PRIMEQUEST 400 Series & SQL Server 2005 Technical Whitepaper (November, 2005)
Fujitsu Limited PRIMEQUEST 400 Series & SQL Server 2005 Technical White Paper PRIMEQUEST 400 Series Server & SQL Server 2005 Technical Whitepaper This whitepaper provides information on the fundamental benefits of running SQL Server 2005 on PRIMEQUEST 400 Series Servers. Trademarks Fujitsu, the Fujitsu logo, PRIMEQUEST and PRIMECLUSTER are registered trademarks of Fujitsu Limited in the United States and other countries. Microsoft, Windows, Windows Server, and SQL Server, are trademarks or registered trademarks of Microsoft Corporation in the United States and other countries. All other trademarks, registered trademarks or company names mentioned in this document remain the property of their respective owners. November, 2005 i All Rights Reserved, Copyright FUJITSU Limited, 2005
Introduction PRIMEQUEST 400 Series Server & SQL Server 2005 Improved reliability and availability for Windows platforms PRIMEQUEST Open Mission Critical Server In today s fast changing business climate, IT systems are no longer locked away within the enterprise. Instead, use-anytime, from-anywhere access from the public infrastructure is the norm. In such a dynamic environment, it is essential to minimize IT costs with a quick and reactive computing infrastructure that handles access to the IT system, provides stable operation in a 24/7 working environment, restricts the effect of problems on the business, and still provides efficient application management. To provide a strong foundation for these needs, Fujitsu developed the PRIMEQUEST open systems mission-critical server. It combines mainframe-class reliability with the flexibility and economical efficiency of an open server, offering robust performance and scalability in the 64-bit Itanium II Windows Server environment. SQL Server 2005 for assured reliability and availability SQL Server 2005 further enhances the excellent capabilities of SQL Server 2000, offering increased reliability and stability suitable for demanding enterprise systems. In particular, the Database Mirroring and improved fail-over clustering functions make the construction of a highly reliable and available database (DB) system much easier. PRIMEQUEST & SQL Server 2005 Together The combination of PRIMEQUEST servers and SQL Server 2005 provides a robust database host platform with enhanced levels of reliability, availability, and scalability. Reliability The PRIMEQUEST server provides complete duplication of its main hardware environment and stand-by technology to cover hardware breakdown. This mainframe-like reliability combined with the DB mirroring technology of SQL Server 2005 offer superior database reliability and availability. Scalability The PRIMEQUEST server state-of-the-art SMP technology enables it to scale to a 32-way system capable of supporting very large scale databases. Combining this hardware scalability with the enhanced multi-processor capabilities of SQL Server 2005 provide an extremely robust scalable database platform. November, 2005 1 All Rights Reserved, Copyright FUJITSU Limited, 2005
PRIMEQUEST 400 Series Server & SQL Server 2005 example solution Solution Highly reliable and available system using PRIMEQUEST 400 Series Server and SQL Server 2005. Customer Profile Major international retail chain with a sales base in North America, Europe and Asia. Credit sales are based on the issue of a distinctive in-house credit card. Credit card users can access their account history using an internet-based online inquiry service. The customer can also check on their available credit limit at participating stores. The load on the system varies area by area, from high to low, throughout the day, but access is available 24hrs. Previous System Configuration The user inquiry system and participating store systems use a Application Server North America #1 North America #2 Europe Asia common DB server, plus there is one application server for each region. (North America has 2 regions, Europe 1, and Asia 1). In order to support 24/7/365 operation, each of the systems is clustered requiring a total of 10 servers to cover the DB server and application servers. DB Server PRIMEQUEST 400 Series Server & SQL Server 2005 provide the answer Maximized business uptime As all four application servers commonly access the DB server, it is essential that DB server downtime is minimized. However, even if the DB server is constructed as a cluster, once a problem occurs the time to resolution will depend on the change procedures and number of transactions being processed. This may mean the DB server takes tens of minutes to restart. During this time the entire system becomes inoperable. Therefore, database downtime must be reduced to zero. The combination of PRIMEQUEST servers and SQL Server 2005 provide this through two complementary features: System Mirror and Database Mirroring. November, 2005 2 All Rights Reserved, Copyright FUJITSU Limited, 2005
System Mirror Mechanism PRIMEQUEST The PRIMEQUEST Server duplicates principal System hardware such as memory and chip sets and Board Memory Memory synchronizes them with the System Mirror CPU CPU Chip Set mechanism. This feature duplicates read/writes to memory via duplex synchronous operation of Synchronous the cross bar and chip sets. As a result, even if a Dual hardware fault occurs service continues and the system does not stop. Cross Bar Cross Bar Database Mirroring The Database Mirroring features of SQL Server 2005 enable the primary server to almost immediately reconnect the application to the database on the mirror server in the event of a host hardware problem. Completion takes place without waiting for recovery to finish. The mirror server detects the problem within 5 seconds and accepts database connection from the application less than 2 seconds later. Covering hardware failures and problems not detectable by the PRIMEQUEST duplex system, database mirroring only requires a few seconds to switch to the mirror server and continue service. SQL Server 2005 Database Mirroring Primary SQL Server Application Server Witness Mirror SQL Server Efficient use of CPU, Memory, and I/O during load fluctuations In the customer s previous configuration, increases in service load could exceed the available physical capacity of the server, causing sluggish response times and negatively affecting the customer experience. Ideally, the customer IT organization would be able to take two 4CPU servers and use them together as a single 8CPU server. Then as service load reduces, the IT organization could reconfigure them as 2 separate servers again. Until recently there has been no effective technology for using CPU, Memory and I/O resources in this way. However, the PRIMEQUEST servers offer flexible partition management that can accommodate changes in business demands and fluctuations. November, 2005 3 All Rights Reserved, Copyright FUJITSU Limited, 2005
