Database Reference
In-Depth Information
Fast, Simple Database Infrastructure Deployments
The first business benefit offered by ODAs is fast and easy database infrastructure and database instance
deployments. Deploying an ODA does not require spec'ing equipment, provisioning Storage Area Network (SAN)
infrastructure, deploying RAC private interconnect networks, installing an Operating System (OS), or installing RAC,
ASM, the grid infrastructure, Oracle database software, or project managers to coordinate all of this work.
The deployment of an ODA simply requires the steps outlined in the setup poster and deployment documents.
Once the server is on the network, and the deployment configuration file is built, the time to build the ODA is slightly
less than one hour. The result is a two-node RAC cluster, including the configuration of the file systems, ASM storage,
grid and database homes, a local OEM database control, and a starter database. Your company can be running on a
RAC cluster in less than a day.
If the time to deployment matters to your company, then ODAs will meet the need to have fully deployed high-
availability databases and virtualized application solutions within days of completing the server cabling and other
network dependencies. An ODA can be deployed by the average DBA by just reading the documentation. If your
company is deploying a larger portfolio of ODAs, then adding additional skill sets to your team may make sense.
After deploying our first ODA, we counted the number of pre-deployment, install, and post-deployment steps
involved in building databases on our other build-it-yourself platforms. The number of steps to cover all of the bases
for the database server, without creating the databases, was between 103 and 107. The equivalent number of steps on
an ODA to accomplish the same result, and also deploy a RAC cluster, numbered only in the twenties. Needless to say,
we estimated that the reduction in the build and setup time was geometric in nature. Every company should perform
their own count of the number of steps and handoffs between teams required to build a database server and high-
availability databases using their current build processes.
We experienced a similar reduction in the costs for deploying ODA database infrastructure. Every company has
their methods and rules for defining organizational roles and responsibilities. Let's just say that you can build an
ODA without the involvement of a SAN team and sysadmins if that aligns with your company's views on roles and
responsibilities. There is also a reduction in the amount networking work required to deploy and ODA.
A number of companies we have talked to regarding what resources are needed to build an ODA vs. how
things are done today at their company, report that this change in paradigm can result in some interesting internal
discussions as they shift their strategy toward engineered systems. The shift in roles and responsibilities can be
worked out up front, or they can slowly change over time, as companies become more familiar with the platform and
gain more experience.
So that leads to the question: “How long does it take to deploy an ODA?”. To answer this question, you have to
realize that an ODA has dependencies just like any other engineered system. You still have to order the ODA and
have it shipped. An ODA requires fairly simple network cabling, and you still need to complete any DNS and firewall
requests. However, much of this work can be done in parallel, even before the ODA arrives. With careful planning, an
ODA can be deployed very quickly, with the actual build work taking less than a day.
The author's first experience with an ODA was for a project that required high availability and a very fast
deployment. We needed to order equipment and have a set of four RAC clusters running in two weeks. At the time, the
ODA platform was new, but we decided to take a chance because this was our best option. We ordered four ODAs and
asked for immediate shipment. The ODAs arrived on a Monday, and we racked and cabled them on Tuesday. Some
Oracle people flew in on Wednesday morning to see if they could help out, since the platform was very new. We drove
to the data center and built all four ODAs at the same time in parallel Wednesday afternoon—with Oracle looking over
our shoulders.
The process went so well that we decided to take one of the ODAs and bare metal a brand-new running
production ODA RAC cluster in front of the entire DBA department over lunchtime. We finished the deployment
just as everyone finished their lunches. People knew that they were looking at a new paradigm in building database
servers. Building four RAC clusters in parallel in an afternoon with just a couple of DBAs was unthinkable before that
day. Table
9-1
lists many of the tasks that we no longer had to be concerned with as a result of deploying an ODA.
