IDC IT Agenda Community –
The recent Microsoft-HP announcement of a range of database appliances based on Microsoft SQL Server and delivered preconfigured on HP hardware is part of a trend in delivering database technology. Microsoft's announcement means that they have joined the other major DBMS software vendors in offering especially customized hardware-software packages (Oracle offers Exadata, and IBM offers both Workload Optimized Systems (including the pureScale Application System and Smart Analytic System) and Netezza Data Warehouse Appliances). At the same time, Microsoft is putting increasing emphasis on the ability of users to deploy their databases in their Cloud service, Azure, using SQL Azure. This puts Microsoft out in front of the major players, but probably not for long. Going forward, it seems certain that enterprises seeking to deploy databases of any substance will have a choice to make: box or Cloud?
High availability and disaster recovery, capturing and processing streaming data, dealing with blends of predefined structured data and ad hoc data, mining social media data, addressing extreme transaction processing demands, rapdily growing databases, and widely fluctuating concurrent session counts are making the setup and management of many large enterprise databases next to impossible. As requirements for mission-critical databases, or even business-critical databases, that involve anything beyond fairly prosaic workloads, become ever more difficult to meet, two very different deployment options are becoming ever more attractive: one involves elastic cloud computing (EC2), and the other involves using a factory preconfigured and optimized hardware and software package. Put more simply, users will, in the near future, tend to turn to clouds and boxes to deal with these headaches.
The "box" concept is neat and simple. Basically, many enterprise database workloads, especially in the area of data warehousing, are so well defined that it is possible for engineers to design and deploy combinations of hardware and software that deliver much better performance, scalability, and robustness than any end-user organization is likely to achieve by buying all the hardware and software parts and piecing them together. These are sometimes called "appliances", though to be precise, that term should really be applied to a hardware-software offering that is closed and offered for a single price. Most of these are more open, allowing for slight modifications and extensions, and they separate the hardware purchase from the software purchase. Sometimes, as is the case with the Microsoft-HP offering, the hardware and the software are provided by different vendors. In all cases, however, they represent an "out of the box" experience that virtually eliminates the costs and risks involved in end-user system deployment. That alone represents significant value, but the fact that the software is optimized for the specific hardware on which it runs, and vice-versa, generally also delivers an ongoing user experience that is far superior to the roll-your-own alternative.
The "cloud" concept is more nebulous (excuse the pun). The promise of EC2 is virtually limitless scalability, even performance despite fluctuating levels of demand, and extremely flexible resource management. An EC2 environment is built using compute and storage grids with load balancing and resource virtualization, and governed by a utility computing automated provisioning system that can deliver additional processor or storage resources on demand. Sounds complicated to set up? You bet! That's why there are only a few enterprises currently building in-house EC2 environments (so-called "private clouds") and a limited number of publicly hosted "public clouds" operating as service platforms on the Internet. That's likely to change, however, as vendors find ways to package operational elements of an EC2 environment and more standards emerge allowing for a non-proprietary way to link those elements together. Right now, concerns about security and online performance are inhibiting user interest in public cloud-based databases, but that should change as the technology matures.
Who will choose which options?
Smaller enterprises may tend to use the public cloud for databases with demanding growth characteristics, and hardware-software packages ("boxes") for well-defined and fairly contained database workloads, such as data warehouses. In many cases, smaller enterprises will actually look to transactional and analytic applications in the Cloud and in boxes that embed databases within them, including application appliances ("boxes") and application software-as-a-service (SaaS) in the cloud, delivering a business rather than technology service.
Larger enterprises may look to Cloud-based SaaS offerings for more prosaic functionality, but for the mission-critical workloads, they will tend to favor boxes, again, for contained problem spaces, and private cloud deployment for the more demanding, scalable workloads. It should also be noted that for larger users there will be boxes available that can be linked together to create a private cloud; in such a case, boxes and clouds are not necessarily mutually exclusive.
One area of data management where users can enjoy the scalability benefits of the Cloud is that of data integration. A number of leading data integration vendors now offer public cloud-based services, including Informatica, Pervasive, and IBM. Over time, the list of vendors offering data management services in the Cloud, including both DBMS and data integration should grow substantially.
The time is coming when most data mangement solutions will no longer be delivered by engineers connecting machines and splicing wires as they do today. Instead, they will get configured solutions as then require, based on their needs, in either boxes or clouds.