Introduction <ul><li>IT managers are challenged objectives: </li></ul><ul><ul><li>Provide continuous application services and data access, even if applications or systems fail or data centers become incapacitated. </li></ul></ul><ul><ul><li>Provide secure universal access to enterprise data. </li></ul></ul><ul><ul><li>Provide application performance that grows as the organization grows </li></ul></ul><ul><ul><li>Contain management costs as the number and complexity of systems grow </li></ul></ul><ul><li>Clustering technologies helps meet these challenges by coordinating interconnected servers to enhance application and data availability, scalability, and manageability. Clusters are conceptually attractive to users because of their potential to solve some of the most significant problems in computing: </li></ul><ul><ul><li>Server Failure: If a server or application crashes, another server in the cluster can take over its application, providing nearly continuous service to clients. </li></ul></ul><ul><ul><li>Network or I/O path failures: If network or I/O links fail, access to applications and data can continue on alternate paths. </li></ul></ul><ul><ul><li>Application growth: If application demands go beyond the ability of a single server, additional servers can be added, and the workload can be shared. </li></ul></ul><ul><ul><li>Management Cost: As the number of enterprise applications grows in, clusters can provide a flexible platform for running multiple applications while allowing the entire cluster to be managed as a single system. </li></ul></ul><ul><li>This need for a high level of availability as well as scalability has led companies to consider technologies such as Oracle 9i Real Application Clusters (RAC). </li></ul>
Introduction (continued) <ul><li>Challenges in a RAC Environment </li></ul><ul><li>This presentation discusses the benefits of using a cluster file system. Cluster volume management and cluster membership services to increase manageability and scalability will also be discussed. </li></ul>
Veritas Failover Overview <ul><li>Seamlessly failover to a Oracle9i RAC </li></ul><ul><li>VERITAS Database Edition /Advanced Cluster uses the VERITAS Cluster Server and the Oracle9i RAC agent to monitor and allow RAC to failover your database clients in the event of a system or application failure. </li></ul><ul><li>Remove all single points of failure </li></ul><ul><li>Database Edition /Advanced Cluster supports Dynamic Multipathing (DMP) to balance I/O loads across multiple storage devices and subsystems for maximum throughput while increasing availability if one path becomes unavailable. Database Edition /Advanced Cluster also allows you to multiple inter-node connection, all of which are actively used, to provide redundant heartbeat connections. However, this is not the case with logical data corruption or human error. </li></ul><ul><li>Replace failed disks while Oracle9i RAC stays online </li></ul><ul><li>Database Edition /Advanced Cluster enables proactive storage management when disk failures occur . When a disk fails, system administrators can replace the failed disk without taking Oracle9i RAC offline. The data is automatically moved to a hot spare disk. The data is then restored to its original configuration after the failed disk has been repaired or replaced. Administrators can set this up to happen automatically. This can also be accomplished through other methods with Hitachi or EMC. </li></ul>
Veritas Proactive Storage Management Note: This is also available on HP-XP and EMC.
RAC Overview <ul><li>RAC is a parallel database environment that takes advantage of the processing power of multiple, interconnected computers. In RAC environments, all nodes concurrently execute transactions against the same database. Each node adds its processing power to the cluster as a whole and can provide scalability in overall throughput and performance. RAC coordinates each node’s access to the shared data to provide consistency and integrity. </li></ul><ul><li>RAC serves as an important component of a robust high availability solution. A properly configured RAC environment can tolerate failures with minimal downtime or interruption to users. With clients accessing the same database from multiple nodes, failure of a node does not interrupt access. Clients accessing the surviving nodes continue to operate, and clients attached to the failed node can reconnect to a surviving node and resume access. </li></ul>Note: This does not help in case of disk failure, data corruption or human error.
