Contents |
Violin & HP Partnership
Violin Memory and HP formed a strong bond in competing against Oracle’s Exadata by combining HP servers connected via PCIe to Violin’s 3000 flash-based products. These solutions outperformed Oracle’s Exadata on cost, response-time and throughput. There was discussion of a closer partnership between the two companies. However, recent statements from Violin as part of its preparation for an IPO change the outlook for any potential partnership between Violin’s latest 6000 products and HP future flash initiatives:
- “HP has stated 3Par is the long term strategic direction for their company;"
- "The current HP-Violin relationship remains unchanged. The VMA product family (the Violin 3000 and vSHARE software) continue to be available to customers via HP as per the [current] announced relationship.”
Since July 2012, HP has been clearly saying "HP 3PAR is our strategic direction for solid state storage".
HP Storage Strategy
HP’s strategic storage emphasis is to provide a software-led storage infrastructure. It envisions a set of software storage services that provide an end-to-end storage management across the enterprise. HP has aggressively grown market share for its $2.35 billion purchase of 3PAR, and as a result 3PAR’s 64-bit InForm storage operating system has been chosen by HP as the strategic software base to support this strategy.
Wikibon has pointed out that almost all active data will reside on flash technology by 2015, with non-active data residing on traditional high capacity hard disk drives. Original writes, very high performance data and metadata will be held at the dense server complex computing level. New IO protocols (e.g., Fusion-io's atomic write API) will allow very low latency access to flash storage and low latency access across the server complex. Flash only arrays will allow data sharing between server complexes across data centers and metro distances. The lowest level of data will be asynchronously written to magnetic drives and tapes and managed by higher-level metadata. The key (and very complex) management layers between manage data migration between the physical layers, manage data consistency and integrity, enable end-to-end data reduction and security (compression, de-duplication and encryption), manage replication and recovery, and ensure compliance.
The current flash strategy on 3PAR storage arrays has been limited to using Solid State Drives (SSDs) in place of the traditional Hard Disk Drives (HDDs) in the same drive bays within the storage array. HP even offers a 100% SSD 3PAR offering. This approach is a low cost way of implementing flash. However, the performance of SSDs is constrained by the traditional array infrastructure, which was designed many years ago for slow disk drives with low data transfer rates. As previously stated by Wikibon, putting solid-state drives into a traditional array architecture built for slow mechanical hard drives is similar to putting a Tesla engine into a Fiat Uno; it fits in the engine space, but the performance constraints of the body and transmission will kill any performance gains.
If HP is to use the 3PAR InForm operating system, 8-controller architecture and ASICs and achieve high performances flash-only storage arrays, HP will have to ensure that the flash controllers and communication infrastructure are developed from scratch.
Flash-only Arrays
The market for flash-only arrays has been a magnet for investor money, with a long and growing list of vendors such as Astute Networks, Fusion-io, Kaminario (using Fusion-io technology), NexGen Storage (hybrid flash/HDD), Nimble (hybrid flash/HDD), Nimbus Data Systems, Pure Storage, Skyera, SolidFire, Starboard Storage (hybrid flash/HDD), Tintri (hybrid flash/HDD), Texas Memory Systems (bought by IBM), Virident, Violin Memory, WHIPTAIL and XtremIO (bought by EMC).
HP’s storage competitors EMC and IBM have acquired flash-only array technologies, and NetApp is developing a closer relationship with Fusion-io. All the storage vendors are betting that they can adapt these technologies to fit in with their storage operating systems functions and management frameworks faster than they can develop flash technologies based on their own technologies.
Conclusions
A sensible conclusion reached by Wikibon and other observers from the Violin Memory disclosures and HP statements is that HP is developing its own flash-only array, based on the 3PAR InForm operating system, and using the 3PAR 64-bit 8-way processor complex to power the system. HP will be betting it will be faster to develop its own flash technology solutions, and fit them into its software-led storage strategy. Of particular importance will be the ability to manage large amounts of metadata and move data automatically between federated high performance flash-only arrays and high capacity storage arrays.
Emerging Non-Volatile Memories
There are many emerging non-volatile memory (ENVM) contenders for replacing NAND flash memory. HP will also be hoping that its next generation Resisitive Random Access Memory (RRAM or ReRAM) based on Titanium Dioxide (called memristor by HP) will be a replacement for flash in the next decade. Developing experience of designing and bringing to market native flash storage arrays will be useful in reducing time to market for ENVM technologies, if and when they emerge.
Action Item: CTOs and CIOs should be deeply interested in software-defined infrastructure and data-led application architectures. Within this framework, software-led storage architectures and end-to-end storage services will play a prominent part in connecting active data, backup data and archive data across the enterprise. The flash-only arrays will be a critical component.
Footnotes: