OW2 Meeting Towards Building a Cloud Platform for Service Oriented Software Development

Towards Building a Cloud Platform for

Service Oriented Software Development

Hailong Sun (Assistant Professor) Beihang University

Agenda

Motivation

Existing work

Traditional software development

Requirement

Specification Design Coding Test Deployment

- the support of various development tools

- the runtime environment for both test and production

running

Service–oriented software design & development

Basic approach

Improve the productivity of software development through reusing of loosely-coupled services over Internet

Applications Component Services Infrastructure

Key issues

Many tools and middleware involved

Tools

Business process modeler

Service composition tool

Verification and test tool

Service deployment tool

Monitoring tool

Middleware

Service container

Composite service execution engine (workflow

engine)

Problems

The extra work for developers

Install and configure various development

tools

Install and configure runtime supporting

middleware

Obtain and maintain underlying resources

Machines

Networks

However they should focus on

“development”.

The crux of cloud computing

Service

IaaS: Infrastructure as a Service

PaaS: Platform as a Service

SaaS: Software as a Service

Can we make use of cloud computing to

save developers from the extra work?

An online development environment

A hosting platform

Service-oriented software architecture

Infrastructural resources

Software Production and Running Applications

R&D goal

Building a cloud platform for service

oriented software development

Instant development without installing tools

Instant deployment without installing

middleware

Moving towards a cloud platform

Online application development

Browser-based IDE

Business process centric development

Online sharing of software services

Transparent deployment and running

Agenda

Motivation

Existing work

Existing R&D work on service oriented

development

Trustie project: SOARProLine (Service

oriented software production line)

SOARWare : SOARProLine+

Service resource management

Design and development

SOARBase Active se_rvice collecting Service mining Service i nfo. coll ecting Service development tool Business process modeling tool Service choreography tool

Service test & verification tool Service deployment tool Other tools Software modeling and development Ser_vice Re g./D isco very deploy

SaaS-oriented Software Service Provisioning and Consuming Environment

Online modeling and

development Service access Service provisioning

Service Info. Mgmt.

Monitoring

Service evolution

mgmt. _{and evolution} Service running

1. SOARBase: service management

Service resources

Atomic services

Composite services

(business process)

Functionality

Collecting

Service mining

Service recommendation

Registration & search

Mgmt of service attributes

(QoS)

Service Oriented SPL

Service Oriented SPL

Developer

Intermediate

assets Business

model Web service

Execution model Composite service Business requirement Software product

Service oriented SPL

SPL config & management tool

Configue software production process Mgmt. of software production process Assign and Mgmt. of production tasks Mgmt of intermediate assets

Various software production tools

BP modeling tool

BPIDE-Business Process IDE

Collaboration between multi-roles

3. Service running and evolution

Service bus

Message exchanging

QoS-aware service scheduling

Monitoring of composite service engines and service

instances

Service container

Axis2 powered

Easy service deployment

Container management

Composite service engine

Service Invocation Proxy Runtime Sys. Mgmt. SOARBus BPMN Parser Evolution Mgmt. Task Pool Execution Processing Communication with SOARBus

SOAREngine Service deployment Processing Result Mgmt. SOARBase Deploy/Unde-ploy

Communication with SOARBus ServiceInfo

Monitoring

Start/Stop Axis2 core

Applications of SOARWare

Funded by 863 program

Digital China

Government tax service

Wonders Information

Agenda

Motivation

Existing work

The state of the art

Service composition Service composition AS API AS API Infrastructure resources (compute, storage, network…) Infrastructure resources (compute, storage, network…) Software Appliance: Service Container、BPMNEngine、App Server Software Appliance: Service Container、BPMNEngine、App Server SA org. and mgmt. SA org. and mgmt. CS API

CS API RM APIRM API

Scheduling, running & monitoring Scheduling, running & monitoring Service deployment Service deployment

AppEngine

(PaaS)

ServiceXchange MyCloudMyCloud

SaaS

IaaS: software appliance provisioning

Software appliance=VM/PM+ middleware

Dynamic provisioning Manageability

Functions

SaaS

Atomic Service API

Atomic Service API Composite Service APIComposite Service API Runtime APIRuntime API

AppEngine

(PaaS)

Online development

ServiceXchange _MyCloudMyCloud _AppStoreAppStore

SaaS

Online App Access

ServiceXchange

A production version of SOARBase

www.servicexchange.cn

Research work behind the scene

Service crawling

Service mining: clustering and conntectability

Service recommendation

Trustworthiness management

Tagging

MyCloud

Personalized development environment

Service subscription

App development tools

Instant development tools: BPIDELite- a

light-weight BPMN editor and service composition

tool suite.