Challenges in RAC <ul><li>Most configurations of RAC imply managing the database on raw partitions, which increases administrative burden and costs. Through cluster file system technology, management of data stored in Real Application Clusters can be dramatically simplified without reducing performance. Use of a cluster file systems can also reduce mean time to recover (MTTR) the database if a failure was to occur. </li></ul><ul><li>Imagine a scenario where a RAC database supports e-commerce transactions for a web site. Now, let’s say that this web site suddenly starts receiving a much greater number of transactions than was expected. While this is a good problem to have from a business perspective, the Oracle DBA needs to ensure that the database will be able to stay up while the tablespaces, and redo and archive log files are filling up rapidly. A raw device based solution can create a situation in which, at this very critical time, the DBA would need to intervene, and ask the system administrator to create a new disk partition and then add the partitions to the tablespace in order to expand the storage. Use of a clustered file system alleviates intervention of DBAs and system administrators. </li></ul>
Managing Raw Partitions <ul><li>Managing Raw Partitions – For most Oracle Real Application Clusters implementations today, there is no choice but to store databases in raw partitions. Raw devices reduce the overall manageability of the system by making growth more difficult as well as being more prone to user error, resulting in unexpected downtime. Cluster File systems significantly ease this management burden while ensuring no performance or data integrity loss. While raw devices allow sharing of data across several systems, which is not possible with conventional file systems such as UFS (Unix File System), they are more costly to administer. Raw partitions are not “visible” to typical users and applications. Backups of raw partitions are limited to using the dd command or specialized backup utilities, whereas file systems can be backed up using many common commands like tar and cpio, as well as third party backup software such as VERITAS Netbackup. </li></ul>
Problems with General Purposes File Systems <ul><li>Often, the use of the file systems means giving up performance. Using general-purpose file systems like the Unix File System (UFS) for database files slows down database performance for several reasons. Database servers like Oracle use their own buffer pools in order to manage page allocation more efficiently than the standard file system cache manager. UFS is not smart enough to recognize Oracle is buffering data and continues to do its system cache manager. UFS is not smart enough to recognize Oracle is buffering data and continues to do its own buffering as well. Due to this double buffering of data, writes to a file system from Oracle results in data being copied twice – once between the database buffer and the system buffer cache and later between the system buffer cache and disk. This double buffering negatively impacts database performance. This additional buffering imposed by the system is also an inefficient use of memory and OS resources, effectively cutting usable memory in half and using extra CPU processing to manage the buffer. </li></ul><ul><li>Asynchronous I/O is a form of I/O that performs non-blocking system level reads and writes, allowing the system to handle multiple I/O requests simultaneously. Operating systems such as Solaris provide kernel support for asynchronous I/O on raw devices, but not on regular files. Even if the database server is capable of using asynchronous I/O, it cannot issue asynchronous I/O requests when the database runs on general purpose file systems like UFS. Lack of asynchronous I/O significantly degrades performance. </li></ul>
Problems with General Purpose File Systems <ul><li>Another performance bottleneck is the exclusive write lock at the file system level. This locking mechanism causes write operations to a file to be serialized and prevents Oracle from processing parallel read and write requests. This problem is heightened when multiple users access the same databases simultaneously. </li></ul><ul><li>VERITAS has addressed these problems on single instance systems (non-RAC) with a feature in the VERITAS Database Edition called Quick I/O (QIO). QIO allows the database to access the storage as if it were a raw device, deriving the performance benefits of raw devices, and the system administrator/DBA has the simplified management of a file system. </li></ul><ul><li>In Oracle 9i, Oracle has created the Oracle Disk Manager (ODM) standard, ODM is an API developed by Oracle for the Oracle wishes to do database I/O. VERITAS has implemented an “ODM Compliant” file system by providing a library for Oracle to use to perform I/O. For example, when Oracle wishes to write, it will call the odm_io() function. Through a single function call, multiple IO requests and status checks can be done, reducing overhead. </li></ul><ul><li>Creating an ODM compliant version of VxFS means VERITAS was able to leverage the exceptional performance of Quick I/O and also add additional management features to better support the database administrator. VxFS with ODM provides all the performance benefits of QIO. The VERITAS CFS and ODM interface supports resizing of data files while the database is in use, as well as supporting all Oracle Managed Files (OMF) features. </li></ul>
Managing Physical Devices <ul><li>Increased Manageability with DBE/AC – Configuring RAC in a DBE/AC 3.5 includes the cluster versions of the VERITAS Volume Manager (VxVM) and high performance ODM compliant file system, the VERITAS File System (VxFS). In the DBE/AC environment, these are referred to as the cluster volume manager (CVM) and cluster file system (CFS). </li></ul><ul><li>Cluster Volume Manager – Cluster Volume Manager is based on VERITAS Volume Manager. CVM extends the concepts of VxVM across multiple nodes. Each node sees the same logical volume layout, and more importantly, the same state of all volume resources. With CVM in a DBE/AC cluster, all storage is managed with standard VxVM commands from a single point, simplifying administration in a multi-node cluster. All other nodes immediately recognize any changes in disk group and volume configuration with no interaction. CVM supports multiple RAID levels including RAID 0 (Striping), RAID 1 (Mirroring), as well as RAID 0 +1 capabilities. Mirror break-off and fast re-synchronization with VERITAS FlashSnap is also available for off host backup or offline reporting and decision support systems. </li></ul><ul><li>Cluster File System – The VERITAS Cluster File System is an extension of the VERITAS File System. CFS is a true SAN (Storage Area Network) Filesystem. All data traffic takes place over the Storage Area Network. Only metadata information traverses the cluster interconnects. (This is different than many implementations that are essentially a Proxy Filesystem, where data is sent through another node to the storage). </li></ul>
