The Eucalyptus Open Source Cloud Computing System Daniel

The Eucalyptus Open Source Cloud Computing System Daniel Nurmi et al, UCSB Presented by: Sanketh Beerabbi University of Central Florida COP 6087 - Cloud Computing

Introduction �A cloud is a collection of networked resources configured such that users can request scalable resources (VMs, platforms, software services) deployed across a variety of physical resources. �Most cloud computing systems are proprietary and rely upon infrastructure that is invisible to the end users.

Cloud Vs Grid � Cloud § § § Users get small fraction of resources. No support for federation. Resources are hidden. � Grid § § § Users get higher resources. Federation is supported. Resources are visible.

Open Source Clouds � Commercial cloud service providers charge CPU time and bandwidth. � For large organizations, it’s more cost effective to purchase the hardware and create own cloud. � Researchers � More � No and academia need open standards. secure and flexible. vendor lock in.

Iaa. S Platforms � Commercial: Amazon AWS IBM Smartcloud Microsoft Azure HP Cloud Rackspace cloud Go. Grid � Open Source: Eucalyptus Nimbus Open Nebula Enomaly Open Stack Cloud Stack

Eucalyptus � Open source architecture for implementing cloud functionality at the Iaa. S level. � Can be installed and deployed without modification to the existing infrastructure (From laptops to data centers) � Linux based, highly modular and compatible with Amazon AWS (uses AWS APIs).

Architectural Overview 1 - Instances 2 - Networking 3 - Compute nodes 4 - Storage 5 - User interface 6 - Overall cloud platform

Hierarchical Design

Cloud Controller (CLC) � External interface to the cloud for users and administrators. � Performs high level resource scheduling and arbitration. � Lets users manipulate properties of instances and virtual networks. � Monitors � Provides system resources and components. user interface, handles authentication and protocol translation.

CLC Services

Cluster Controller (CC) � Acts as the front end for a cluster of machines within the cloud. � Receives requests from CLC to deploy instances. � Manages instance execution on specific NCs. � Controls virtual network available to instances. � Collects data about resources from a set of NCs.

Node Controller (NC) � Present on every node that is designated for hosting VM instances. � Controls execution, inspection, and termination of VM instances. � Fetches, installs and cleans up local copies of instance images (kernel, the root file system, and the ramdisk image) � Reports physical resource availability and utilization to CCs.

Storage Controller(Walrus) �A put/get storage service (similar to AWS S 3) � Stores user data of any type. � Stores machine images and volume snapshots � Is available to any instance in any cluster. � Implements the REST and SOAP interfaces, which are compatible with S 3

Walrus Storage

Virtual Networking � Every VM must have network connectivity with every other and partially to public Internet. � Public interface handles communication outside a given set of instances. interface used only for inter-instance communication across availability zones. � Private � Provides Isolation - Users have super user access to VM’s network interface and may cause interference with other VMs if unchecked.

Virtual Networking 3 different modes: � Manually define static MAC and IP for each VM. � Directly attach VM interface to software Ethernet bridge and assign IPs via DHCP. � Fully § § § managed by Eucalyptus: VMs attach to user defined networks using VLAN tags and IP subnets. CC acts as router between VM subnets. Cross cluster VM communication via tunneling.

Virtual Networking

Instantiating process 1 -Request a VM. 2 -VM image pushed to CN 3 - Disk image packaged for Hypervisor 4 - The CN provides a virtual NIC with virtual MAC 5 -DHCP on head node provides IP 6/7 - VM starts/User can login via SSH

Overall Architecture

Perormance 1000 TCP throughput of EC 2 and EPC TCP Throughput mb/s 900 800 700 600 500 400 300 200 100 0 EC 2 1 Zone EC 2 2 Zones EPC 1 Zone EPC 2 Zones

Advantages � Open source and modular- allows components to be modified or replaced. � Compatibility with AWS API allows Amazon EC 2 users to transition easily. Existing AWS tools, images and scripts can be used with Eucalyptus. � Hybrid cloud capability: Users can move instances between Eucalyptus private cloud and Amazon public cloud to create a hybrid cloud.

Comparision

Comparison

Comparision

References � The Eucalyptus Open-source cloud computing system, Daniel Nurmi et al. �A Survey on Open-source Cloud Computing Solutions, P. T. Endo et al. �A Comparison and Critique of Eucalyptus, Open. Nebula and Nimbus, P Sempolinski, 2010 � Source code available at: www. eucalyptus. com/download/eucalyptus/source

Questions?