Partitioning is an effective solution PRIMEQUEST servers provide partitioning functions where a maximum of 8 independent servers can be hosted in the same server frame, and up to 8 system boards () can be freely associated for flexible partitioning. It is possible to modify the system so lower loaded services use 1 (4 CPU) while a higher loaded service uses 2 (8 CPU) or more, as required. PRIMEQUEST Possible to add to any partitions AP #1 AP #2 AP #3 Cross Bar Floating System Board Flexible I/O Mechanism PRIMEQUEST Servers provide a floating I/O mechanism, where s and (I/O Units) are physically separate. and connect via the cross bar, enabling the floating I/O mechanism to perform flexible modification to the components that constitute a partition. PRIMEQUEST Partition #1 Cross Bar Partition #2 Partition #3 Optimum Partition Configuration The CPU and Memory or I/O resources required by the partition can be assigned and adjusted to system requirements in as flexible a combination as required. Now it is possible to reduce hardware costs to a minimum by enabling the efficient application of hardware, minimization of wasted resources and the provision of optimum partitions. Consolidating 10 servers of the 5 systems into one. Previously, growth in a service area or additional load required the addition of an extra server. The participating store systems required 10 servers just for the user inquiry system. But that is not the extent of the customer s systems. In addition to these two externally focussed systems, there are dozens of servers for administration and sales management within the company. Unless it is possible to consolidate, there is no floor space or systems administration capacity left to handle further service addition or load increases. November, 2005 4 All Rights Reserved, Copyright FUJITSU Limited, 2005
Consolidation of fail-over clustering into 1 server. SQL Server 2005 s failover cluster fault DB Server handling provides high availability support for the entire server. With failover clustering, redundant hardware is set up. Then SQL Server 2005 and the Operating System work together for fault protection. If a fault North America #1 North America #2 Europe Asia Consolidate 10 dispersed servers to 1 PRIMEQUEST ( x6 systemboards) #1 DB Server #2 North America #1 #3 North America #2 #4 Europe #5 Asia #6 Free occurs in the primary server, an automatic mechanism moves the database server to the secondary hardware. SQL Server 2005 failover clustering was extended to Analysis Services, Notification Services and SQL Server replication. Ability to scale up the DB Server without service interruption The original DB server was operating using a 4 CPU server. Because the daily service load can be high, the organization decided to increase the DB server size to 8CPUs. But it is not possible to stop the DB server to scale up; if the DB server stops, all services stop. To date, when a new DB server is introduced, it is necessary to move the data from the operational server to the new server. During this data transition, the service must stop. Downtime reduced to near zero for scale up operation. By using the partitioning functions and flexible I/O mechanisms of the PRIMEQUEST Server combined with the database mirroring of SQL Server 2005 it is possible to scale up the DB server with near zero downtime. What s the procedure? As shown on the right, partition#1 on #1 contains the operational SQL Server 2005. Partition #2 on #2 contains the SQL Server 2005 mirror. Step1 Database Mirroring Halted The database mirroring operating on #2 is stopped. Ongoing data updates are maintained on the operational system. Once the database mirroring is restarted all updates occurring during the downtime and maintained on the operational system are updated on the mirror. Step2 OS Stopped on Mirror Partition Database Mirroring #1 Primary SQL Server #2 Mirrored SQL Server #3 Free November, 2005 5 All Rights Reserved, Copyright FUJITSU Limited, 2005
OS on #2 is stopped. Step3 Reconstructing Partition #1 First #2 and #3 are combined forming a new partition #2. This modification is efficiently performed using the built-in, web-based GUI interface to the unique server management board (MMB) found on each PRIMEQUEST server. Step4 Restart OS on Partition #2 #1 #2 #3 Primary SQL Server Mirrored SQL Server Restart Database Mirroring Partition #2 which now contains #2 & #3 is restarted. Step5 Restart Database Mirroring Using partition #1, database mirroring is restarted. Now the updates that occurred during Steps 1-4 are automatically reflected on both sides of the mirror. Step5 Database Consistency Verification The updates to the mirror are verified with the operational database. #1 Mirrored SQL Server #2 Primary SQL Server #3 Exchange Primary to Mirrored Step6 Manual Fail-over with Mirror Side Automatically Restarting as the Operational System Manual fail-over is performed. As a result, the operational system automatically starts on the previous mirror side. This switching time only takes a few seconds. November, 2005 6 All Rights Reserved, Copyright FUJITSU Limited, 2005
Summary Mission critical systems can be built using PRIMEQUEST 400 Series Servers & SQL Server 2005 The PRIMEQUEST System Mirror mechanism provides mainframe-class reliability. In addition, PRIMEQUEST partition functions and I/O mechanisms deliver flexibility and economic system configuration. SQL Server 2005 s advanced database mirroring capability minimizes business downtime and improves business continuity. Database Mirroring Fail Over Clustering Flexible I/O Partition System Mirror SQL Server 2005 PRIMEQUEST The combination of PRIMEQUEST 400 Series Servers and SQL Server 2005 enables the construction of highly reliable mainframe-class DB systems using open systems technology with superior economic efficiency. Fujitsu Limited is now using PRIMEQUEST Servers and SQL Server 2005 to tackle the building of highly reliable mainframe-class mission critical systems for its customers. November, 2005 7 All Rights Reserved, Copyright FUJITSU Limited, 2005
About this whitepaper The contents of this document may change without notice. Please inform Fujitsu Limited of any errors found in this document. Fujitsu Limited does not provide any guarantee as to the contents of this document. In addition Fujitsu Limited does not take responsibility for any damage resulting from use of the contents of this document. Fujitsu Limited URL:http://www.fujitsu.com/primequest/