HP StorageWorks

The HP StorageWorks Enterprise Virtual Array EVA for short, released in 2001 and has since become the only fully virtualized storage system on the market. The system was developed by DEC, Compaq and taken in connection with the acquisition by Hewlett -Packard integrated in 2002 into the product range.

EVA generations

The product range is divided into different Enterprise Virtual Array models.

Currently, five models are available in which EVA6400 EVA8400 and represent the latest models of a new generation.

Technical structure

All EVA systems work from the hard drive to the host port Fibre Channel, partly over copper, partly fiber optic. The disk shelves and hard drives are connected via two redundant (for EVA6400 EVA8400 two FC loops and three loops FC ) FC -AL connections connected to the controller pair. Each controller (two per controller EVA) has two or four host ports to which the controllers are connected to the fabric. The disk shelves are directly connected to the controllers at the beginner systems, EVA4x00. Again, the disk shelves are connected redundantly. For models 5000, 6x00 and 8x00 Loops Witches are used (For EVA4400, 6400 and 8400, this task is taken over by so-called " Switched IO Modules", they have no switches). These loops Witches break the physically existing ring structure in a physical star topology, but still is a logical ring. Through the use of loops Witches even entire disk shelves during operation can be added. Theoretically, this goes for a EVA4x00, but HP does not recommend. The entire back end, so the connection of hard disks to the disk shelves, and the connection of the disk shelves to the controller works with 2 GB FC. The EVA4400 will change this condition, and works from the hard disk to the host port with 4 GB FC. The claim that 2 GB FC a bottleneck, has not yet been proven. To create a EVA8100 up to 210,000 IOPS and 1500 MB / s An EVA can manage 8-240 disks. Hard drives are available with FC and FATA interface. FATA drives are ATA hard drives with Fibre Channel interface. This way you can mix FATA and FC disks in a disk shelf, a great advantage over many other manufacturers in this environment. For systems that SAS and SATA HDDs, this is also possible, but these techniques tend to come before in the entry level segment. Currently different hard drives are available:

The controllers are each housed in a separate, 2U controller shelf high. Thus, these two controllers shelves need 4U in a rack. The controllers are redundantly connected to each other via Mirror ports. Depending on the controller, the controller pair has 4 to 8 GB of cache. The controllers work with PowerPC CPUs and operating at XCS (name of the firmware) based controllers always in an active / active network. Older EVA3000 and 5000 systems with VCS 4.x ( name of the firmware ) is able to work in a restricted active / active network.

Memory management

The memory in an EVA system is fully virtualized. While traditional storage systems work with RAID sets and logical drives, an EVA works with Disk Groups and Vdisks. Instead of the classical case to connect hard drives to RAID sets with a specific RAID level, and then to make it logical drives, hard disks are organized in an EVA Disk Groups. There are at least eight disks of size and speed necessary to form a Disk Group. Therefore, the smallest EVA system always consists of a so-called 2C1D 8 HDD Configuration: 2 controllers, 1 disk shelf, and 8 hard drives.

Protection level

Within a Disk Group are always all disks active. There is no spare disks. Instead of using spare disks, is kept free as much space on all disks in a Disk Group, the largest hard disk can be restored to the Disk Group there. By eliminating spare disks go no IOPS lost by inactive disks. This releasing of space is called a Protection Level. There are three levels to choose from: None, Single, and Double. When Protection Level None no memory is allocated. For single double the capacity of the largest disk is reserved in the Disk Group at Double four times the capacity. Thus, the EVA cushion the failure of one or two hard drives. The current, twice the capacity is for technical reasons, since the EVA combines all hard drives internally in pairs. If one drive fails, the data of the failed disk, as well as her partner, called the Widow, recovered in the free space. As long as free space is left in a Disk Group, the controllers detect the disks in the disk usage restores. Is not that enough of this, the reserved space is used.

