What are you looking for ?
Infinidat
Articles_top

Datrium Doubles Per-Server Flash Performance Density to 100TB

Predictable latency enabled by server IO isolation architecture sets new bar for mixed-use private cloud infrastructure

Datrium, Inc. announced that it has doubled performance density to 100TB (1) of effective flash capacity per server in the  DVX, while improving flash affordability to 10TB per $1,000 (2).

DatriumDVX_1
This announcement illustrates how the company enables a broader range of data-intensive applications to remain entirely in local server flash, keeping all read IO processing off the network for lower latency and faster results in a shared private cloud infrastructure.

Unlike arrays or hyperconverged infrastructure (HCI), DVX-based infrastructure enables VM IO to be accelerated on-demand, and can provision speed independent of capacity. To achieve this, DVX solves the converged infrastructure problem a new way, keeping IO services and flash for read IO strictly local per server, and separating shared durable storage to a simple, off-host repository.

As a result, DVX can support a spectrum of server configuration types and flash densities. Since servers remain stateless in the Datrium approach, serviceability is a key advantage and unplanned server failures are not subject to data rebuilds of any kind.

“It’s become clear that storage will continue to move closer and closer to the server processor, where it is fastest. With an I/O architecture like Datrium, it’s not only the fastest but ends up being the cheapest,” says Steve Duplessie, founder and senior analyst, Enterprise Strategy Group. “This is the future of server-storage convergence, and marks the beginning of the end of arrays as we’ve known them.

Datrium_ProductOverview-01
Performance density keeps IO processing off network
By increasing the effective capacity of low-cost flash to keep all application data local, organizations can maintain high and predictable service levels for mixed-used private clouds. High performance density is made possible because durable data is off-host, so local flash is not used for persistence replicas. In addition, raw flash capacity is expanded by inline deduplication and compression. With this announcement for support for up to 16TBs of raw server-local flash, DVX can offer up to 100TB of effective flash capacity, keeping more IO processing local to the server and off congested networks.

I’ve been looking at the market for a long time, and this is really the first storage system that feels like it’s built for an infrastructure-as-a-service provider like us,” said Martin Skojec, director, IT, EdgeConneX, Inc.

In the future, reads will never leave the server,” said Brian Biles, CEO and co-founder, Datrium. “Server flash and other memory types are just getting too cheap and abundant, and IO queuing keeps read latency much faster on a server than across a SAN. Thanks to Datrium’s always-on data reduction technologies, our customers can experience that future today.

Performance density features are available, and are bundled at no additional cost with the DVX version 1.1 and later.

(1) With DVX Software version 1.1, 16TB of flash is supported per host. With typical data reduction ratios of 2x to 6x and no need to store replicas from other servers, users will experience effective capacity of up to 100TB.
(2) Pricing for 10-Dell 1.6TB SATA read-intensive 2.5″ SSDs at $1,021 total $10,200.  Where data reduction achieves effective capacity of 100TBs, flash costs will amount to 10 effective TB per $1,000.

Articles_bottom
AIC
ATTO
OPEN-E