There has been major news on the Oracle virtualization front with the announcement of
the Oracle Virtual Compute Appliance ("OVCA" for short). This is an Oracle engineered system designed for virtualization, joining Oracle's family
of engineered systems such as Exadata,
Oracle SuperCluster T5-8.
OVCA is is intended for general purpose use for a wide range of applications in virtual machines rather than being optimized for a specific workload.
It is especially designed for quick deployment into production and ease of use.
Customers can start up virtual machines about an hour after OVCA installation.
That provides faster "time to value" than taking general purpose systems and
designing, adding, and configuring the network, storage and VM software needed to be useful.
OVCA is part of the trend toward "converged infratructure", in which compute resources, storage, networking, virtualization, OS and
application components are engineered to work with one another and acquired and managed as a unit.
This provides better results by eliminating two categories of error often seen in the field: misconfiguration, and problems integrating hardware and software from multiple vendors that may never have been tested together.
Many customer problems result from misaligned hardware and software or from non-optimal configuration.
It's not trivial to build a robust infrastructure that provides performance, isolation, and resiliency, and it's easy to make mistakes
- especially when scaling to many servers.
(When I ran systems as a customer, I used to joke that nobody ever came up to you in the hallway to thank you for nothing crashing
- in effect when "nothing happened", but you sure heard from them if there was a problem. You start at zero points, and can only stay at zero
if everything is perfect; you can only get points deducted and go into the red. But I digress...)
Engineering out possibilities for error is better than removing them in the field.
One-stop shopping with solutions that are known to work well together is a big plus.
OVCA is based on robust configurations of components, and the engineering details have been worked out - even to the extent of
pre-cabling racks at the factory so adding a new compute node uses previously connected, color-coded cables, reducing hassle and risk.
When a compute node (a Oracle VM Server host) powers up, it is automatically added to an Oracle VM server pool,
complete with a repository for OS images, virtual disks and metadata.
This removes most of the manual steps needed for setting up a Oracle VM environment.
A natural question is how to differentiate OVCA from existing Oracle engineered systems.
Tbere are both similarities and differences. Like the other engineered systems, OVCA uses the concept of
"hardware and software engineered to work together", which provides benefits in reliability, reduced customer engineering effort,
and balanced performance.
What is different is that products like Exalogic and Exadata are designed for specific workloads, respectively to be premiere
platforms for the middleware application tier and for database.
Oracle SuperCluster T5-8 is a premiere platform for both application and database tier, but it shares with Exalogic and
Exadata the orientation towards maximum performance and scale on a completely Oracle application stack.
These engineered systems use Oracle Linux and Oracle Solaris 11 as the operating system.
OVCA addresses a different market requirement. It provides very good performance for a wide range of workloads running in virtual machines
(and only in a virtual machine)
and supports much more heterogeneous workloads, and exclusively inside virtual machines. In addition to Oracle Linux and Solaris, it supports
virtual machines running Red Hat Linux. Centos, SuSE, and Microsoft Windows.
The OVCA "sweet spot" is multitenancy with a wide range of guest OS and application choices, and with an emphasis on easy deployment.
Oracle provides several unique advantages here.
In my opinion, the most important advantages is that we can provision the entire technology stack from the storage and networks
up to the application software.
A big part of this is the availability of close to 100 pre-created Oracle VM templates, which contain
virtual machines with pre-installed operating system and application. Download the templates you want from Oracle, import them
into the Oracle VM instance within OVCA, and minutes later you can clone virtual machines from them, with pre-installed applcations like
I've had the opportunity to work with OVCA the last several weeks.
A few days ago, I put an OVCA rack into a "still at the factory" mode, and reprovisioned it from "empty" state. That process was fully
automated (well, okay, I told a management server to do a network install using two commands, and used a third commmand to watch it run),
and took about 45 minutes. It was now at the state that a customer would see it at first on-premises power-up.
At that point, I logged into the OVCA "dashboard" and the Oracle VM console, and saw that it discovered the compute nodes (Oracle VM Server),
and prebuilt a clustered server pool, defined a repository on the ZFS storage appliance, and defined network channels on 10GbE and Infiniband for virtual machine traffic, management, heartbeat, and live migration purposes. This was all done without any effort on my part.
At that point, I imported a simple Oracle VM assembly containing an Oracle Linux template from a nearby web server, and then cloned the
template to make a new virtual machine. From the moment I started the upload to the point where I booted up the guest VM and connected to the console: under 5 minutes.. That's very powerful. Later I'll add some screenshots!
Here are some handy links for reference: