Host-side software (VIB) offloads reads and accelerates writes. Performs all dynamic performance management. Provides VM-centric administration—no storage artifacts to manage.
Designed for strong isolation to scale mixed-use environments. Provides locality of faults and performance.
Performs inline dedupe and compression to local flash, offloads instant vDisk clones, and performs RAID-parity calculations for persistent data stored on the NetShelf.
An HA server appliance to centrally store persistent data. Dual active/standby servers with mirrored NVRAM for writes. Disk drives for long-term retention of durable data.
All IO performed by DVX Software on VMware hosts. Data on the DVX NetShelf is organized for streaming host-flash uploads.
Customers install a NetShelf, then upload DVX Hyperdriver software onto vSphere; hosts to do the rest.
I/O speed resources are server-based so the DVX gets faster as you add servers–up to 32 servers/3M IOPS.
Server-based instance flash avoids SAN latency (more cores = more VMs = higher queue depths).
With high performance density, all application data and most IO remains server-local.
Insane Mode delivers in-place VM acceleration with the touch of a button; allocates up to 40% of underutilized CPU cores to IO processing.
Alternatively if local CPU use is high, use vMotion to move the VM to a less busy server to increase IO performance.
Administration & troubleshooting are simple, cohesive, and end-to-end through the vCenter plugin.
Respond to unpredictable workloads; dynamically change the number of hosts, cores/host or SSDs/host.
Server configurations need not be the same nor remain static within a cluster.
RAID calculations and IO are performed locally but stored centrally, eliminating host-to-host IO overhead.
Hosts remain serviceable without lengthy rebuilds; all can fail at the same time without data risk.
Leverages under-utilized compute within your existing ”brownfield” server infrastructure.
Supports commodity flash–including inexpensive read intensive drives–typically 1/8th the cost of the same drive in an All-Flash Array.
Performs inline deduplication and compression prior to writing to flash.
Global dedupe occurs on shared persistent data throughout the day for further efficiencies.
Deduplication and compression are always-on for both server-flash and durable data.
Presents a single addressable capacity pool of up to 32 hosts.
A single persistent data server provides 12 x 4TB drives with approximately 100 to 180TB of effective capacity.
The pool appears as an NFS datastore to vSphere, but terminates within each server. Data moves to the persistent data server over a proprietary protocol.