Precautions for storage planning before Oracle11g RAC + ASM Installation

Precautions for storage planning before Oracle11g RAC + ASM Installation

When should we start to better optimize the performance of Oracle databases ?, Based on my years of practical experience, I think it is best to start with the storage plan before installation. Storage Planning is not about adopting the most advanced technologies, but about adopting the most appropriate business needs. The general principle is as follows:

1. Data Files, backup files, and archive log files are stored in different disk groups instead of different Luns in the same disk group, however, the disadvantage of this method is that each disk group requires a hot backup disk, which seriously wastes space;

2. For OLTP systems with frequent read/write operations, it is recommended to use Raid1 + 0 for the disk groups where data files and archive log files are stored, so as to ensure fast read/write operations. Online redo logfile and control file are stored in both the data zone and archive zone at the same time. This improves availability and recoverability while ensuring read/write performance. RAID 5 redundancy can be applied to the disk groups where the backup files are stored.

3. Generally, data files and backup files and archive log files are stored in two different disk groups, instead of different Luns in the same disk group, perform Raid1 + 0 for both disk groups. This will not affect the performance, and can save a hot spare disk to improve the space utilization. However, if the data volume is large, partitioning the backup disk space by Raid1 + 0 may also lead to a waste of space;

Based on the above contradictions, we can see that there is an important factor in the disk size and quantity. It is also necessary to consider the size of the database and the specific application.

In addition to the backup area, too large hard disks are not recommended for other areas. If the database is not large and the number of disks is too small, it is better to make all disks into RAID 1 + 0, which can increase the overall I/O bandwidth, it can also save space waste caused by multiple hot spare disks.

With the sharp decrease in the price of SDD disks, SDD and SAS can be used properly because the IOPS of a 10 k SAS disk is only 120 to 150, the IOPS of SSD hard disks is even more than 10 times. Therefore, it is worth trying this combination for systems with high I/O performance and concurrency requirements. For example, assume that the current size of a customer database is 1 TB, the data increment is 100 GB per month, and the concurrency is more than 1000. You can use the following two methods:

• 6 900 GB hard disks. After completing RAID 5 groups and hot spare disks, the actual available capacity is: 4*3.6 GB = TB. 7 SSD disks mainly run online data. After RAID 5 is completed, actual available capacity: 5*400 GB = 2 TB;

• Configure 11 SAS disks with a rotation speed of 15 kb for RAID 1 + 0. After two Luns are partitioned, further partitioning is performed in ASM;

For databases with large space capacity, in addition to the above factors, another important factor is the number of ASM diskgroups and the size of each asm disk required by the DISK group. ASM disk groups are not too many, and names can be normalized as much as possible. For example, they can be divided into DATADG1, DATADG2 ..., DATADGn to store data files and log files, FRADG1, FRADG2 ..., FRADGn to store archives (including backup files), BACKUPDG1, BACKUPDG2 ..., BACKUPDGn to store backups. For the ASM disk under each ASM disk group, it is best to use an appropriate size, which is not too small or too large, in combination with the database size and future growth. For example, if the incremental size of the database is 3 TB in three years, it is best to keep the size of each asm disk in the range of GB to GB. Why, then 3000G/300G = 30 asmdisks. In this way, when the disk group is composed in ASM, it looks neat and reasonable. If the number of disks is too large, it will not only lead to chaotic management, but also lead to human errors. In addition, managing so many disks for ASM also brings a certain amount of management overhead.

Finally, some DBAs may put archiving and backup separately on independent Luns and adopt the file system instead of the ASM disk management mode. The advantage is that even if the ASM fails to log on, however, backup and archive files can be seen in the file system intuitively, which can avoid certain risks and make people feel at ease. If you use this mode and use the LINUX operating system, we recommend that you divide the file system by Linux LVM, which will be very convenient for future expansion. However, this method also has some disadvantages, that is, because it is a Linux local file system, rather than a Cluster file system, the implementation of archiving and backup for two nodes is that the two nodes write independently only belong to their own partitions and can share read-only data. If you accidentally configure the partitions to the mount mode that both parties can read and write, the data will be overwritten and lost.

 

How Does Oracle ASM Add a new disk to a disk?

 

Oracle 10g manual creation of the ASM Database

 

Solutions to various problems after installing Oracle 11gR2 in Ubuntu 12.04 (amd64)

 

How to change the sys password of Oracle 10g ASM