Software-defined storage (hereinafter – SDS) are considered to be the next milestone in the development of data storage technology. This technology is new on the market, so a clear and precise definition of this term yet. A similar situation was present at the dawn of the “cloud” as the term was called everything, not especially going into its meaning. By analogy with the above, “clouds”, the basic idea of SDS is abstraction from the hardware component and coming to a business-oriented model of the construction of storage and processing systems.
Now, in most cases, we are talking about software virtualization of data warehouses. This is not to say the substitution of one another at the expense of storage built-in functions or hardware locks, and the virtualization of computing local drives customer sites.
Architecture HPE StoreVirtual VSA
HPE VSA is a software product installed in a virtual cluster nodes. At the moment, they supported products VMware vSphere, Microsoft Hyper-V, and KVM. For vSphere and Hyper-V, there are plug-ins for integration with their management console that allows you to manage the entire infrastructure from a single point.
For a typical deployment HPE StoreVirtual VSA requires 3 assembly – two nodes for failover and the third to house Quorum Witness, who is responsible for data consistency across all nodes in the case of disconnection. Possible scenario using only two nodes, but in this case requires independent NFSv3 file storage service to accommodate the Quorum Witness.
From the point of view of the virtual cluster administrator, StoreVirtual VSA is a virtual machine under the control of Enterprise Linux, one for each node in the cluster. Installation can be made either through a special wizard, as well as through the deployment of OVF-image. In both cases, the user will need to configure the drive created by the virtual machine. The virtual disk size will depend on the block size specified when formatting data storage.
As with any software, StoreVirtual VSA requires computing power of the processor and memory, and to develop new solutions (or to verify compatibility with the existing cluster) HPE experts have developed a number of recommendations:
this product is licensed by the volume of which we highlight the hypervisor as a data store. It is important to understand that it would be “raw” the volume from the viewpoint of the end user. First of all local disks of your servers will be combined into a RAID-level group on the node RAID-controller. Get useful volume and will be licensed. After running each VSA they are combined in a cluster and summarize the disk space into a single pool, available then users (servers) on the block protocol (iSCSI).
Features HPE StoreVirtual VSA
In our time is not enough just to provide some capacity for data storage. The storage system must have the capacity to protect data from failures, optimizing the storage, management and processing, and StoreVirtual VSA can offer us all the basic functions that are present in the hardware storage systems:
- Thin provisioning – «delicate» volume, allowing the system to allocate space as the actual content data;
- Peer Motion – the migration of volumes without interrupting access to them;
- Multi-site SAN– distributed on two or three areas of a single volume;
- Synchronous and asynchronous replication;
- The creation of consistent snapshots at the application layer;
- Network RAID – building fault-tolerant schemes at the level of the cluster nodes
- Split Site – the creation of geographically dispersed clusters;
- Adaptive Optimization – duplex automatic tearing a 256 Kbyte granularity
All functions are available both through the CMC (Centralized Management Console), or via the plug-in for VMWare / Hyper-V management console.
The most interesting option would be Network RAID, which protects against data loss in the event of failure of a cluster node. In fact, this synchronous replication between nodes, and RAID level that regulates the number of copies of data blocks that are stored in the cluster.
For example, in the case of the level 10 Network RAID level recommended by the manufacturer as the most optimal, in cluster 2 will always stored copies of each data block. This redundancy scheme provides guaranteed protection against data loss in case of failure of the first cluster node (and in particular good fortune – up to half the nodes), but overhead will be 1/2. This means that when you buy a license for StoreVirtual VSA 2TB, we get a virtual store 1 TB of usable capacity.
The next level of protection – Network RAID level 10 + 1. In this case, the cluster is stored in three copies of each data block. The permissible level of “loss” – 2 knots. Overhead will be 1/3. This algorithm Network RAID 10 + 1 redundancy is the basis SplitSite technology, allowing you to create a cluster in three geographically diverse sites. The most remarkable thing in this technology is that, unlike the classical replication storage hardware we have no such thing as a primary and a secondary site. The application does not matter which of the areas is currently processed block, and in the event of failure of the treatment unit and to duplicate blocks at another site, the application does not notice the difference and, consequently, no downtime. On the other hand, it imposes strict requirements on the channel connecting the platforms together. The delay in data transfer (round trip) shall not exceed 5 ms.
Level Network RAID level 10 + 2 creates a copy of the blocks on all cluster nodes, which enables us to lose n-1 nodes, but ovrehed will be equal to 1 / n (where n – the number of nodes in the cluster). The minimum supported Configuration- from 3 units that is excessive and not applicable in real-world architecture.
By analogy with hardware raids are levels of redundancy checksum: Network RAID level 5 and 6. In these architectures, there are blocks with checksums that are uniformly distributed across all nodes in the cluster. This redundancy scheme is the most economical in terms of overhead (n-1 to level 5, and n-2 for level 6), and fault tolerance level is equal to level 10 and level 10 + 1, respectively. However, the need to consider checksums leads to a noticeable reduction in computational performance of the entire cluster, so that these levels are not recommended for use with heavy duty applications (the most obvious example – the database)
Product StoreVirtual VSA is part of a family giperkonvergentnyh systems solutions. And, like any other software component giperkovnergentnyh solutions, it is an analogue of the budget of its original hardware – a dedicated storage. The main objective of this product – the creation of failover secure storage for your virtualized cluster without buying additional equipment. We should not underestimate the functionality of Network RAID, which allows to implement synchronous replication between the nodes in the cluster, including geographically dispersed.
As can be seen from the description, this solution can be called functionally independent, giving its users the ability to store an entry-level hardware for a modest money.
In conclusion, it is worth noting that currently operates HPE promotional program : when buying any server HPE Proliant Gen9 you get a license StoreVirtual VSA 1 TB free. This is a great occasion not only to download a trial version for a detailed study of the product, but also to begin to introduce it into new projects without increasing their budgets.