This article belongs to SQL Server AlwaysOn from getting started to Advanced series articles
Objective:
This section explains the contents of SQL Server storage, which is relatively short compared to other subsections. This section provides some suggestions for using storage in a clustered or non-clustered system. Of course, the focus is on the selection and configuration of storage for a standard AlwaysOn available group configuration process.
AlwaysOn deployments are first established on a Windows Server Failover Cluster (WSFC). And each server typically has a separate instance of SQL Server. In addition, each server uses its local storage to hold database files (data files, log files, backup files, and so on) for a separate instance of SQL Server. Although all partner nodes belong to the same cluster, they do not need to be based on a hard disk witness or a failover instance, and there is no requirement to share storage. This avoids the risk of a single point of failure for shared storage in an FCI. However, AlwaysOn available groups can use an FCI as an available replica. This not only re-introduces the risk of a single point of failure, but also increases the complexity of the cluster to the node.
Now, let's take a look at the core content of the storage system:
- Local Storage (localized)
- Networked Storage (networked)
Here is a detailed introduction:
Local attached storage (locally attached Storage):
In this mode, the local storage is directly attached to the server, the hard disk is inserted directly into the hardware backplane (backplane), and then connected to the server's motherboard. Older configurations may contain an extended RAID controller that will connect to the PCI bus via a 68/80-pin cable.is a typical local store. This is relative to the way the path is broken and the complexity is low, providing quick access to the hard drive. The backplane has an input and output BIOS that can be used to control the hard disk redundancy capability of the RAID array across the local hard disk, but due to hardware server limitations, it is usually only 16 hard drives available.
This is a typical separate store in a node that does not have networked storage, and in WSFC there is no separate storage share to other nodes. This also makes it unnecessary to replicate the storage during the physical location of the sub-node.
Networked Storage (Network Storage):
Networked storage can be provided as a resource to multiple computer systems. There is a central repository that makes it easier to manage these hard drives by reducing the number of touchpoints that are running through multiple arrays of each server. As shown, there are often many servers in the system that are interconnected via fiber optic (Fibre CHANNEL,FC) networks, often referred to as "Fabric." The computer passes a host Bus Adapter (HBA card,A host bus adapter is a network and exchange, a board that can be plugged into a computer or a mainframe), the HBA card is actually similar to a network card.Each server can also be interconnected via the iSCSI network, which is relatively new but bandwidth constrained (1Gbps). It runs on a standard, isolated TCP\IP network. Servers typically use a dedicated network adapter that is used only for iSCSI and TCP traffic to reduce load, meaning that iSCSI traffic control is distributed. Modern iSCSI is already capable of handling bandwidth data up to 10Gbps. One of the benefits of iSCSI configuration is that it is more economical than traditional FC networks. But that's not always the case.
When many servers send requests to and receive results from the storage process, you can quickly discover how much traffic has been generated in the FC network. As with the TCP\IP network, you will find that the FC network is blocked. The network of the storage area can then produce performance problems. In a complex SAN configuration, multiple switches are connected to a large number of network cables and additional power requirements. As you can see, the data flow path and complexity are significantly larger in this case.Making I/O requests in such a long path and complex route consumes a lot of events and other overhead. What does this kind of system offer about consolidated storage? This type of storage makes it easier to provision and deliver resources to large-scale data. And then just like virtualization, not every entity can be used.This mode of storage is also commonly used in SQL Server FCI, LUNs from the disk array, and the number is huge. The disadvantage here is that the array may be formatted with a block size of 128KB. This size is not optimized for SQL Server. The advantage is that, when properly configured, storage requests can be almost less than the end array. Because the request can occur directly in the high-speed memory buffer, the cached data is then brushed to the hard disk at the appropriate point in time to reduce the impact on performance. In the case of a power outage, the backup power source will also flush the cached data to the hard drive to avoid data loss.There is also a network store that can be used to share storage in highly available nodes and avoid the overhead of multiple host connections. This type of storage, called direct attached Storage (DAS, directly attached storage), is designed specifically for applications that can use private fiber-based connectivity and can basically be localized. The following is a typical private high-availability storage configuration:
In this scenario, if you want to create a private, highly available cluster, it is a little better than local storage. Some storage vendors provide fibre-connected devices and can have up to two hosts connected to multiple paths in highly available scenarios. Multiple array enclosures can increase the amount of available storage in the way.This type of storage can also be used in SQL Server FCI. This approach is suitable for sharing storage for a small or simple cluster of nodes in a particular environment. You may have noticed that the LUNs way is a box, because not all Windows systems have a separate physical array on the bottom of the logical hard disk. Scenario is also the most common configuration in which the disk is set to a larger array.Imagine a big cake. Or, in this case, the array created from the physical hard disk pool. Cut down a piece of cake or draw a LUN from the array to use for Windows as a logical hard disk.
Summarize:
This article describes the typical storage requirements that are stable and robust for clustered and stand-alone instances of SQL Server. The next section focuses on the infrastructure required to support WSFC and FCI and AlwaysOn available groups.
SQL Server AlwaysOn from getting Started to Advanced (2)--storage
