Automated Application and Resource Management in the Cloud Nikos Parlavantzas HDR defense—15th June 2020
The demand for collaboration and entertainment services is skyrocketing 12 million 2,900% New daily active Growth of daily users in 1 week, up participants between 37.5% December and April 2.5 million 15.8 million New connected users New subscribers in 1 week, up 25% between January and March 2
Made possible by the cloud Estimated Cloud Growth in US Worldwide Enterprise Spending on Cloud and Data Centers https://internetassociation.org/publications/ex amining-economic-contributions-cloud-united- states-economy/ http://ec.europa.eu/newsroom/document.cfm?doc_id=41184 3
Cloud brings benefits to both providers and customers APP PROVIDER CUSTOMERS 4
Cloud brings benefits to both providers and customers APP PROVIDER CUSTOMERS Pools and shares Obtain and release resources among resources on demand customers while paying only for actual use 5
To optimise these benefits, these actors require automated management PROVIDER CUSTOMERS 6
Building automated management systems is difficult PROVIDER CUSTOMERS Satisfying provider objectives (e.g., increasing profit) despite variations in • customer workload • QoS and prices of underlying infrastructure and cloud services 7
Building automated management systems is difficult PROVIDER CUSTOMERS Satisfying customer objectives (e.g., maintaining performance, reducing costs) despite variations in: • application workload and requirements • cloud service capabilities, QoS, and prices 8
Outline 1 2 Application Management for Customers Resource Management for Providers CUSTOMER PROVIDER Application and Resource Management 3 in Private Clouds PROVIDER CUSTOMERS 9
Outline 1 2 Application Management for Customers Resource Management for Providers CUSTOMER PROVIDER Application and Resource Management 3 in Private Clouds PROVIDER CUSTOMERS 10
We discuss two management systems for public cloud providers Both provide SLA support and profit optimisation ► Resource and execution management for SaaS providers ► SLA-based resource management for PaaS providers 11
How to manage service delivery in order to increase the SaaS provider profit? SaaS system that requests ► delivers a (master-worker) application as a requests service SaaS ► supports SLAs that specify response time and reliability QoS ► uses a single IaaS cloud IaaS 12
Our approach provides mechanisms for creating SLAs and managing service delivery SLAs QoS Requests translation Execution Resource management management Under-provisioning Worker Contract cancellation replacement Request cancellation [A. Lage PhD][CPE'17a][CLOUD’12] Decides how to use the mechanisms [ECOWS'11][MONA'10][CIT'10] based on cost-benefit calculations With A. Lage, J.-L. Pazat 13
Applying the mechanisms increases provider profit ... Experiments with an audio encoding Customer 1 Customer n negotiation, provisioning, cancelation SaaS service on Grid’5000 Service Qu4DS § Contract cancellation à 4x profit increase SLA Management Execution Management Negotiation Templates Templates Qos Translation Qos Translation § Under-provisioning à 20-50% profit Resource Management increase PaaS Configurable mechanisms § Worker replacement à 60% profit Contract Rescission Under-provisioning increase Booking Allocation resource booking, request availability (re)allocation management IaaS 14
How to share resources under SLA constraints in order to increase the PaaS provider profit? PaaS system that ► hosts applications of various types Public PaaS cloud ► supports SLAs ► uses private resources and resources rented from IaaS clouds 15
Our approach decomposes resources into application type-specific groups ► Groups decide independently how to allocate their resources to applications, exchanging, if necessary, resources with other groups and renting resources from public clouds Public cloud ► Decisions are based on a profit optimisation policy [D. Dib PhD] [CPE'17b][CCGrid'14] [ORMaCloud'13] With D. Dib, C. Morin 16
Optimisation policy at a glance ► Three options for obtaining missing resources: § waiting for private resources to become available § obtaining them from running applications Public cloud new § renting them from a public cloud app 17
Optimisation policy at a glance ► Three options for obtaining missing resources: § waiting for private resources to become available bid § obtaining them from running bid applications Public cloud new § renting them from a public cloud price app ► Bids correspond to estimated penalties for each option 18
