Windows Server 2012 Virtualization Test: Storage (i)

In the computer world we can see in a way, that is abstraction. 1946 von Neumann presented the basic structure of the computer comprising: calculators, memory and I/O devices. This is probably the most important abstract of computer, which directly affects the development of computer hardware and software in the next few decades. Each of us has a different understanding of computer calculators, memory, and I/O devices, and you may not know all the computing chips, the various storage devices and input, but this will never prevent you from distinguishing them. The various parts of a computer system may be indispensable, but they must be replaceable, and that is the power of abstraction. Without abstraction, we might have to develop different operating systems for each set of hardware, and it would be unthinkable to develop different applications for each operating system.

Of course, we still feel that there are too many kinds of hardware and software systems, and we need new levels of abstraction. Virtualization is an abstract process, an abstraction of hardware and software resources. One of the purposes of virtualization is to allow the upper layers to not perceive the underlying differences and to provide a unified interface for the upper layer to use the underlying resources.

In the ideal state of hardware virtualization, it may be possible to use hardware without any limitations, and the application can take full advantage of hardware resources without the need to perceive differences in hardware devices such as compute, storage, and networking. But the reality of the situation, we still need to consider more or less to consider the limitations of hardware. Next, let's consider the most critical role in hardware virtualization-storage.

First, the status of storage development

For the store everyone will have an intuitive understanding, from tape, floppy disk, CD-ROM to hard disk. According to von Neumann's theory of computer architecture, the concept of storage is literally as simple as any device that can be used to store data, and as an abstract independent concept, storage devices can even be relatively independent in the computer architecture. With the development of the network, storage is no longer confined to the inside of the computer shell, the network storage is becoming the mainstream in the server field.

In the storage device is the most important should be hard disk, the drive to read and write speed, capacity and quality of the pursuit of hard disk technology has undergone several changes. From the interface standard of the motherboard with the ATA (IDE), SATA, SCSI, SAS, FC and InfiniBand, they have a certain difference in interface, transmission media and protocol. Where ATA interface is often connected to familiar IDE devices, SATA (Serial ATA) is serial Ata;sas (Serial attached SCSI) is serial scsi;sata with SAS is twin, SATA hard drive can connect SAS interface, but not compatible The FC (Fiber channel) is fibre-optic channels, which in fact do not necessarily use fiber optics, but also use copper cables. Need to learn more about the technical details of each interface please Google.

With the development of networked storage, storage modes such as Das (direct-attached Storage), NAS (Network attached Storage), SAN (Storage area Network) are distinguished. is a comparison between them, where each dashed box is represented as a relatively independent whole, and the left side of the arrow describes an interface or connection, and the right side describes the technology scheme for implementing the storage. The diagram shows just the simplest way to organize or connect in each pattern. The way the file system in Das connects storage is not limited to the various interfaces and cables inside the machine (such as a computer connected to a built-in SAS hard drive), or external interfaces and cables (such as the storage via an external SAS cable), which can complicate the situation. Nas or San as a whole can easily become part of the DAS. But the biggest difference between Das and the other is that DAS does not require network support. The biggest difference between Nas and SAN is that NAS is file-based storage, and SANS are block-based. NAS storage is more of a standalone file server, but the San behaves more like a disk, so the SAN can be a much lower part of the NAS network.

Second, Windows Server 2012 Storage

Next, let's look at several important improvements in Windows Server 2012 that support storage virtualization. New and improved multi-storage features in Windows Server 2012 support storage virtualization, the most interesting of which is the iSCSI target server, SMB3.0, and storage space. Windows Server 2012 and its clusters can easily implement current mainstream storage scenarios for testing or production environments, and Microsoft has a unique storage solution based on file servers and file server clusters (SMB3.0 and storage technology).

1. iSCSI Target Server

ISCSI (Internet Small computer system Interface) is an Internet small computer system interface. iSCSI uses an Ethernet connection between the server and the storage system to encapsulate the transmission of SCSI instructions and data based on the TCP/IP protocol, creating an IP SAN. iSCSI and IP sans should be the most cost-effective storage solution available today.

In Windows Server 2012, iSCSI target servers (iSCSI Software Target server) become a server role built into the file and storage services, integrated in Server Manager, Additional download installs are no longer required (previous versions of the server need to download a standalone installation package for installation), so deployment and updates are easier. After finding the document, we found that the iSCSI target server can provide the following services:

• Network and disk-free booting: You can quickly deploy hundreds or thousands of disk-free servers by using a network adapter or software loader that supports startup. With differencing virtual disks, you can save up to 90% of your operating system image storage space. This is useful for large deployments of the same operating system image, such as deploying a large room or deploying servers in a large-scale cluster.
• Server application storage: Some applications require block storage (for example, Hyper-V and Exchange server). ISCSI target servers can provide continuously available block storage for these applications. Because storage can be accessed remotely, you can also merge block storage for Central or branch office locations. This is the most important feature of iSCSI.
• Heterogeneous storage: iSCSI target servers support non-Windows ISCSI initiators so that they can be stored on Windows Server that is shared in a hybrid software environment.
• Dev, test, demo, and lab environments: When the ISCSI target Server role service is enabled, it turns any Windows server into a block storage device that can be accessed over the network. Storage arrays are typically very expensive, and in a test environment we can use a Windows Server computer that deploys the iSCSI target server role to act as such a storage device. This is a very useful feature, and if you want to do virtualization testing without a separate storage array, you can use the iSCSI Target Server service to make any server that is installing Windows Server a storage array.

