You are on page 1of 25

Red Hat Cloud Architecture

Carl Trieloff Technical Director Cloud, Red Hat cctrieloff@redhat.com May 6, 2011

Red Hat CloudForms Everything required to build a private cloud, a hybrid cloud, or be a cloud provider

Private/Public/Hybrid Cloud
Resource Providers

Resource: An architecture paper with full deployment example worked is available


4

Build a Cloud, Pick a Cloud. Any Cloud.

~]$ yum install cloudforms

The large functional pieces

Layering in the key components

Setting up a basic hybrid deployment

Setup: - Providers - Quota - Services - User creation policy - User Quota (defaulted) - Pools (optional)

Creating a Deployment Definition

Templates The recipes to generate disk images from content repositories (e.g. list of packages, services offered, dependencies, etc.) Assemblies An aggregation (e.g. stack) or association of one or more templates & service configuration. Can be thought of as a system/'machine' definition Deployables A set of one or more assemblies and the information (operational parameters and configuration) in order to target, instantiate, and initialize the deployable Definitions are independent of target resource provider (Private, Public)

Example: Templates, Assemblies & Deployables

Under the covers, creating the Deployable

1->2. User defines the Deployment 2->4. Image Factory assembles the artifacts required 5->6. Meta data + disk images + configuration stored in iwhd 7. Deployable ready to launch

Deploying my application (Deployable)

1) Evaluation to determine placement of resources based on

Dependencies (Services, Infrastructure) Policies (Security, QoS) Artifacts (what artifacts are build / can be built)

2) Instantiate the resources 3) Configure the resource via post boot processing 4) Monitor 5) We're up & running ...

Under the covers, Instantiation & Configuration

The basic resource match work flow

Notes: Resources can be matched based on just about any criteria that can be measured. Cost (Data / Compute / Time /bandwidth / or associated cost basis) QoS (time to launch, Security Zone, etc) Other....

What is this 'post boot' processing


Mechanics to be able to 'incarnate' a machine from a 'Definition' and the implicated artifacts

Disk images Service configurations (network, disk, software/ hardware services) Optional packages Configuration

Allowing for cloud, virtualization and metal instantiations from a single definition.

CDL Post Boot execution


The post boot daemon 'instantiates' the assembly

Each service & node has a schema type,

This type is used to identify which 'script' to execute the metadata with If the 'script' to execute the type is not in the base image, it is downloaded dynamically Mount and link a disk image from NFS

Example:

Disk Image script in base image, so call with metadata My custom script type which not local gets downloaded, then executed

Mount a custom Service

What about importing OVF images?


What if I have an OVF image created elsewhere?

can run image on backend's that support OVF

If we inject the post boot daemon into the OVF image,

can run on any resource provider / public & private cloud that Red Hat supports / certifies can combine/ use in a Deployment / Assembly description. can still use Image in a OVF environment

Adding QoS::Availability to my Deployable


Deliver maximum application service availability for a collection of deployables with CPE
Monitoring Recovery Failure Escalation

Control Start/Stop of Applications Monitor for application instance failures Monitor for assembly instance failures Monitor for deployable instance failures Recover from Application Instance failure Recover from Assembly Instance failure Recover from Deployable Failure Relocate deployable Escalate Application Failures

Escalate Assembly Failures

Escalate Deployable Failures

Managing my content sources

Content includes my application

This content is archived via Catello services, via Pulp for inclusion into my Deployments For JBoss content, the Application Bundle is uploaded as a service from JON. This JON bundle publishes that it requires the JON Managed Service

Dealing with storage,

Archival Store (iwhd)

Operational Store (CloudFS)


large objects explicit whole-file get/put operations (via HTTP) explicit data placement virtual-machine images, media assets, etc S3, Google, CloudFiles, Azure replication comparable

Portable Cloud Shared Storage filesystem-like directories, small files, singlebyte read/write, etc. higher performance caching and/or replication Participates in iwhd meta data naming structure (go see Jeff's talk right after this)

Meta data daemons and repository manage storage domains

OpenShift Power, play with a hosted version

OpenShift Power, Coming Soon

Messaging Services

Today MRG / AMQP is used to run OpenShift. MRG / AMQP as a cartridge for flex

Allows any application to use Portable Messaging in the cloud Allows anyone to consume and use messaging, securely and at scale in the cloud to any application without worrying about managing the infrastructure.

AMQP as a Service

Come join us...


As you have heard, CloudForms will be released as a subscription offer from Red Hat, also, ITS ALL OPEN SOURCE So, feel free to come innovate, develop, run, deploy with us at http://www.aeolusproject.org http://incubator.apache.org/deltacloud/ http://openshift.redhat.com/

Abstract

In this session, Carl Trieloff will provide insight into the Red Hat cloud architecture and will discuss how it provides the next generation of private and hybrid cloud deployments. He will cover a range of topics ranging from CDL (Content Definition Language), to running and mapping services, to cloud, managing quota, federating storage, dealing with replication, and monitoring. Attendees should leave this session with an understanding of how Red Hat is enabling cloud for enterprise use cases.

You might also like