| Back in the 90's a group of German engineers put | | | | couple of low-cost powerful dual or quad core |
| together the world's first grid computing network | | | | servers and spread the workload over multiple |
| with over 100 PCs running on the first version of the | | | | servers. Not only do you get instant high availability, |
| Linux operating system. It was a great success and | | | | but you gain added scalability beyond your MPP |
| everybody called it the dawn of a new technology | | | | platform's physical limitations. There are two major |
| that will change the computing world forever. | | | | methodologies in achieving grid databases; shared |
| What these engineers didn't understand was how | | | | everything and shared nothing. |
| database engines worked at the time and how they | | | | The shared everything category is dominated by |
| actually set the trends for hardware development. | | | | Oracle and Sybase. Both systems are able to |
| The database engines were calling for Massive Parallel | | | | instantly failover database processes should one |
| Processing (MPP) systems that offered dozens of | | | | participating server go down, aka. high availability. And |
| CPUs in one single server platform. Just when we | | | | both can dynamically scale their CPU power by adding |
| thought the mainframe was dead - with grid | | | | more servers to the grid, and both systems can |
| computing, we witnessed the birth of the open | | | | balance the workload among all participating servers. |
| systems mainframe. | | | | Oracle RAC and Sybase ASE-Cluster Edition are the |
| To offset these MPP systems, software architects | | | | most sophisticated systems available today. If you |
| created a middleware layer to get away from these | | | | want to squeeze the maximum out of your existing |
| monsters. Now we have these MPP systems in the | | | | hardware or if you are seeking to replace |
| center of the universe and all these middleware | | | | energy-wasting and maintenance-fee-eating MPP |
| servers dancing around them. | | | | monsters, these two databases are the weapons of |
| For years to come, companies filled their data | | | | choice. |
| centers with hundreds of middleware servers and | | | | The second methodology is shared nothing. Microsoft |
| dozens of MPP systems. Cooling and power supplies | | | | SQL Server 2008 Federation Data Store represents |
| were running at their peak and data center managers | | | | the leader in this category. Unlike the shared |
| didn't know how to support this hardware excess | | | | everything technology, the shared nothing approach |
| into the future. | | | | has a clear distinction between local and global data. |
| This became the birthplace of server virtualization. | | | | The data federation approach allows combining data |
| Within a few years, server virtualization became the | | | | stored locally on multiple individual databases. It acts |
| main focus for every company searching for | | | | as an aggregator of multiple databases. This is not as |
| infrastructure savings. Consolidating middleware | | | | sophisticated as the shared everything approach, but |
| servers was an easy task and a huge success story | | | | it gets the job done as well. |
| shared with pride by the project managers. | | | | There's a third contender, Sybase IQ Multiplex. This |
| But one area was not so successful in consolidating | | | | system uses a hybrid approach, shared data, but no |
| hardware. MPP systems didn't go away quietly. It | | | | shared cache. This is very unique and no other |
| turned out that the database systems were too | | | | database vendor has anything like it. Sybase IQ is |
| much to handle for the server virtualization frenzy. | | | | Sybase's data warehouse engine. A column-vector |
| Yes, countless efforts have been undertaken to | | | | database that set new performance benchmark |
| move the databases off these monsters, but the | | | | records by having just one distinctive writer node in |
| virtual world couldn't provide the performance | | | | the cluster and a nearly unlimited number of reader |
| needed to support the databases. | | | | nodes. But there's a caution: Never try to run an |
| Remember our grid computing story? The vision of | | | | OLTP application on this system. This system is built |
| sharing an army of small computers to produce the | | | | for data warehousing and massive analytical |
| same computing power as the massive MPP systems | | | | reporting, a perfect match for data hungry BI tools. |
| seemed lost forever. Until the ugly truth about MPP | | | | There are a couple of other database vendors that |
| systems became obvious. Running huge MPP systems | | | | are offering grid database technology of some flavor. |
| are not only energy intensive, but the associated | | | | This article is not meant to create a competitive |
| maintenance costs are also burdensome. Buying a | | | | analysis between all of these systems, but a starting |
| replacement MPP system was the only option. | | | | point to get your imagination going. The bottom-line is |
| However, spending money got increasingly tight over | | | | that preserving energy becomes more and more |
| the past several years. This all became a Catch-22. | | | | important and software vendors are providing |
| You needed to spend more to increase your costs! | | | | solutions to maximize low-cost and low-energy |
| Database vendors to the rescue: grid database | | | | consuming hardware. The future belongs to grid |
| technology seems to be the way out of under the | | | | databases. |
| massive weight of these MPP systems. Take a | | | | |