IBM's Cloud storage technology

In 09, IBM announced the latest storage strategy for cloud computing, an application called "Enterprise Intelligent Cloud storage", a private cloud based storage and archiving technology designed to provide application support to enterprise customers. This cloud storage is implemented primarily through storage virtualization. This is in contrast to the existing Low-cost cloud storage application environment offered to customers by EMC and other vendors, with a distinction between two or three-level data replicas or storage applications in development and testing environments. IBM's cloud storage is an integrated solution based on IBM storage virtualization with a variety of storage devices.

IBM Cloud Computing solution is an advanced infrastructure management platform developed by IBM Cloud Computing Center after years of exploration and practice. The solution combines the industry's latest technology, fully embodies the concept of cloud computing, has been successfully operating within IBM for many years, and has a number of customer cases around the world.

The solution can integrate the enterprise's existing infrastructure, through virtualization technology and automation technology, build the cloud computing center owned by the Enterprise, realize the unified management, unified distribution, unified deployment, unified monitoring and unified backup of the enterprise's hardware resources and software resources, and break the monopoly of the application on resources, To help enterprises realize the concept of cloud computing.

The IBM Cloud Computing solution is composed of the following components:

Need to incorporate the software and hardware resources of the cloud computing center. Hardware can include network devices such as x86 or power machines, storage servers, switches, and routers. Software can include a variety of operating systems, middleware, databases, and applications such as AIX, Linux, DB2, WebSphere, Lotus, and rational.

Cloud computing management software and IBM Tivoli Management software. Cloud computing management software is developed by IBM Cloud Computing Center, dedicated to providing cloud computing services.

Cloud computing consulting services, deployment services and customized services. Cloud computing solutions can be developed two times according to customers ' specific needs and scenarios, integrating the cloud management platform with the customer's existing software hardware.

The following figure is the IBM Cloud Computing System architecture diagram

Cloud computing solutions can bring the following value to the enterprise:

Unified management of IT infrastructure (servers, networks, storage, software, etc.) to enable effective control of IT resources

Integrate standard processes and templates into IT management to reduce operational risk

Share resources, improve resource utilization

Reduce power consumption

Reduce system maintenance costs

Quickly respond to enterprise demand for IT resources

As you can see, IBM's storage virtualization is part of the IBM Cloud Computing solution.

So what is storage virtualization?

Add a virtual layer between the physical storage system and the server that manages and controls all storage and provides storage services to the server.

The server does not directly deal with storage hardware, the storage hardware changes, replacement, split, the integration of the server layer completely transparent.

Hidden complexity

Allow existing functionality to be integrated with

Free from the limitations of physical capacity

Storage virtualization for three-tier infrastructure

Modern storage virtualization technologies pool products from heterogeneous storage vendors in a special way, providing advanced features such as non-stop data migration and streamlined configuration.

This level of abstraction can be implemented in the three-tier infrastructure of servers, storage networks, and storage controllers.

Server-based

Some of the earliest storage virtualization was not in the storage infrastructure, but within the server and, more specifically, in the server's operating system.

While file systems help to easily invoke information, as more and more file systems are established, the storage space for the physical LUN will be exhausted, and a LUN will need to be established and assigned to the operating system to continue storing the files. To see which data is stored on which LUN, the operating system needs to assign a volume number, name, or identifier to each LUN. In Microsoft Windows, you may be most familiar with the letters assigned to each volume, such as C:\ or d:\, while in Unix these letters are/DEV/HDO or/DEV/HD1. As more and more files are created by applications and users, and more volumes are required, management will soon become very difficult.

As a result, the operating system vendor presents the concept of a Logical Volume manager (LVM). Similar to the way the file system combines chunks together to provide files, LVM combines volumes or LUNs together to provide larger, more flexible storage pools for applications, as shown in Figure 1. When the LVM space is nearly exhausted, you can expand the space by concatenating or adding a volume without having to reconfigure the application or turn it off. Conversely, if you want to split a larger volume into smaller chunks, LVM can help you partition to differentiate between different information, such as one block for the operating system and another for user data.

server-based storage virtualization, which was originally incorporated into the operating system as software, is still very popular today. The following are the main advantages of this approach:

Server-based Storage Virtualization is highly configurable and flexible because it is built into system software.

Because most operating systems incorporate this functionality into their system software, they are very inexpensive.

The storage infrastructure does not require the configuration of additional hardware and can be used in conjunction with any device recognized by the operating system.

Server-based virtualization also has many drawbacks:

While it helps maximize the efficiency and recovery level of storage resources, it can only be optimized on a server basis.

Mirroring, splitting, and computing parity tasks need to be handled separately, consuming the valuable CPU resources of the application and the internal

Each operating system has a different approach to file system and volume management, and businesses that use multiple IT vendor products have different technologies and processes that increase costs.

It is difficult to ensure data-tracking protection across the environment when data is migrated or replicated, whether local or remote.