Instant deployment interface

Instant running-Virtual runtime environment

MyServiceContainer

MyBPMNEngine

BPIDELite

Flex-based development toolset

BPMN-based business process modeling

Service choreography

MyServiceContainer

Virtual service container

Service deployment/un-deployment

WSDL file browsing

MyBPMNEngine

Virtual composite service execution engine

BPMN compliant

CS deployment

Online monitoring

Flex-based composite service monitoring

Instance state

App Store

Application management

Agenda

Motivation

Existing work

The state of the art

Conclusion and future work

Conclusion

We aim at building a cloud platform for service oriented software development

Instant service for developers

Instant development Instant deployment Instant running

Future work

Richer online development tools Better multi-tenant support

Load balancing Separation

Dependability

36

iVIC

The Internet-oriented Virtual Computing Infrastructure

Towards resource virtualization

Design Goals (**)

Support resource allocation based on VM instances

To provide another software encapsulate method

Image creation, dissemination and optimization

Container for VM Images: Load VM images dynamically

To provide internet-scale resource allocation and adaption

Wide-Area Network VM Migration

Support the network application virtualization

Linking virtual machine instances via virtual network overlays

vHaaS: virtualized Hardware as a Service

38

Motivation

To support wide-area resource sharing, like department, organization, city, or anything bigger than an office

underlying resource may not be in the same LAN (link, subnet)

Current VM technology lacks in support of WAN Migration

Network Problem: VM’s IP address will be unavailable to other Internet host

Storage Problem: Hosts must share a network storage, which is inefficient for WAN hosts

App App

Provide Service

39

Similar problem with Mobile IP

Mobile IP deals with actual physical device movement VM migration only involve with runtime-state movement

Our solution: Hypervisor Controlled Mobile IP

HyperMIP: Hypervisor Controlled Mobile IP

Combine Proxy-mode mobile-ip stack with hosts VM Migration:

40 Network performance of VM1 (HyperMIP) and VM2 (non‐HyperMIP) Network restoration time of live migration Network traffic during live migration(w. HyperMIP) Topology for Exp1: Network Performance Topology for Exp2: Network restoration time of migration

Service down-time In Migration Average 3.1s across two LANs

~400ms overhead than migration in LAN 2.7s

Service down-time In Migration Average 3.1s across two LANs

~400ms overhead than migration in LAN 2.7s

Network Performance

No significant overhead for Web Server workload and File Server workload (<1%)

Network Performance

41

Problem

To migrate VM, two host must share NAS

Otherwise migration downtime is unacceptable

NAS is infeasible among distributed resources

NAS disk performance is limited by network bandwidth

Solution: MigLoop, two level Vdisk(remote/local)

NAS VM VM VDisk VDisk VM VM Migrate VM VM VDisk VDisk VM VM Migrate HTTP/FTP/NFS VM VM Lvl1 VDiskLvl1 VDisk VM VM Migrate VDisk VDisk Lvl2 VDiskLvl2

VDisk _VDiskVDiskLvl2 Lvl2 Poor disk performance

Poor disk performance _{Long migration}

down-time

Long migration

42

Virtual Disk Performance

Migration Performance

Disk Performance In Migration MigLoop vs Loop overhead < ~6%Disk Performance In Migration MigLoop vs Loop overhead < ~6% Disk Performance

MigLoop vs Loop DBench overhead <2.1%

Disk Performance

43

iVIC Clients

vProcess Servers vSpace Servers

vSpace User Storage

vStore (Software Libs)

Presenta tion Streami ng Application Streaming

44

From you laptop

By using

browser with iVIC Client Plugins

.

By plug a

USB Flash Disk

into any computer.

From you mobile phone

iVIC Platform

Programmable virtual computing environment

Low cost, high flexibility and reliability Virtualized Software as a Service

Key technologies

VM Migration over WAN P2P based iVIC-VPN VM high availability Software streaming

Applications

Software lab

Massive data processing (MapReduce) HPC - vCluster