Was wondering when that would get born. They both use the same PR company, so an alliance was pretty much inevitable. Solid crossover, makes sense to do. Now to go convince them that one of these need a couple week's worth of testing by El Reg to prove it out...
Tegile to shove flash into DIMMs, it would seem
Tegile's marketing veep figures that "PCIe flash and NVDIMMs will make their way into shared storage devices, further driving latencies down." Cue Tegile supporting flash DIMMs in 2015, then. NVDIMMs (or flash DIMMs) are DIMM memory sockets filled with NAND dies so that the flash is a read extension of the memory's address …
-
-
Friday 19th December 2014 07:41 GMT Anonymous Coward
Re: Sure NVDIMM is new, but PCI Flash?
This is true but the problem with WAFL/OnTap is that it wasnt built with flash in mind so that whilst moderate improvements were made with flashcache, it wasnt the blistering improvement which comes from newer systems which were designed to use this magic black stuff more effectively.
-
-
Thursday 18th December 2014 15:31 GMT Lynrd
NVDIMM- a good idea!
Nimble uses NVDIMM for the NV Layer in all the latest Generation product, which has been shipping for several months now. Write latency was reduced with the change, but even the NVRAM card on the PCI bus that the previous generation used (very similar to the NTAP architecture mentioned above) delivered sub millisecond latency.
The switch to NVDIMM frees up a PCI slot that can be used to add additional front end ports and also so far has been easily as rock-solid as the legacy PCI Flash solution it replaced.
-
This post has been deleted by its author
-
Friday 19th December 2014 18:42 GMT Alistair
slightly confused.....
We go with NVram on DIMMs in the storage controller at the head of the array -
--- so we can buffer writes at the controller and reduce write latency to the storage subscriber, without loosing coherence or the write itself (given a dead moment in the datacenter).
With all flash arrays - are we really improving writes? - or are we talking about putting this in front of spinning rust arrays?
Or is the objective to increase the overall performance of the storage controller, expanding its capabilities to the subscribing systems?
Biting the hand that feeds IT