From the above, it is an indispensable helper in a test environment if the production environment of a high-performance iSCSI target server is not considered. In addition, the iSCSI target server can be configured as a cluster role for the Windows Server 2012 failover cluster, while achieving high availability by configuring MPIO. There is no difference between initiating a connection to an iSCSI target server and initiating a connection to another iSCSI device.

2. File Server/smb3.0

In Windows Server 2012, the file server is another important server role. Access to the file server has two Optional protocols: the NFS (Network file system) and the SMB (server Message Block)/CIFS (Common Internet file System) protocol. When configuring a file server, you will be asked to make a selection, so let's look at their differences:

• NFS was first developed by Sun Corporation and is the most common network file sharing protocol. NFS allows the system to share its directories and files with other systems on the network. Users and applications can access files on the remote system as if they were accessing local files. NFS is more used in Unix or Unix-like systems, and of course Windows supports that protocol.
• SMB originally originated from IBM, but then Microsoft supported and improved it, initially SMB was based on NetBIOS to establish a file sharing protocol, but in order to extend SMB to the Internet, to get rid of the reliance on NetBIOS, Microsoft to defragment the SMB protocol, and rename it to CIFS. At the heart of the Windows Networking environment is SMB/CIFS, where many Microsoft Department network applications can be based on the SMB/CIFS protocol, including file sharing between Windows systems. The SMB/CIFS protocol, such as Samba, is required to provide file sharing services to Windows customers on UNIX class systems.

SMB3.0 is the latest version of SMB and offers many new features such as SMB transparent failover, SMB multichannel, SMB Direct (SMB Direct), and more. The SMB3.0 can achieve 1Gbps file access speed over Ethernet, which can rival Das performance. Thus, in a Windows network environment, SMB can be another alternative to iSCSI.

In Windows Server 2012, a file server can also be configured as a cluster role for a Windows Server 2012 failover cluster, and in order to meet the needs of the file server for different applications, there are several options for NFS-and SMB-based options in the cluster role configuration process.

The core Hyper-V Server for Windows Server virtualization has high requirements for storage networking, and here are a number of options: While Hyper-V virtual machines can be deployed in Das, Nas, or SAN storage, for high availability, SAN-based is the most common (iSCSI-based IP San and FC-channel-based FC San), and you can now choose a file server cluster-based Hyper-V over SMB (a small network that can be used in conjunction with the "Storage space" feature to be mentioned later in this article), which will be implemented later.

3. Storage space

Storage space is a new feature in Windows Server 2012 that provides an efficient and convenient storage virtualization solution.

In simple terms, it can put a standard storage device into a storage pool and virtualize it to create a virtual disk called storage space. Currently compliant storage devices include disk or JBOD disk arrays that are directly connected via ATA, SATA, SCSI, SAS, or even storage devices that are connected via USB, but not those that are virtualized and tiered, such as VHDs virtual disks, pass-through disks in virtual machines, RAID arrays, and other storage pools do not support iSCSI and FC Fibre Channel.

To put it simply, storage space implements features like RAID cards, but the implementation level is on the operating system, not on hardware, which is the implementation of the so-called software-defined storage.

If a RAID card is present in the storage device, it is wasteful to disable all features of the raid (if allowed) to be compatible with the storage space feature. Therefore, the use of storage space virtualization, there is no need to purchase a RAID controller disk array, and direct procurement of JBOD disk cabinets can be, save a lot of money.

such as the Xia Guan Online has clearly demonstrated the concept of storage space:

A virtual disk created in storage space is an elastic store that cannot change the storage layout (that is, cannot switch freely in simple, mirror, and parity modes) and provide a way (thin and fixed), but can adjust the capacity when the virtual disk is defined. Microsoft offers three storage layouts, similar to RAID levels, as defined on the tednet:

• Simple: Data is striped across physical disks, which maximizes capacity and improves throughput, but reduces reliability. This storage layout requires at least one disk and cannot provide protection against disk failures.
• Mirroring (MIRROR): Data is replicated on two or three physical disks, which improves reliability but reduces capacity. This storage layout requires a minimum of two disks to protect against a single disk failure, or at least five disks to protect against the two disk failures that occur simultaneously.
• Parity check (Parity): Data and parity information are striped across physical disks, which improves reliability but somewhat reduces capacity. This storage layout requires a minimum of three disks to protect against a single disk failure, and at least seven disks are required to protect against two disk failures.

