Every Database Administrator (DBA) faces the challenge of degrading system performance as their SANs struggle to keep up with the demands of their enterprise application Service Level Agreements (SLAs). Such performance degradations might be temporary in nature, but they can also occur multiple times every month, week or even day—especially in mixed workload environments. Unfortunately, given the workloads of today’s database administrators, many DBAs don’t have the time to tune databases successfully to meet maintenance demands. They need a self-managing solution. Now, with the Cirrus Data Solutions Data Caching Server (DCS) utilizing Seagate’s Nytro flash accelerator cards, that solution has arrived.
DBAs address a performance problem first by determining whether the issue is related to storage; then, if so, they identify the LUNs affecting the storage performance. Is it hot data? Frequently accessed data? Hot disk inside the SAN? Only after the problem is identified and diagnosed could the DBAs consider possible avenues to remedy the problem. As database applications demand high transaction rates with minimal response times, the DBAs try to address storage issues using any of the following strategies:
- Tuning the database
- Adding more servers and licenses
- Adding more spinning disks
- Migrating to a faster storage solution: faster disks and/or implementing a flash- based solution
Each of these strategies, however, comes with its own set of potential issues and risks, including system-wide downtime to install, associated costs, space constraints and unavailable DBA resources to manage them.
The DCS appliance addresses these performance issues. DCS is a 2U Linux-based server equipped with multiple pairs of FC ports and with up to three Seagate Nytro flash accelerator cards (and another 10 Nytro cards when used with an optional external PCIe extended enclosure) and 512GB of RAM. It was designed to be dynamically inserted between the host and the SAN with no downtime—no adding more work for the DBA, no database changes, no cost of adding more spinning disks or adding more SAN arrays.
DCS also provides the user with an enormous amount of information and statistical data concerning every component within the SAN. It can also dynamically configure the cache based on discovered hot spots. This cached data will reside on the appliance’s PCIe-based flash cards, such as the Nytro flash accelerator cards.
The evolutionary algorithm utilized by DCS automates the fine-tuning to optimize settings dynamically. This is accomplished using an LRU/LFU (least recently used cache/least frequently used cache) algorithm. Therefore, whether you’re having a boot storm or running a database report that would take multiple hours, the performance increases dramatically without having to implement a complete change to the environment. DCS also utilizes a write-through cache to guarantee write coherency.
DCS then defines the hot data and identifies how much cache is needed to address each hot spot. It then dynamically maps the cache autonomously, or allows the DBA to create their own cache policies if they so choose.
For more information, visit this co-written brief with Cirrus on how to increase database performance.
Also be sure to visit the Cirrus booth at Oracle OpenWorld in booth #724!
—
Seagate creates space for the human experience by innovating how data is stored, shared and used. Learn more at www.seagate.com.
