Questions about storage Clusters

What are the common storage solutions provided by 3par, bluearc, equallogic, exanet, isilon, lefthand, netapp, onstor, panasas, polyserve, and SGI? If your answer is that these vendors provide Cluster Storage in different ways, congratulations. Similar to cluster servers, Cluster Storage has many different meanings and implementation methods to meet different application and business needs.

As a widely used technology, clusters can provide performance, capacity, reliability, and availability of proportionally increasing servers or storage resources, breaking through various restrictions of Single-host devices. Traditional storage systems are limited by their physical composition (such as the number of disk drives, number of connected servers, memory size, and controller performance, this may cause many functional limitations (for example, the number of file systems supported, the number of snapshots or copies ). Once the bottleneck of the storage system is encountered, it will constantly urge users to upgrade to a larger storage system and add more management tools.

Adjustable features in a common cluster include:

1. performance (bandwidth, iops-input/output times per second-etc.) can be increased to meet the needs of large-scale sequential read or write operations, or time-sensitive) the random read/write of is transaction-oriented (transaction-oriented) processing.
2. Availability-eliminates spof, transparent failover or self-healing capabilities.
3. storage capacity and service connection access (FC, Ethernet, and InfiniBand interface ).
4. Accessibility-including block-level (iSCSI, FC, InfiniBand) or NAS (NFS, CIFS, or other private file systems) and data sharing.
5. based on Open or private hardware and software, a tightly or loosely Interconnected Architecture (shared nothing, shared something or shared everything ubuntures) is implemented ).
The following table shows various Cluster Storage solutions, including cluster and parallel file systems, cluster file servers, cluster NAS, cluster iSCSI and FC storage. Most cluster implementation methods meet the capacity and availability requirements. Some Cluster Storage Solutions also support performance adjustment by controlling throughput and I/O operations to simplify usage and management.

Many features and examples of Cluster Storage

Storage	Block-level cluster (San FC/iSCSI)	File-level cluster (NAS)	Cluster and parallel File System
Access	Iscsi or FCP	NFS, CIFS	NFS, CIFS, HTTP, or other
Features	Performance, capacity, or availability can be upgraded	Virtual NAS server, performance upgrading	Host-Based Software or applications
Applicable Environment	Microsoft Exchange, SQL, and other sensitive block-based applications	Common File Sharing and Related Application Requirements	Specialized software or suitable hardware may be required for applications with high bandwidth requirements.
Representative manufacturer	Equallogic, lefthand networks, 3par	Netapp, onstor, bluearc	Ibrix, polyserve, isilon, verari, panasas, lustre, SGI, crosswalk

The evaluation of the actual capabilities of Cluster Storage products mainly relies on the capacity and performance scaling capabilities, and the accessibility (accessibility, block or file level) availability and ease of use. Cluster Storage is not synonymous with a large sequential bandwidth or parallel file system that can be used together with High Performance Computing (HPC) environments. Multi-purpose Cluster Storage supports traditional commercial applications, such as email, database, and online transaction processing (OLTP.

Different applications and environments of any level can benefit from the Scalability (including performance, capacity, availability and adequacy) and virtualization features provided by the flexible Cluster Storage System. For example, a small multi-node storage system can be deployed in a small and medium-sized enterprise environment to meet special application requirements, and then the system performance, capacity, and functions will be constantly increased as the enterprise grows.

Meet different application performance and service requirements

Eliminating spof is important for enhancing data availability, accessibility, and reliability. The cluster solution effectively prevents single point of failure (spof). Its n + 1 Redundancy feature, and the hot swapping feature and self-diagnosis capability of components can ensure that errors are detected, isolated, and eliminated before they cause problems.

There is also a storage system with n + 1 Redundancy architecture that has been in the gray area, and there is still controversy about whether it belongs to a cluster. In the n + 1 Redundant Architecture mode, there are two or more (n) Major I/O nodes or controllers, that is, the so-called NAS header, and slave or failover nodes. For example, EMC's celerra NSX and pillar axiom. The vendor's naming system makes the n + 1 model very messy. For example, the solution that includes raid arrays with dual controllers or dual NAS headers is called a cluster that improves availability.

Is a cluster a grid? It depends on your definition of grid. Consider grid as a service, architecture, or hardware or software based on other capabilities that span the distance. For this reason, there are many different vendors and industry definitions and opinions on how servers and storage environments form a grid or cluster. Although I have been working in this field for many years, I have found that I am still often on the basic components of the grid, the data collection and monitoring system (supervisory control and data acquisition, SCADA) lost in the discussion. Therefore, the discussion of Grid-cluster is full of variables.

Differences among various Cluster Storage solutions include:

1. How nodes are connected to each other (loosely or closely connected, open or private );
2. I/O performance and load balancing between nodes;
3. Suitable hardware, open, existing products (off-the-shelf), or support third-party servers or storage;
4. file sharing, including cluster file system software, host-based proxy or driver;
5. local or remote image or copy, point-in-time copy or snapshot;
6. Increase In virtualized storage modules to achieve automatic load balancing;
7. Performance adaptive Sequential read/write or random access;
8. The Distributed Lock Management (Distributed Lock Management) is consistent with the cluster features.

Understanding the meanings, types, and implementation methods of Cluster Storage helps you select the most suitable solution. Cluster Storage is ideal for environments of all sizes that continue to grow, allowing instant-in-time storage to avoid disruptive upgrades and increased management complexity.