In what has become somewhat of a tradition at HP Discover, the HP 3PAR StoreServ platforms were extended this week with the announcements of new products and features, including the massively scalable 20800 series. HP released two new models:
- 20800 “Converged Flash” – a hybrid system, scaling up to eight nodes, 33.8TB of cache, 1920 HDDs and 1024 SSDs and 6PB raw capacity.
- 20850 “All-Flash” – an all-flash system, scaling to eight nodes, 3.6TB of cache, 1024 SSDs and 4PB of capacity.
I suspect the delivery of a specific all-flash model is to ensure the 3PAR systems continue to sit on Gartner’s all-flash magic quadrant. These systems are truly huge, with more flash in a single system than any other vendor (as far as I am aware) and connectivity from up to 160x 16Gb/s Fibre Channel ports.
The 20800 products are based on a new Gen5 “Thin Express” ASIC that includes support for SSD and implements improved (inline) data de-duplication and T10-PI ( also known as T10-DIF end-to-end checksum validation from HBA to array drives) for protection against data corruption issues. HP are calling this feature Persistence Checksum. With T10-DIF supported host HBAs, checksums are added at the host layer; with unsupported HBAs, the 3PAR system adds the checksum at the array front-end port and validates the data to the back-end disks/SSDs.
HP also announced improvements to data replication (known as HP 3PAR Remote Copy) that implements asynchronous streaming mode. Streamed replication provides a more granular level of RPO, compared to periodic replication that uses the shipping of snapshots to update the remote (or secondary) array. This means that HP 3PAR systems now support synchronous, asynchronous (snapshot) and asynchronous (streamed) modes, providing flexibility in implementing data protection based on application/business requirements. Incidentally, I expect there will be discussions on why synchronous replication is offered for flash-based systems (expect that comment from those who don’t implement it). It is true that synchronous replication introduces latency, but only for write I/O; read I/O is unaffected.
There may be good reasons why (especially in Metro or 3DC configurations) where synchronous replication is acceptable to provide a sufficient level of application resilience. It’s also worth mentioning that replication can be to/from 20800 series and the existing 7000 series (e.g. midrange to/from enterprise).
HP has been “all-in” on the 3PAR systems since acquisition, primarily I expect because of the focus placed on storage by David Scott, ex-HP Storage VP, who left the company earlier this year. I’ve yet to meet Manish Goel (I’m not attending HP Discover US this year), so it will be interesting to see how the focus changes and whether 3PAR continues to be the core product in HP’s storage portfolio now that Scott has departed. There’s no doubt that the 3PAR range is fully featured, however there is a risk that there are now so many features on the platform, customers may be unaware or confused as to exact what each one is used for.
For more information I recommend looking at the documents on this page – HP 3PAR StorServ 20000 Storage – Product Documentation. Despite a nice revamp, HP’s website(s) are still awkward to navigate, so here are two documents that are particularly useful:
Comments are always welcome; please read our Comments Policy first. If you have any related links of interest, please feel free to add them as a comment for consideration.
Disclaimer: HP is a client of Langton Blue Ltd
Copyright (c) 2009-2015 – Chris M Evans, first published on http://blog.architecting.it, do not reproduce without permission.