The approach increases provider profit ► Experiments in Grid’5000 (90 nodes) Private VM Public VM Users § MapReduce and batch applications Results Submission SLA negotiation Client Manager ► The optimisation policy generates 9.02% Request transmitting VC. b VC. a App. 2 more provider profit than a baseline policy App. 4 App. 4 App. 2 Controller Controller App. 1 App. 3 and has minimal QoS impact App. 1 App. 3 Controller Controller Cluster Manager Cluster Manager VMs exchange Add/Remove public VMs Resource Manager Add/Remove private VMs VM manager Public resources Private resources 19
To sum up ► A complete, SLA-driven management solution for SaaS providers ► An SLA-based PaaS solution hosting various application types on a hybrid cloud ► No support for dynamically adding resources to running applications ► Only homogeneous, coarse-grained resources 20
Outline 1 2 Application Management for Customers Resource Management for Providers CUSTOMER PROVIDER Application and Resource Management 3 in Private Clouds PROVIDER CUSTOMERS 21
We discuss two application management systems for IaaS customers ► Managing modular multi-cloud applications ► Managing epidemic simulation applications with monolithic structure 22
Deploying and managing applications in multi-cloud environments is challenging ► Producing an initial deployment that satisfies requirements ► Dynamically adapting the deployment to react to environment changes (e.g., changes in workload, resource prices) 23
Platforms that address this challenge adopt a similar architecture Modelling Deciding Executing Application Monitoring 24
Platforms that address this challenge adopt a similar architecture [PDP’18] [ARMS’17] [IOT360'16] Modelling Deciding Executing Application [UCC’16] Monitoring [PaaSage] [C. Ruiz PhD] With L. Pham, A. Sinha, C. Morin, C. Ruiz, H. Duran-Limon 25
Platforms that address this challenge adopt a similar architecture [PDP’18] [ARMS’17] [IOT360'16] Modelling Deciding Execu3ng Application [UCC’16] Monitoring [PaaSage] [C. Ruiz PhD] With L. Pham, A. Sinha, C. Morin, C. Ruiz, H. Duran-Limon 26
How to continuously optimise both the performance and cost of a multi-cloud application? ► Existing solutions fail to consider adaptation costs together with adaptation benefits 27
We assume that the application generates revenue for the customer depending on performance Depends on app performance APP CUSTOMER Objective : Optimise profit (i.e., revenue – cloud charges) 28
We propose a continuous deployment optimisation process Current Deployment Model Proposed Reconfiguration Deployment Plan Model Reasoning Comparison Validation Workload, Reconfiguration Performance, Actions Prices 29
Validation at a glance Profit Current profit now time 30
Validation at a glance Profit a@er Total profit of Profit reconfiguraAon doing reconfiguration Current profit Profit during reconfiguration now time Reconfiguration duration Stability Interval 31
Validation at a glance Total profit of Profit doing reconfiguration Current Adapt profit > Total profit of doing nothing now time Reconfiguration duration Stability Interval 32
The approach is effective in optimizing performance and cost ► Evaluated the approach using experiments with a web application on multiple clouds ► Showed effectiveness of reconfiguration validation ► Demonstrated PaaSage using applications developed by partners 33
How to enable legacy epidemic simulation applications to run in the cloud? Challenges ► decomposing applications into services ► facilitating their deployment and operation on multiple clouds ► supporting elasticity and handling failures 34
DiFFuSE supports structuring simulators as distributed, interacting services ► Provides reusable code for commonly required functionality ► Provides data exchange mechanisms supporting replication and failure handling ► Builds on PaaSage to support multi- cloud deployment and elasticity [CPE’20][CloudCom’17] With L. Pham, C. Morin, S. Arnoux, G. Beaunée, L. Qi, P. Gontier, P. Ezanno 35
DiFFuSE enables simulation applications to fully exploit cloud platforms ► Used to restructure two legacy epidemic simulators developed by INRAE § Spread of bovine viral diarrhea virus (BVDV) § Spread of Mycobacterium avium subspecies paratuberculosis (MAP) 36
Recommend
More recommend