RAID level

All EVA systems dominate the RAID Level 0 VRAID, VRAID 1 and 5 VRAID A vdisk, so a logical drive that can be presented to a host, always extends across all the disks in a Disk Group. Includes a Disk Group, for example, over 56 hard drives, and it is a vdisk with 1GB affixed this 1 GB vdisk is distributed across all 56 disks. According performant is this vdisk. Vdisks with different RAID levels can be easily mix within a Disk Group. Also this is not possible in all systems of market competitors. HP strongly recommends not to mix different plate sizes and techniques within a Disk Group. It is a recognized best practice to create as little as possible, Disk Groups, each with as many plates. From this one can conclude that it is better to use 32x 72 GB hard drives instead of 16x 146 GB hard drives.

It is quite possible that a productive Vdisk is operated in a VRAID 1, a clone for tests but with VRAID 5

Internal management

Internal arrange the controller all hard drives a Redundant Storage Set, shortly RSS, too. This RSS feature 6 to 11 hard drives. With an EVA system with 16 disks so two RSS are available. Each RSS is a single RAID Protection domain. The EVA used, for example, for VRAID 1 a 4D 4 P, at VRAID 5 a 4D 1 P mechanism. Each four Datachunks stored on four different plates in a RSS, as well as the one Paritychunk. In VRAID 1 it is comparable. A hard drive can therefore contain different data and Paritychunks. If one drive from now, is only used an RSS affected. Now it becomes clear why the EVA disks binds to pairs: These couples always form a part of a 1D 1 P VRAID 1 If one disk fails, the data in the free area to be restored, and there also mirrored again. The data are therefore after recovery before mirrored back. Depending on the status of RSS, this may None, mirrored or parity be, can even be intercepted an entire disk shelf failure. The EVA controllers always ensure an optimal distribution of plates. For example, if eight disk shelves and 64 hard drives present, the EVA will try to make more RSS with eight disks, with an RSS perpendicular of all eight disk shelves. To protect the metadata are those at the beginning of a disk group and are distributed over multiple disks. It will be kept up to five copies.


The system is managed via a software called HP StorageWorks Command View EVA software. This software is installed on a management server. This server is only necessary for the administration. If it fails, the system continues to run as usual, only an administration is no longer possible. With a Command View EVA installation up to 16 EVA systems can be managed. It is planned to relocate in the next generations, the management on an embedded Command View server. The EVA4400 is prepared here already.

HP StorageWorks Business Copy EVA snapshots, clones, and Snapclones Mirror Clones can be created. So you can, for example, simplify backups, or very easy to generate test data and prepare migrations. Yet 2011 will be possible online RAID migration.

If necessary, EVA systems can be mirrored using HP StorageWorks Continuous Access EVA. A mirror is synchronous or asynchronous possible.

HP StorageWorks EVA Dynamic Capacity Management software simplifies the management of storage space and helps make better use of a storage system. This product is among other things Thin provisioning possible.

The long-awaited "right" Thinprovisioning came in the summer of 2011 to do so.

HP StorageWorks Secure Path for HP- UX and multi -path failover software ensure an optimized path management, provided that the operating system can not handle out of the house with redundant paths.


About the EVA iSCSI Connectivity Option Vdisks can be exported via iSCSI. Here, an iSCSI -to -FC bridge in the EVA is included, which can be administered via the Command View EVA. Thus, it is possible to provide high-performance storage and inexpensively via iSCSI. For larger file server environments, there are EVA File Services, which is a solution of EVA storage systems ProLiant servers and HP PolyServe file serving. Here, a composite of ProLiant servers works, running Windows or Linux, as a virtual file server. The file server can be accessed via CIFS, NFS or PolyServe Direct I / O. At two EVA systems to reflect over long distances, there is the IP Distance Gateway. For longer distances with low-cost IP routes can be bridged.

In total, around 200 petabytes in the form of EVA storage systems are currently installed.