NDLI: CloudLEGO: scalable cross-VM-type application performance prediction

Please wait, while we are loading the content...

Pico replication: a high availability framework for middleboxes

Hierarchical scheduling for diverse datacenter workloads

Small is better: avoiding latency traps in virtualized data centers

Orbe: scalable causal consistency using dependency matrices and physical clocks

Limplock: understanding the impact of limpware on scale-out cloud systems

Scalable lineage capture for debugging DISC analytics

EventWave: programming model and runtime support for tightly-coupled elastic cloud applications

Leveraging data deduplication to improve the performance of primary storage systems in the cloud

On fault resilience of OpenStack

Apache Hadoop YARN: yet another resource negotiator

Go, server, go!: parallel computing with moving servers

Leveraging sharding in the design of scalable replication protocols

When the network crumbles: an empirical study of cloud network failures and their impact on services

Memory-efficient groupby-aggregate using compressed buffer trees

Tolerating business failures in hosted applications

FAST: near real-time data analytics for the cloud

COLO: COarse-grained LOck-stepping virtual machines for non-stop service

Natjam: design and evaluation of eviction policies for supporting priorities and deadlines in mapreduce clusters

Virtual network diagnosis as a service

Understanding and mitigating the impact of load imbalance in the memory caching tier

vTube: efficient streaming of virtual appliances over last-mile networks

Memory footprint matters: efficient equi-join algorithms for main memory data processing

Consistency without borders

Towards a performance-as-a-service cloud

jVerbs: ultra-low latency for data center applications

Scale-up vs scale-out for Hadoop: time to rethink?

MRM: delivering predictability and service differentiation in shared compute clusters

Client-centric benchmarking of eventual consistency for cloud storage systems

Introducing service-level awareness in the cloud

Mortar: filling the gaps in data center memory

PoWER: prediction of workload for energy efficient relocation of virtual machines

Simple and efficient coupling of Hadoop with a database engine

The wisdom of virtual crowds: mining datacenter telemetry to collaboratively debug performance

Extending modern PaaS clouds with BSP to execute legacy MPI applications

Adaptive partitioning for large-scale dynamic graphs

Towards a general framework for secure MapReduce computation on hybrid clouds

CATS: a linearizable and self-organizing key-value store

FastMR: fast processing for large distributed data streams

XDB: a novel database architecture for data analytics as a service

CloudLEGO: scalable cross-VM-type application performance prediction

Comparing SSD-placement strategies to scale a database-in-the-cloud

VMShadow: optimizing the performance of virtual desktops in distributed clouds

Fault-tolerant industrial automation as a cloud service

High performance clustering of social images in a map-collective programming model

Does RDMA-based enhanced Hadoop MapReduce need a new performance model?

