over PCIe SSD?
PCIe SSDs have recently gained notoriety largely
based on their high performance and capacities.
While these are attractive, there are system-level
compromises to be considered.
First, server manufacturers are driving towards
smaller and smaller footprints. In a 1U server, for
example, only one or two PCIe slots may be available.
Occupying one of these valuable resources with a PCIe
SSD may not be the best choice in maximizing overall
system performance.
Second, the PCIe SSD’s impressive computing
capability comes at the cost of high power
consumption and the domination of server CPU
cycles. A typical PCIe SSD may consume well over
the specified 25 watts of power per slot. Additionally,
these SSDs can consume more than 50% of the
server’s CPU cycles, thus robbing computing power
from virtual applications.
In terms of flexibility, PCIe SSD’s are application and
system specific, requiring drivers for each operating
system and platform. And unlike native PCIe SSD,
a “bridged” PCIe solution with multiple SSDs behind
a RAID chip does not wear level host writes over the
entire array, thus allowing some of its SSDs to wear out
Lastly, employing a PCIe SSD is an all-or-nothing
proposition. A lack of scalability forces the designer
to buy at the high end of the price, performance, and
capacity spectrum. There is no ability to incrementally
increase storage resources. And if a portion of the
card fails, the entire capacity is lost and must be
Viking Technology’s SATADIMM offers the solution