclusions, Ongoing and Future Work 6

Introduction Wrapping up Survey on Adaptability in VMs PaaS Models Mechanisms Evaluation IaaS Models Mechanisms Evaluation Adaptability in virtual machines  How to analyze?  R esponsiveness Monitoring Decision  how fast can the system adapt?  C omprehensiveness Collect data What needs to Adaptation Loop from sensors  which is the breadth and scope of be changed the adaptation process? Action  I ntricateness Act on available effectors  which is the depth/complexity of the adaption process?  Conjecture: A given adaptation technique aiming at achieving improvements on two of these aspects ( R esponsiveness, C omprehensiveness, I ntricateness) can only do so at the cost of the remaining one.  Distributed system in general: C onsistency, A vailability and tolerance to P artitions [5]  P2P: H igh availability, S calability and support for D ynamic Populations [6] 7

Introduction Wrapping up Survey on Adaptability in VMs PaaS Models Mechanisms Evaluation IaaS Models Mechanisms Evaluation Adaptability techniques IaaS PaaS Higher density 8 JS, LV @ ARM workshop 2012 (IaaS) JS, LV @ IEEE CloudCom 2013 (PaaS)

Introduction Wrapping up Survey on Adaptability in VMs PaaS Models Mechanisms Evaluation IaaS Models Mechanisms Evaluation RCI framework internals R min R R M I max Normalization M onitor D ecide R t j Adaptation D C(M,A) C Loop I A ct R min I System under I I max A classification Step 1 Step 2 Step 3 Decomposition Mapping to Aggregation and of techniques a qualitative value normalization 9

Introduction Wrapping up Survey on Adaptability in VMs PaaS Models Mechanisms Evaluation IaaS Models Mechanisms Evaluation RCI conjecture in practice  Currently, 17 influential systems IaaS 1 were analyzed in depth, assessed 0,8 and classified. 0,6 0,4  New systems and techniques R 0,2 can be added without changing C 0 the classification framework I 1 0,8 0,6 0,4  In both types of VMs R is R 0,2 dominant C 0 I  Overbooking exchanges R by C  In Control and Learning , a higher I lead to a reduced R PaaS 10

Outline  Introduction  Adaptability in virtual machines  PaaS  Models  Mechanisms  Evaluation  IaaS  Models, Mechanisms, Evaluation  Energy and Community Clouds  Models, Mechanisms, Evaluation  Publications, Conclusions, Ongoing and Future Work 11

Introduction Wrapping up Survey on Adaptability in VMs PaaS Models Mechanisms Evaluation IaaS Models Mechanisms Evaluation PaaS-level motivation and goals  How to influence an application behavior, effectively (wide range and impact), efficiently (low overhead) and flexibly (with no or little intrusive coding)?  Line of work : Extend managed runtimes (e.g., Java VMs such as Jikes RVM [3] and OpenJDK [4] ) to operate efficiently in multi-tenancy scenarios such as those of cloud computing infrastructures  Accurately monitor resource usage  Monitor application progress  Resource management  Elasticity and horizontal scaling 12

Introduction Wrapping up Survey on Adaptability in VMs PaaS Models Mechanisms Evaluation IaaS Models Mechanisms Evaluation Economic yield o yield is a return/reward from Savings ( S , S )  r a b applying a given allocation strategy Yield ( S , S ) r a b Degradatio n(S , S ) (S) to some resource (r) a b   U ( S ) U ( S ) P ( S ) P ( S )   r a r b b a Savings r ( S , S ) Degradatio n(S , S ) a b a b P ( S ) U ( S ) a r a o Savings represents how much of a given o Degradation represents the impact of resource ( r) is saved when two the savings, given a specific performance management strategies are compared. or progress metric (e.g. execution time). o It relates the usage ( U ) of a resource o It relates the progress ( P ) made with the with the old and the new configuration old and the new configuration 13

Introduction Wrapping up Survey on Adaptability in VMs PaaS Models Mechanisms Evaluation IaaS Models Mechanisms Evaluation Mechanisms Application  Mechanisms incorporated in QoE-JVM Unified Resource Jikes RVM, « winner of the Management framework ACM SIGPLAN Software Class award , cited for its "high Loader Alternative Heap Resizing quality and modular Policies JIT design" » in Compiler http://en.wikipedia.org/wiki/Jikes_RVM Progress Monitoring Framework GC  Progress monitor supported on Java instrumentation State checkpointing for agent infrastructure Threading Migration and Resilience New mechanisms Existing mechanisms 14

Introduction Wrapping up Survey on Adaptability in VMs PaaS Models Mechanisms Evaluation IaaS Models Mechanisms Evaluation Unified Resource Management Framework Application Resource-aware JVM Reconfigurable components (e.g. Distributed scheduling, Migration, GC plans, JIT optimization level) Resource Sensors Internal & External Resource RA-JVM Consume RA-JVM # Files # Connections aware Adapt JVM Resource attribute Resource Awareness # Threads Data Sent/Rcv and Managment Module CPU Usage Used Memory (RAMM) Environment (OS, Network, CPU, ...)  Extension of Jikes RVM [3] , and the GNU classpath, with JSR 284 – The resource management API  Monitoring and enforcement points include  Memory allocation ( heap growth rate), CPU usage, Thread creation JS, JL, LV @ CoopIS 2011, LNCS 15 JS, LV @ DOA-SVI 2012, LNCS

Introduction Wrapping up Survey on Adaptability in VMs PaaS Models Mechanisms Evaluation IaaS Models Mechanisms Evaluation Heap Policies: Base and alternatives  GC-Economics in Jikes RVM  heap growth rate driven by wasted CPU on GC M0 M1 Shrink Growth M3 M2 JS, LV @ CSSE, CRL Publishing, 2013 16 JS, LV @ DOA-SVI 2012, LNCS