Current Filter: Storage>>>>>Case study>

   

The 888 Holdings plc is the umbrella organisation for one of the world's most popular online gaming providers, the 888 Group, having been at the forefront of its industry for over a decade. Facilitating online gaming requires a high-performance IT infrastructure that can process multi-player transactions in real-time, discover usage patterns and data relationships that provide business intelligence (BI) and deliver a world-class gaming experience. To support these requirements, they implemented a Microsoft SQL Server RDMS (relational database management system) as part of its initial enterprise infrastructure in support of data warehousing, OnLine Transaction Processing (OLTP) and OnLine Analytical Processing (OLAP) applications.

To keep up with the system's ever-increasing demand for performance the infrastructure was upgraded with a new server, and a high-end SAN comprised of multiple RAID sets, a full SSD tier featuring auto-tiering and internal caching that resides in front of the hard disk drives (HDDs). In order to meet the compute requirements created by the online gaming market, the IT department was required to constantly upgrade its powerful tier-1 SAN in an attempt to keep up with server IOPS and application performance demands. As the tier-1 SAN continued to grow, significantly more power and associated cooling was also required, driving up data centre total cost of ownership (TCO) and total operating cost (TOC). The advantage of flash caching at the server level is in enabling access to hot database data on flash locally for immediate use by SQL Server, not only accelerating application performance but also reducing I/O pressure on the SAN. With this objective, 888 conducted a 'bake-off' between two selected cache solutions to obtain necessary performance statistics, which included OCZ's ZD-XL SQL Accelerator.

The ability for ZD-XL SQL Accelerator to partition flash volumes provided 888 with the best of both worlds; firstly, an optimised solution where tempDB data files benefit from high flash performance while secondly, hot areas of the database are flash cached for immediate use by SQL Server.

Through the vendor's 'Direct Pass Caching Technology' the solution is able to employ advanced SQL Server policy-based algorithms and optimised 'application-specific' caching policies providing 888's IT team with statistically-optimised decisions on what data to store in cache while achieving high hit ratios. A cache director dynamically sends data access metadata to the cache analysis engine, which then performs deep statistical 'out-of-band' analysis without interfering with the data path itself.

TESTING TIMES
In order to compare the 3.2TB single card solution with another existing solution the 888 Group's IT department created a test benchmark system using a number of tests. The tests were conducted over a one-month period to obtain statistics, assess SQL Server performance, and compare cache results.

The 888 Group was provided one ZD-XL SQL Accelerator card with a storage capacity of 3.2TB and 5 acceleration cards from another vendor, each supporting 800GB of storage capacity or 3.5 TB of available cache. During the deployment phase, the implementation wizards built within ZD-XL SQL Accelerator enabled plug-and-play installation through a simple to use GUI, while the competitive approach required each of the five cards to be individually configured.

The first set of tests evaluated the key requirements of lowering pressure on the SAN. For these tests, The 888 Group measured the storage workload placed on the storage processor from a utilisation perspective, the I/O load delivered to the SAN, and the bandwidth required to transfer data from the SAN.

The second set of tests evaluated performance as seen from the application server itself where the objectives are a reduction in wait times, improved latency responses and a reduction in average I/O workloads.

The third and last set of tests evaluated end user impact, which included the time it took to complete an extraction, the time to complete key analysis and the time to complete their top ten reports.

Page   1  2