Venture-backed startup
Formulus Black today announced that it has teamed with leading IT
consulting and engineering firm Nth Generation to deploy ForsaOS into
the marketplace, offering those in the hyper-converged infrastructure,
high performance computing and public cloud compute industries with the
ability to run all applications entirely in-memory to provide huge gains
in compute efficiency.
Ideally suited to solving the needs of enterprises in the financial
services, automotive, telecom, energy, university and healthcare
industries, ForsaOS will be integrated by Nth Generation as part of a
series of HPE-branded servers designed to best meet their customers'
needs. Testing of ForsaOS on a HPE DL380 Gen10 server featuring Intel
Xeon Gold 6126 processors, and 24-32GB HPE DDR4-2666 MT/s DIMMs yielded
performance from 4.5 to 7.8 million random 4K IOPs with average
latencies for writes at 2.8µs and reads at 1.8µs per IOP with CPU
utilization below 50 percent - a level of performance unmatchable by any
SSD or other I/O-bound technology. The results were achieved running
VDBENCH on ForsaOS in host mode (non-virtualized).
"While we understood that by ForsaOS running all applications in memory
we would achieve significant gains in compute efficiency, we didn't
expect to see such incredible performance results in terms of execution
speed and reduced latency that we saw in testing," said Lee Solomonson,
Nth Generation Labs Manager. "We see ForsaOS as a transformative
technology and look forward to offering it in our product portfolio so
that we may better help our customers achieve their business goals now
and into the future."
Developed as a software only solution that utilizes fast DRAM memory as
storage while providing all the necessary management tools and features
needed to increase effective memory capacity by up to 24x while
improving processing speed as much as 450x, ForsaOS enables larger
database and analytics workloads to persist and run in memory on
commodity hardware without any application modifications. ForsaOS
organizes data for use with any application and OS for a truly
platform-agnostic approach. A built-in hypervisor leverages the
company's patented Formulus Bit Markers (FbMs) technology to allow many
more virtual server instances on standard hardware than conventional
hypervisors.
"While storage vendors look to add hyperconverged solutions to move
closer to compute, and compute vendors add additional peripheral
channels to servers in an attempt to address storage I/O bottlenecks,
Formulus Black enables a software-only solution without the negative
technical tradeoffs made by other vendors," said George Crump, Lead
Analyst of Storage Switzerland. "Others who have attempted a similar
converged approach have struggled to overcome the technology hurdles
that allow both memory and storage to coexist in the memory channel."
Formulus Black's ForsaOS exponentially increases cost effectiveness,
data security, processing speed and memory capacity without application
changes, compression, encryption or added peripherals. The software
stack is the first in the world to keep data in persistent memory and is
completely safe against power loss. By identifying and encoding data
patterns using algorithms, the software enables more data to be securely
persisted in memory while also providing complete in-memory system
backup and restore in minutes instead of hours or days.
"We are pleased to have the opportunity to bring ForsaOS to market by
partnering with leading solutions providers like Nth Generation who will
be able to deliver tailor-made solutions that satisfy the needs of
customers requiring exceptional performance, even when utilizing
more-affordable, mid-tier Xeon processors" said Wayne Rickard, Chief
Strategy and Marketing Officer. "Because we designed the ForsaOS
software stack to run any workload in memory without modification on any
server, reseller partners and integrators can quickly and easily
incorporate it into their product portfolios to deliver solutions that
break new ground in satisfying performance-hungry workloads at
memory-channel speeds."