A quiet revolution is occurring in systems integration. The idea of the IT department being the point of integration is being superseded by vendors who will deliver a pre-integrated appliance or cluster, complete with code, so you can build a private cloud.
Oracle picked up a company, Cobalt, in the late 90's, whose claim to fame was integration of app stacks on a server. While the Cobalt brand and concept got lost in the giant company, the idea may have taken root, since Oracle was the forerunner of a new concept. A year or so ago, they started shipping pre-integrated clusters, complete with database code.
The idea was that most of the work to set up the systems was being done inside Oracle anyway, and like blade servers, there was a strong benefit in handling that prior to shipment. These units leave Oracle fully tested, and ready to be cabled up. Once connected to the switch unit, the servers and the rest are ready to go -- no long software installs or upgrades, and no tuning for best performance, with all the associated decisions. These units are like gassed-up cars at the dealership: ready to go.
The savings in time and stress is substantial, since the units are tested thoroughly in the form they'll be used. The old process required a server test, followed sometimes by staging, followed by teardown and reinstallation, and further software installation and testing. Now, there is a single installation and test.
A further benefit, at least to the vendor, is not having to deal with as many configuration variables. Limiting the customer to well-tested configurations, and providing the units to those configurations, will cut warranty and support costs substantially.
This approach is a necessary way of life in the cloud, where the CSP delivers a set of standardized virtual machines, so in some ways, the faster fulfillment that pre-integrated machines bring is a response to cloud flexibility.
The idea is building. HP and SAP have announced a HANA appliance, and Data Direct Networks has a pre-integrated Hadoop cluster.
The real opportunity, though, may lie in the open-source market. The intricacies of picking up one of the new packages, such as Ceph, Hadoop, or OpenStack, are currently slowing market acceptance and growth, and opening up a business opportunity for experts in those areas to solve configuration, integration tuning, and management issues with a strong value-add proposition.
One example of this is Nebula, founded by OpenStackers from NASA. They are taking on the challenge of shrink-wrapping the management and VM-migration tools required to build an OpenStack Cloud cluster, and are delivering the result as a 2U high expert appliance.
The Nebula CTR-1500 is an x86-64 system married with a 10GE switch and delivered with a complete OpenStack controller software stack. The product provisions server, storage, and networking to deliver OpenStack VMs to users. It is a well-tuned design, with SSD cache and storage, and a HDD for slow data like logs as part of the configuration.
Their value proposition is stark. They are experts at doing this, and their customers are not, so why try to grow to expert level before you start, when you can buy the expertise?
The CTR-1500 has 48 10GE ports on the switch, and a cluster will run two links to each server, to a total of 20 servers. The switch is under Nebula control, and this approach allows path management and optimum data flow to storage. Multiple CTR units are typical, and allow up to 5 racks of gear in the Nebula cloud.
The Nebula product also allows hybrid clouds to be built, where the Nebula Cloud instances can be moved into AWS or a similar service. This is a challenge that most hybrid cloud users are just starting to tackle.
I've discussed the Nebula product at length because it is an excellent example of solving integration problems using an expert appliance. It brings a lot of value to the use case, and will definitely put a big dent in the time to implement.
We can expect many more of these types of expert systems. The "hot" areas of the industry such as big-data, searching, and data conversion for mobile devices will benefit from those who know how to solve the problems for those who don't. We'll see expert appliance solutions in storage too, as the issues of cloud connectivity, SSD, and flash dramatically change integration and tuning modes from what is considered current best-practice.
Further along, the migration of GPU-based performance computing into mainstream clients and servers will create a similar opportunity for integrated expert appliances. The concept will likely take large segments of the simple server market and move them to a new level of value-add.