Most operating system vendors, such as Microsoft, IBM, HP, and Redhat (Linux), have at least a certain capacity to provide virtualized storage resources. Vendors such as Symantec provide more advanced, server-based storage virtualization.

Based on storage network

In the late 90, with the advent of network-attached storage (NAS) and storage area networks (Sans), disks (and their controllers) were separated from the server, and all applications in the IT environment were able to share storage resources more efficiently. Storage networks become hubs for all information exchanged between servers and storage devices, and some storage vendors believe this is the best place to manage virtualization.

Based on networked storage virtualization, embedded storage resources Intelligent Management in the network layer, abstract the actual storage resources between server and storage array, divided into band or Out-of-band two ways.

In-band mode, sometimes called symmetry, embeds virtualization functionality into the I/O path between the server and the storage array, as shown in Figure 2, and can be configured in the San switch itself or in a dedicated device. All I/O (input/output) requests are routed through the device along with the data, and the server interacts with the virtual device without directly interacting with the storage device. The virtual device analyzes the request, looks up its mapping table, and sequentially executes the storage device I/O. Not only can these devices transform storage requests, they can also use their onboard memory cache data, provide data usage metrics, manage replication services, coordinate data migrations, and streamline configuration.

Out-of-band, sometimes called asymmetric, is not strictly defined in the I/O path as in-band, and it performs a specific lookup with a specific virtual San switch, as shown in Figure 3. The server maintains direct interaction with the storage layer through the smart switch. Out-of-band devices maintain a map of all connected storage resources for Sans (often called metadata) and instruct the server where to find them. In this two-step process, the server uses proprietary software or agents because the instructions must be sent over the SAN to work. Because the data does not pass through the virtual device, performance is only slightly affected, however, features such as data caching are not available.

Both in-band and Out-of-band modes provide storage virtualization with the following capabilities:

Pool Heterogeneous vendor storage products into a seamless access pool.

Performs replication between dissimilar devices.

Provides a single management interface.

However, only in-band methods can cache data, thereby improving performance.

Figure 3: Out-of-band Networked storage virtualization uses intelligent switches to maintain the interaction between the server and the storage array.

Vendors such as IBM, EMC, and Falconstor provide network-based storage virtualization solutions.

Base Storage Controller

Enterprise-class storage arrays for large enterprises in performance and functionality, with enhanced virtualization for physical storage resources (to some extent stronger than other systems). For example, RAID protects data from disk failure. Host storage domains can virtualize front-end connection ports, allowing multiple operating systems to use the same physical port. Many enterprise-class devices employ advanced switching architectures that support multiple physical connections to disk drives, provide balanced performance and resiliency, and logically segment the internal global cache to ensure quality of service and security.

In the last 15 years, various forms of virtualization have been applied to storage controllers, and some storage array vendors believe these functions should be extended to storage resources outside of the device. Controller-based storage virtualization arises (see Figure 4).

This new generation of storage arrays allows other heterogeneous vendor storage arrays to connect directly to their controllers. As a result, the external storage resources that are based on them are "present" and can be managed in the same way as internal disks. This approach has many advantages, including the need to remap LUNs or scopes, which means reducing one layer of management and greatly reducing network complexity. Virtualization in this way, advanced microcode software installed on storage controllers becomes external storage resources, as in the array, and the host does not know where they are actually connected.

Figure 4: Controller-based storage virtualization makes external storage as internal storage.

Hitachi Data Systems (HDS), NetApp v series, Sun Microsystems and HP all use this form of storage virtualization.

The choice of the virtualization architecture-which is superior to the inferior

The above diagram is in-band, Out-of-band, storage-Controller based virtualization architecture.

svc-Advantages of using band-block level virtualization

IBM SAN Volume Controller implementation of storage virtualization

IBM SAN Volume Controller (abbreviated SVC) is a hardware and software integrated product, including IBM System x server, Linux kernel based storage operating system and professional virtual storage software. SVC can support more than 120 disk storage systems, including IBM and non-IBM, and help you simplify your storage infrastructure, lifecycle management of information, and maintain business continuity by consolidating storage capacity into a single storage resource pool.

SVC Logical Architecture

SVC Outstanding Performance:

The fastest controller in the SPC benchmark

Svc has the fastest SPC-1 benchmark test value (SVC 4.2:272,500iops)

Svc has the fastest SPC-2 benchmark value (SVC4.2:7080MB/S)

Many cases show that SVC performance has been significantly improved (up to 10 times times)

SVC can improve the access performance of storage part in application system

Especially for older storage devices with small caches

For caching-friendly business flows, performance improvements are most obvious

For cache unfriendly business flows, performance is as good as before use

The performance growth of SVC presents a linear trend, both for iops and bandwidth.

Introducing SVC into a storage network typically does not negatively affect performance.

In the worst case scenario, about 0.6% of the performance is reduced.