Syndicate: democratizing cloud storage and caching through service composition

Coloring the cloud for predictable performance

Decentralized privacy protection strategies for location-based services

Recommending just enough memory for analytics

Process-oriented recovery for operations on cloud applications

CloudSSI: revisiting SSI in cloud era

Firewall placement in cloud data centers

Harmony: coordinating network, compute, and storage in software-defined clouds

Pregelix: dataflow-based big graph analytics

Wide-area streaming analytics: distributing the data cube

High performance in-memory caching through flexible fine-grained services

Compiling machine learning algorithms with SystemML

SuperCloud: economical cloud service on multiple vendors

Dynamic performance profiling of cloud caches

DEDIS: distributed exact deduplication for primary storage infrastructures

CloudLEGO: scalable cross-VM-type application performance prediction

Content Provider	ACM Digital Library
Author	Tan, Jian Rouvellou, Isabelle M. Silva-Lepe, Ignacio Iyengar, Arun K. Liu, Ling Wang, Ting Meng, Shicong
Abstract	Understanding the performance difference of a multi-tier Cloud application between different provisioning plans and workloads is difficult to achieve. A typical IaaS provider offers a variety of virtual server instances with different performance capacities and rental rates. Such instances are often marked with a high level description of their hardware/software configuration (e.g. 1 or 2 vC-PUs) which provides insufficient information on the performance of the virtual server instances. Furthermore, as each tier of an application can be independently provisioned with different types and numbers of VMs, the number of possible provisioning plans grows exponentially with each additional tier. Previous work [10] proposed to perform automatic experiments to evaluate candidate provisioning plans, which leads to high cost due to the exponential increase of candidate provisioning plans with the number of tiers and available VM types. While several existing works [8, 6, 7] studied a variety of performance models for multi-tier applications, these works assume that an application runs on a fixed deployment (with fixed machine type and number for each tier). We present CloudLEGO, an efficient cross-VM-type performance learning and prediction approach. Since building a model for each possible deployment is clearly not scalable, instead of treating each candidate deployment separately, CloudLEGO views them as derivatives from a single, fixed deployment. Accordingly, the task of learning the performance of a targeted deployment can be decoupled into learning the performance of the original fixed deployment and learning the performance difference between the original deployment and the targeted one. The key to efficiently capture performance difference between deployments is to find multiple independent changes that can be used to derive any deployment from the original deployment. CloudLEGO formulates such "modular" changes as VM type changes at a given tier. To capture changes of performance at a tier caused by VM type changes, CloudLEGO uses relative performance models [5] which predict the performance difference between a pair of VMs (rather than the absolute performance of a VM) for a given workload. Moreover, training relative performance models requires only performance data from Cloud monitoring services [1, 4] rather than fine-grain data such as per-tier response time which requires application instrumentation. Training relative performance models with traditional passive learning techniques would require a large amount of training data as performance data are collected uniformly in a single batch. We find that different types of VMs often share similar performance for many "regions" of workloads. To leverage this characteristic and guide the profiling to regions with high performance differences, CloudLEGO uses active learning techniques [2, 3, 9] that split the profiling process into multiple stages where data collected in one stage are used to identify high-value regions for the next profiling stage. As a result, it significantly speeds up the convergence of models and the profiling process due to substantially reduced measurement. We deploy CloudLEGO in IBM's Research Computing Cloud (RC2), an Infrastructure-as-a-Service Cloud, to evaluate its effectiveness. Our results suggest that CloudLEGO provides accurate predictions for various deployments and workloads with only a fraction of training cost incurred by existing techniques.
Starting Page	1
Ending Page	1
Page Count	1
File Format	PDF
ISBN	9781450324281
DOI	10.1145/2523616.2525948
Language	English
Publisher	Association for Computing Machinery (ACM)
Publisher Date	2013-10-01
Publisher Place	New York
Access Restriction	Subscribed
Content Type	Text
Resource Type	Article

Central Library (ISO-9001:2015 Certified)
Indian Institute of Technology Kharagpur
Kharagpur, West Bengal, India | PIN - 721302

See location in the Map
03222 282435
Mail: support@ndl.gov.in

Sl.	Authority	Responsibilities	Communication Details
1	Ministry of Education (GoI), Department of Higher Education	Sanctioning Authority	https://www.education.gov.in/ict-initiatives
2	Indian Institute of Technology Kharagpur	Host Institute of the Project: The host institute of the project is responsible for providing infrastructure support and hosting the project	https://www.iitkgp.ac.in
3	National Digital Library of India Office, Indian Institute of Technology Kharagpur	The administrative and infrastructural headquarters of the project	Dr. B. Sutradhar bsutra@ndl.gov.in
4	Project PI / Joint PI	Principal Investigator and Joint Principal Investigators of the project	Dr. B. Sutradhar bsutra@ndl.gov.in Prof. Saswat Chakrabarti will be added soon
5	Website/Portal (Helpdesk)	Queries regarding NDLI and its services	support@ndl.gov.in
6	Contents and Copyright Issues	Queries related to content curation and copyright issues	content@ndl.gov.in
7	National Digital Library of India Club (NDLI Club)	Queries related to NDLI Club formation, support, user awareness program, seminar/symposium, collaboration, social media, promotion, and outreach	clubsupport@ndl.gov.in
8	Digital Preservation Centre (DPC)	Assistance with digitizing and archiving copyright-free printed books	dpc@ndl.gov.in
9	IDR Setup or Support	Queries related to establishment and support of Institutional Digital Repository (IDR) and IDR workshops	idr@ndl.gov.in