Driving Data Warehousing with iomemory Fusion s iomemory offers dramatic performance improvements and cost savings for data warehousing environments. Introduction The primary challenge data warehousing environments face is balancing the need for large capacities of data with the high I/O necessary to keep business intelligence applications running at top performance. Fusion-io enables architectures that dramatically improve data warehousing performance, while slashing complex storage infrastructure needs. This white paper describes how Fusion-io can be used in both small and large data warehousing environments, and includes examples of how it has delivered results in practice. Small Data Warehouses Small data warehouses (data marts) generally have a single server host the business intelligence or database application. However, in conventional systems, servers can rarely hold all the data a warehouse needs and thus requires backing by wide-striped disk arrays. While racks of disks can provide the required capacity and I/O performance, storage network complexity makes latency and deep I/O queuing the performance bottleneck for business intelligence applications. Add to this, the reality that disk infrastructure (and cost) must continually increase to cope with ever-growing volumes of data and it is easy to see that current technologies simply aren t a capable, long-term solution. iomemory delivers blistering performance for data marts of up to 20TB capacity. It allows organizations to rethink data infrastructure and put active data on an ultra-low latency Flash memory tier on the server, scaling performance up instead of out. Its revolutionary design puts inefficiently used resources like expensive, multicore CPUs to work, eliminating latency issues and deep I/O queues. This dramatically accelerates business intelligence performance, while simplifying systems to significantly shrink infrastructure and overall costs. countries. All other product and company names and marks mentioned in this document are property of their respective owners. 1
An Example One customer s data warehouse consisted of an Oracle database backed by eight 25-disk MSA70 disk arrays (200 disks). This system could only run one report at a time, so it had to serialize them and was rarely able to complete them in an overnight run. If they did not complete overnight, they would have to be stopped and rerun at a lower traffic time so as not to impact other operations. The company completely resolved its performance issues by moving the database to ten 640GB iodrives, allowing it repurpose its disk arrays. The diagrams below illustrate the systems before and after implementing Fusion-io. BEFORE AFTER 200x DISK ARRAY MSA70 MSA70 MSA70 MSA70 iodrive 640 GB iodrive 640 GB MSA70 MSA70 MSA70 MSA70 DATABASE SERVER DATABASE SERVER countries. All other product and company names and marks mentioned in this document are property of their respective owners. 2
This upgrade offered the customer the following benefits: Report job times improved 8-12x Could run multiple reports in parallel rather than serially Performance for additional reports scales linearly Could run backups at the same time as reports Frees DBAs to create better reports instead of performance tuning Slashed future capital spending on disk arrays Eliminated maintenance on the arrays Recovered eight cases of rack space Slashed power and cooling costs A DBA at the company told us, What will we do with all the rack space that currently contains disk cases if the entire database is in the chassis? And without the comfortable warm breeze the disks generate in the freezing server room? Larger Data Warehouses While the scale up solution offers the best performance, many enterprise-level data warehouses are larger than 20TB. These systems often consist of a SAN, NAS, or other complex storage technology accessed by business intelligence applications across high-bandwidth networks. The problem is that these systems are large, complex, and expensive, and scaling performance requires huge outlays of capital (for collocation fees, servers, racks, cables, switches, and so forth) along with permanent increases to the operating budget (for maintenance, power and cooling, and floor space). iomemory delivers blistering performance for large data warehouses by allowing organizations to rethink data infrastructure and put much, if not all, active data on ultra-low latency Flash memory tiers on servers. Its revolutionary design puts inefficiently used resources like expensive, multicore CPUs to work, eliminating latency issues and deep I/O queues while offering over 1GB of bandwidth per iodrive. This dramatically accelerates business intelligence performance, while taking the performance burden off the storage systems, allowing organizations to significantly reduce spending on high-performance storage systems. countries. All other product and company names and marks mentioned in this document are property of their respective owners. 3
An Example Financial organizations are always looking for ways to improve performance and are among the biggest spenders on high-performance systems. One such institution found that by implementing iomemory it could stop forking out money for a SAN. This customer s environment included servers hosting a business intelligence application, Oracle Data Warehouse and Oracle Reporting that were backed by a SAN. The first change it made was to add iodrives to its Oracle application servers. This eliminated the need for applications to cross the network for active tier data. Next, it replaced its SAN with iodrive-powered servers. The iodrives improved replication times so significantly that they were able to maintain high availability with much less hardware. The diagrams below illustrate the systems before and after implementing Fusion-io. BEFORE AFTER DATA WAREHOUSE REPORTING BUSINESS INTELLIGENCE APPLICATION iodrive 320 GB iodrive 640 GB iodrive 640 GB DATA WAREHOUSE REPORTING INFINIBAND BUSINESS INTELLIGENCE APPLICATION STORAGE SERVER STORAGE SERVER SAN iodrive 320 GB iodrive 320 GB countries. All other product and company names and marks mentioned in this document are property of their respective owners. 4
The financial institution gained the following benefits Improve business intelligence capabilities, by processing much more data, much faster Run reports 75% faster, ensuring job completion before business hours Run many more reports simultaneously to run more jobs each night Run backups on massive amounts of data nightly Eliminated a SAN to deliver complete ROI in under six months just from collocation savings Slashed operational costs for maintenance, power, cooling, and rack space Summary iomemory dramatically improves data warehousing performance, while shrinking infrastructure and overall capital and operating costs. It does this first by delivering business applications large capacities of data in an ultra-low latency Flash memory tier on the server. Second, its revolutionary design restores balance and efficiency to host system resources such as multicore CPUs, maximizing application performance. Finally, because storage no longer needs to bear the primary performance burden, organizations can cut costs dramatically by reducing outlays on SAN, NAS, and other high-performance storage systems. countries. All other product and company names and marks mentioned in this document are property of their respective owners. 5