Storage space features can also be applied in a failover cluster in Windows Server 2012, but the requirements are higher:

• The virtual disk supply must be fixed (fiexed, which occupies a fixed capacity when it is created).
• Virtual disk parity storage layouts are not supported in Windows Server 2012, but are supported by R2.
• All physical disks must be linked via SAS.
• All physical disks must support PR (persistent reservation) and fail over cluster validation tests.
• The cluster requires shared storage, so the JBOD disk must physically connect all nodes through the SAS (all nodes can be connected via SAS or use SAS switches), so DAS is not supported.

From the above requirements, to use the cluster storage space feature limitations are more obvious, to compete with the San solution is still a long way to go. For more details on storage features, you can access storage Spaces frequently asked Questions (FAQ)

Third, virtualized storage networks

The key to implementing virtualization in Windows Server 2012 is to deploy a Hyper-V server that guarantees the Hyper-V virtual machine to run. In principle we can place Hyper-V virtual machines in any storage, based on Das, Nas or SAN storage, which is optional, and Hyper-V Server 2012 has even implemented live migration of virtual machines between standalone hosts that are not shared storage. But to achieve high availability, we typically deploy a hyper-V server failover cluster, which requires the cluster nodes to have shared storage, so we have two options for shared storage: San-based (iSCSI-based IP SAN and FC Channel-based FC SAN) and a file server cluster-based Hyper-V over SMB.

1. IP san and FC San based

As a typical two-node cluster-attached storage for high-availability scenarios, you can see that there is no essential difference between an IP san and an FC San in a network topology, both of which can achieve IO high reliability and load balancing through Multipath IO (MPIO). The storage server here is the traditional storage server (storage array with CPU and controller) that provides shared storage for the Hyper-V server cluster. In a test environment, you can take advantage of the storage server that is required by the iSCSI target server role in Windows Server to implement an IP San in a normal PC.

2. Hyper-V over SMB

Hyper-V over SMB is a unique storage solution for Microsoft. From the network topology, there is more than one level of Windows File Server cluster as a provider of shared storage for Hyper-V clusters, which you might think is superfluous, but there are several benefits to such a scenario:

• Save money, reuse old assets, and use inexpensive hardware. Without the high availability of a file server, any PC connected to a storage device can act as a provider of shared storage for the Hyper-V cluster, and the storage of the file server can take any form of storage mode (DAS, Nas, and SAN) and can even be a local disk in the PC. If you implement high-availability file servers, the file server failover cluster itself also requires shared storage, so the file server cluster either uses a SAN or a shared SAS connection JBOD, but it should be recommended to connect to JBOD via a shared SAS as shown. JBOD is cheaper than a traditional RAID storage array, it has no controller, no raid high availability, but with the Windows Server storage space feature, you can achieve high availability and scalability in a file server cluster.
• Scalability. The nodes of the file server cluster are extensible. Additional storage space can be extended, such as each file server node can connect more than 2 JBOD, with storage space characteristics to achieve large-capacity virtual disk (storage space features require a Microsoft-validated JBOD to support, you can check the JBOD products verified).
• Compatibility. A single layer of file server or file server cluster, the upper-level application only needs to access the file server without knowing the implementation details of the underlying storage, so it can be compatible with different models of products. such as two jbod can come from different manufacturers.
• Other benefits, the perfect fit for Hyper-V and other Microsoft products, and the ease of use of Microsoft's products are not mentioned.

In fact, the file server cluster can also share the storage through the NFS protocol and ISCIS through SMB. However, SMB 3.0 has significantly enhanced the benefits of adopting the SMB protocol.

Iv. Cluster Shared Volumes

Cluster shared volumes are a new feature of Windows Server R2 for failover clustering. Simply put, a cluster shared volume enables multiple nodes to have true synchronization of access to the same location in the network, with each node having the latest state information for that location.

When we used the Windows file Sharing service, we had the experience of adding file A to the shared folder \\fs01\share opened in Node 1, but \\fs01\share the same location opened on Node 2 without file a. Network location \\fs01\share Although it is a shared folder, only one node can maintain its latest state at the same time. There is no such problem when accessing cluster shared volumes on each node of the failover cluster.

The benefits of cluster shared volumes are far more than this:

• Each cluster shared volume is mapped to the Clusterstorage folder under the system partition of each node, because each node accesses the cluster shared volume as if it were accessing the local path. If the system partition is in drive C, the path to cluster shared volume 1 defaults to C:\ClusterStorage\Volume1\, where the name of the Volume1 can be changed arbitrarily.
• Multiple virtual machines can be placed on the same cluster shared volume, which is the same LUN, and can fail over normally (the unit of failover is no longer a LUN, but a virtual machine), which takes advantage of disk space and eliminates the tedious task of configuring LUNs for each virtual machine.
• Even if the connection between a node and the San is interrupted or part of the network is interrupted, the cluster can respond normally, and the cluster will reroute the cluster shared volume traffic through the San or the intact part of the network.

Windows Server 2012 has made great improvements to the cluster Shared volumes feature, and the implementation principles for cluster shared volumes can be referenced in detail in Cluster shared Volume (CSV) Inside out and Cluster shared Volumes Rebo RN in Windows Server 2012:deep Dive these two articles.

Windows Server 2012 Virtualization Test: Storage (i)