Reference Architecture Overview
Expressed here is a "catalogue" of reference architectures that can be used as the basis for production deployments. These architectures are categorized by use case, where each begins with a short description of the use case focusing on the intended use of the resulting production deployment, plus a short discussion of the balance between virtual machines, storage, and certain features/capabilities of Eucalyptus (i.e. boot from EBS instances), and finally a discussion of all technologies/techniques that are requires to run a production Eucalyptus private cloud.
It is not expected that all deployment projects will be able to *exactly* implement one of the reference architecture deployments, but it is expected that production deployments start with one of the architectures from within this catalogue. As an actual deployment diverges from the reference architecture, knowing from which reference architecture the ultimate deployment came from will help to determine/record and communicate precisely the reason why deviation was needed for specific aspects of the deployment.
Currently, reference architectures exist for the following use cases:
- Dev-test: small
- Dev-test: large
- Scalable Web-services: small
- Scalable Web-services: large
- Virtualization Management
- Common Reference Architecture Content
It should be noted that not all features of Eucalyptus are exercised by these reference architectures. Many features of Eucalyptus (multi-cluster tunnelling, LDAP integration) are present for integration purposes, but are not implemented as 'deployment best practice' if they are not needed.
In addition to the base reference architecture, certain use cases also include variants which are (currently)
- VMWare Hypervisor
- Eucalyptus Service High Availability (HA)
If a use case includes both a VMWare and HA variant, they may be combined to result in a base+vmware+ha reference architecture. If a variant does not exist for a given use case (i.e. dev/test large does not have an HA variant, but does have a VMWare variant), a different use case variant should not be used in conjunction with the chosen use case, and customized attention may be needed to ensure production readiness.
Not included in this reference architecture catalogue are POC/Trial architectures, nor will these architectures address the use case of starting with a POC/Trial and moving it into production. If the intended target is a production deployment, starting with a POC/Trial, then the POC/Trial must start from a production capable reference architecture. In other words, a production reference architecture may always be used for POC/Trial, but a POC/Trial architecture may not always be used as the basis for a production deployment.
These reference architectures each specify 'maximums' that the deployment can support. If the maximums in place for each reference architecture need to be substantially exceeded for a deployment, the deployment should be taken into consideration as one that will require devoted, customized attention to ensure its production readiness.