Managing Physical Devices (continued) CFS is used in DBE/AC to provide the manageability of a file system in a large database environment. When used in DBE/AC with Oracle 9i RAC, Oracle accesses data files stored on CFS file systems via the ODM interface. This essentially bypasses the file system buffer cache and internal file system locking and allows asynchronous I/O. This means Oracle coordinates writing and locking of data. A single point of locking and buffering ensures maximum performance . Cluster Membership Services – VERITAS Cluster Server (VCS) is an open system clustering solution. It provides cluster membership services and high performance cluster interconnect software for RAC. The cluster membership services performs cluster integrity functions such as node monitoring in the RAC system and reconfigure the system automatically if a node failure is detected. The cluster interconnect provides the communication path for the cluster membership services and provides interprocess communication (IPC) for high-speed reliable communication between RAC processes. Oracle’s CacheFusion technology is an example of a RAC processes that utilize the high performance interconnect .. Oracle 9i RAC VCS RAC Extensions Cluster Server (VCS) Database Accelerator ODM Cluster File System Hardware Cluster Volume Manager
Veritas Advantages <ul><li>Benefits of Database Edition – Advanced Cluster – VERITAS Database Edition – Advanced Cluster provides many tangible benefits to IT organizations, including the following: </li></ul><ul><ul><li>Simplified Storage Layout – With the benefits of cluster volume management, the user is free to create standard size, standard performance RAID LUN’s in the enterprise array. The user then creates logical volumes comprised of standard LUNs of storage. This is keeping with the Oracle Stripe and Mirror Everything (SAME) principle. This completely removes the need for the array to be reconfigured for any changes in storage needs. The array can be configured once, and all future storage needs met online, with no interruption to users. </li></ul></ul><ul><ul><li>Simplified Management of Tablespaces with File Systems – Management of tablespaces in DBE/AC is as simple as managing regular file systems in a stand-alone system. With modern day enterprise RAID systems providing outstanding I/O performance, it becomes a matter of creating a volume and a file system of necessary size and putting it into use in the RAC environment. As storage needs grow, existing volumes and file systems can also be grown, without any application downtime. </li></ul></ul>
Veritas Advantages (continued) <ul><li>Shared Oracle Home and Application Binary Directories – The use of a CFS allows the ability to install Oracle on a single shared directory rather than one the individual systems in the systems in the application cluster. A single shared ORACLE_HOME means only one set of Oracle binaries must be modified during patch install and upgrade, simplifying management. The same capability exists for applications used for accessing the database. Since all servers in a cluster have access to cluster file systems, maintaining consistent reference data or application images and libraries across multiple servers is automatic. Moreover, redundant storage space required for multiple identical copies of applications and reference date in a shared nothing cluster is unnecessary. </li></ul><ul><li>Faster Recovery – With cluster file system and cluster volume management technology in place, mean time to recover of a database is greatly reduced. To recover a RAC cluster today, only one system in the cluster can perform database recovery. Therefore, archive logs of all the instances must be made accessible to the node performing recovery, to ensure that committed transactions are reflected in the database. Archive log access typically requires NFS mounting or copying those logs onto the surviving node. The process of giving the recovery node access to all of the archive logs can become additionally complex if a node in the RAC cluster will not even boot up and mount the file system containing archive log files. By setting up the archive logs on a cluster file system this step can be skipped because any surviving host in the cluster has access to the logs, resulting in faster and simpler database recovery process. </li></ul>Note: This can be done on NFS and other methods.
Veritas Advantages (continued) <ul><li>Improved Storage Utilization – The ability to create volumes, and file systems in these volumes, capable of being grown as necessary, provides a very important benefit that is sometimes overlooked. The administrator can provision only the storage necessary to support the application today, rather than what is necessary in the future. Since CVM and CFS allow transparent growth of volumes and file systems, expensive storage purchases can be held off until absolutely necessary. Also, the ability to add storage without interruption means the administrator can run the system with far higher utilization since space does not have to be pre-allocated. </li></ul>Note: This can also be accomplished via other methods .
Summary <ul><li>To handle the intensive demands to today’s IT environments, many businesses are turning to clustering solutions for both scalability and availability. In the Oracle environment, this means using the enhanced Real Application Clusters product. Clustered file systems along with the underlying clustered volume management greatly simplify the administration of the critical environments. Without sacrificing performance, clustered file systems allow for centralized system administration and reduces the time to recover in the event of system failure. The result is a reliable clustering platform that offers substantial costs savings and the flexibility to adapt to changing business requirements. </li></ul>