Skip navigation
Help

non-volatile RAM

warning: Creating default object from empty value in /var/www/vhosts/sayforward.com/subdomains/recorder/httpdocs/modules/taxonomy/taxonomy.pages.inc on line 33.
Original author: 
Sean Gallagher


The ArxCis-NV DIMM combines DDR3 dynamic memory with a flash memory backup.

Viking Technology

The server world still waits for DDR4, the next generation of dynamic memory, to be ready for prime time. In the meantime, a new set of memory boards from Viking is looking to squeeze more performance out of servers not by providing faster memory, but by making it safer to keep more in memory and less on disk or SSD. Viking Technology has begun supplying dual in-line memory modules that combine DDR3 dynamic memory with NAND flash memory to create non-volatile RAM for servers and storage arrays—modules that don't lose their memory when the systems they're in lose power or shut down.

The ArxCis-NV DIMM, which Viking demonstrated at the Storage Networking Industry Association's SNW Spring conference in Orlando this week, plugs into standard DIMM memory slots in servers and RAID controller cards.  Viking isn't the only player in the non-volatile DIMM game—Micron Technology and AgigA Tech announced their own NVDIMM effort in November—but they're first to market. The modules shipping now to a select group of server manufacturers have 4GB of dynamic RAM and 8GB of NAND memory. Modules with double those figures are planned for later in the year, and modules with 16GB of DRAM and 32GB of NAND are in the works for next year.

The ArxCis can be plugged into existing servers and RAID controllers today as a substitute for battery backed-up (BBU) memory modules. They are even equipped with batteries to power a last-gasp write to NAND memory in the event of a power outage. But the ArxCis is more than a better backup in the event of system failure. Viking's non-volatile DIMMs are primarily aimed at big in-memory computing tasks, such as high-speed in-memory transactional database systems and indices such as those used in search engines and other "hyper-scale" computing applications.  Facebook's "Unicorn" search engine system, for example, keeps massive indices in memory to allow for real-time response to user queries, as does the "type-ahead" feature in Google's search.

Read 2 remaining paragraphs | Comments

0
Your rating: None