Correction:This story originally indicated that the HP EVA4400 SAN system required a header server running Windows, this is incorrect.
The HP EVA4400 is a Storage Area Network (SAN) system that provides data storage and security on a much larger scale than a simple RAID or NAS system. We found it to be a reliable and scalable high-end data system.
We see several key reasons for considering a SAN system. Administrators often do not want to separately access several physical machines in order to perform system backups and they don't want a battle to locate faulty drives.
Scaling up your storage should not mean having to predict how much space you are likely to need in three years time, carefully selecting which machine to add the extra drives to, and then specifying the OS they need to be formatted for.
You should be able to feel secure that data backups and drive failure recovery will proceed without trauma or downtime. Such requirements are certainly beyond the capacity of simple RAID systems or even NAS devices. However, all of the above can be done with a SAN system.
Design and Features
The EVA4400 is actually a collection of units supplied in its own full-sized rack. With a few options on how you would like it delivered, you can have it sent to your premises pre-wired so all you have to do is plug in the power and connect it to your existing network, you can have the option of configuring the system yourself, or having HP partners do it for you. Physical drives are contained in drawers which bear up to 12 volumes each in a 2RU space. Up to eight drawers allow for as many as 96 individual physical drives. The unit we looked at had just a single drawer containing eight 146GB drives. Power and data connectors were all duplicated to ensure reliability.
While the storage organisation has some similarities to traditional RAID configurations, there are some fundamental differences to the implementation which vastly improve performance and reliability. The EVA agglomerates all the available physical drives into a single virtual unit where logical drives (in fact, logical RAIDs which are striped across the full set of physical devices) are allocated from this space. Connected servers are then applied as hosts to these virtual drives. A single virtual drive can be "hosted" by multiple servers. Furthermore, each virtual drive on one SAN array may be formatted for a different operating system.
With a standard disk, data must be accessed one byte at a time. With RAID, the number of bytes accessed simultaneously can equal the number of physical drives in the array. In the case of the EVA4400 as many as 96 physical disks may be included, thus vastly increasing data throughput. So, if you want to build a 20TB storage facility, it is far better to use 40x 500GB drives than it is to use 20x 1TB drives. Further redundancy and load balancing is achieved by using Multipath I/O. Dual-copper fibre-channel cables connect disk shelves with the SAN header and attached servers.
Drives can be added or replaced with a minimum of fuss. The EVA4400 automatically adjusts itself to the number of physical disks available to it. If a unit fails, then RAID 1 & 5 redundancy data is used to rebuild the lost drive using whatever space is available. "Spare drives" are not physical units as in traditional RAID, but are in fact just another stripe on the virtual disk. As drives are removed due to failure or added to increase available space, the SAN restripes all the logical drives to cover each drive.
Physical disks are also grouped so that data loss can only occur if two physical disks are lost from the same group at the same time. In almost all circumstances the system would compensate for any disk failure long before another failure could occur. Drive rebuilding has the potential to restrict user access to data, but the EVA4400 gives these tasks a lower priority than user requests — therefore, in effect, it does its housekeeping when the system is idle. Also, clever data arrangement means that restriping does not require rearranging all the data on the disk but only a portion equivalent to the amount of space lost or gained.
The Snapshot allows you to create a backup copy of a drive so you can safely explore software upgrades knowing that you can rapidly restore to the earlier configuration if things go pear-shaped. The copy is initially created as a pointer to the original drive with files only being copied as changes are made thus making the initial duplication process almost instantaneous. Similarly, there is a Snapclone function that, as the name suggests, creates clones of an existing logical disk.
Management software must be installed on a Windows server, but once it's done, SAN management can be managed remotely from a Web-browser based in any other OS. OS support includes Windows , Linux (Red Hat, SUSE), Vmware ESX, HP-UX, Solaris, AIX, Netware, Mac OS X and OpenVMS.
HP's disk management guides the administrator through the process of creating new logical drives and bypasses the need for messing about with Windows disk management tools. HP's iLO (Integrated Lights-Out) gives you full remote control over the hardware ensuring that you and emergency administrators can troubleshoot problems from wherever they are.
The "Command View" management interface allows you to check on the status of both logical and physical drives, and adjust logical drive properties (including determining which servers have access to them). Servers, and thus individual end users, see these logical drives as they would any other network drive. Limitations of Windows Server 2003 mean that resizing logical drives is a bit messy (we had to abandon the Web interface to do this), but still not troublesome; however, Server 2008 should allow simpler drive resizing. If you need more than one SAN unit — whether for geographically separated redundancy or extra space — these can all be administered from the same Web-interface.
HP representatives will do all the initial installation and provide a briefing on equipment usage. HP can even migrate your existing data to the new system — for an AU$3,000 fee (which may be okay if you currently have a chaotic collection of data storage devices). The user interface is relatively simple and should not result in any hair-loss to your system administrator.
The HP EVA4400 provides extremely high data availability by facilitating much higher I/O rates and eliminating downtime due to maintenance. The EVA also simplifies data management by physically and logically centralising all your network storage systems.
A starter pack complete with 8x 146GB drives is priced at AU$20,999 (including GST). This price also includes the full installation process and a two-hour orientation session with HP staff. This pricing makes the system accessible to medium-sized businesses — even relatively small businesses that are heavily reliant on highly accessible and well protected data may find this an attractive deal. The warranty period is normally two years, but in Australia the warranty period has been increased to three years.