Set up DB2 and Aix to match striped technology to improve I/O performance

Stripe technology is a widely used I/O load balancing technology, can greatly improve I/O performance without additional hardware input. However, in an environment where this technology is used, only the I/O processing of the database and operating system is well planned and set up to match the stripe technology to achieve the best I/O performance. For this topic, the DB2 database and AIX operating system are examples of how to plan and set up relevant content.

Here are a few ways to introduce:

Introduction of Stripe technology

Reasonable planning and setting of bar with off parameters in DB2

Properly plan and set bar with off parameters in AIX

Summarize

Introduction of Stripe technology

Introduction and application of strip technology

Disk conflicts may occur when multiple processes are accessing a disk simultaneously. Most disk systems have restrictions on the number of accesses (I/O operations per second, IOPS) and data transfer rates (the amount of traffic per second, TPS). When these limits are reached, the subsequent process that requires access to the disk needs to wait, which is called a disk conflict.

Avoiding disk conflicts is an important goal of optimizing I/O performance, and the optimization of I/O performance is very different from the optimization of other resources, such as CPU and memory, and I/O optimization is the most effective means of balancing I/O to the maximum. This involves sharing the I/O access load on a hotspot disk to other relatively idle disks, that is, I/O load balancing. Before the advent of some mature disk load balancing technologies, DBAs need to understand and predict the amount of I/O load on each system, by manually configuring each data to a different storage location to share I/O load to achieve load balancing purposes.

Stripe technology is a kind of automatic load balancing I/O to multiple Physical disk technology, stripe technology is a piece of continuous data into a lot of small and separate them to different disks. This enables multiple processes to access several different parts of the data at the same time without causing disk conflicts, and can achieve maximum I/O parallelism with the need for sequential access to this data, resulting in excellent performance. A lot of operating systems, disk device vendors, all kinds of third-party software can be striped. With stripe, DBAs can easily balance the I/O load without having to manually configure it.

Figure 1 depicts a continuous data distribution without strip processing, and figure 2 depicts the distribution of a continuous piece of data that has been striped, and from which we can see that the maximum concurrency capability for continuous data reads and writes is found in Figure 2.

Figure 1. Continuous data without stripe processing

Figure 2. Continuous data that has been processed by strips

Because of the superiority of stripe in I/O performance issues, it is possible for multiple tiers or platforms in the computing environment in which the application system resides to involve striped technologies, such as operating systems and storage systems, which can be used in two levels. The following figure shows the two-level I/O architecture.

Figure 3. Operating system and storage System two tiers of I/O architecture

At the operating system level, we can use software strips (such as the LV bar in LVM to bring implementation stripe technology) or hardware bands (such as various RAID). At the storage-system level, the vast majority of products present on the market offer a wide variety of stripe technologies (such as various RAID).