In-Memory compute vendor Formulus Black rethinks market strategy with Forsa 3.0 release

Formulus Black, which lets memory be used as high-performance media, now will run on Intel Optane as well as DRAM, and will now support CentOS as well as Ubuntu, with RHEL support coming later this year.

Jing Xie, COO, Formulus Black

Jersey City-based startup Formulus Black has launched the 3.0 version of their Forsa software stack. While it makes some major enhancements, including new support for Intel Optane DC Persistent Memory that they believe will have major significance, the biggest is a fundamental reboot of their approach to the market. While the product was originally launched as ForsaOS, and was based on Ubuntu, with 3.0, Forsa has been decoupled from Ubuntu, with support for CentOS, and forthcoming support for Red Hat Enterprise Linux [RHEL]. It thus takes a broader approach to the market, becoming more like middleware, in order to expand their customer base beyond Ubuntu users.

Formulus Black has made major corrections to its strategy before. The company was originally Symbolic IO, which developed a technology that made DRAM non-volatile, inexpensive and usable for storage. However, while Symbolic IO was a hardware appliance, Formulus Black was reworked into a software-designed approach, which was much less capital intensive, and allowed it to work off any modern Intel platform. They rethought the Go-to-Market approach at the same time, moving from a direct sale model to a channel model.

The core problem they are trying to solve remains unchanged. While DRAM is super-fast, it is also super-expensive, and super-volatile, major defects for customers looking for a stable environment. Formulus Black lets memory be easily and efficiently used as a high-performance storage media with a patented data encoding technology called Formulus Black Bit Markers, patented algorithms that process application codes to find and remove repetitive bit patterns.  Real-time pattern matching reduces redundancies and makes storing data in-memory more efficient by a factor of 2-24x. This removes capacity issues – and dramatically cuts the price.

Another major innovation is making DRAM persistent, to eliminate the nightmare scenario that if you lose power, you lose all the data that was stored in DRAM. Formulus Black addresses this by taking system memory from DRAM in the event of a power failure, and migrating it to external SSDs. This wasn’t possible before, because the drives were too slow and you would run out of power before the job could be completely copied. Removing the capacity issues makes it possible because it treats the whole system like a big NVDIMM.

Originally, this was all based around a modification of the Ubuntu kernel, tuned so that the hypervisor could see system memory as either memory or storage. Increasingly, however, that Ubuntu lock-in became perceived as more of a problem than a solution.

“We had a chance to validate the technology on multiple applications, as well as on custom applications that customers had developed,” said Jing Xie, Formulus Black’s Chief Operating Officer. That led to a decision to fundamentally architect the product to decouple Forsa from the Ubuntu OS.

“That’s why with 3.0, we are just Forsa, not ForsaOS,” Xie noted. “Previously, you HAD to use a Ubuntu OS. Now we have expanded the support to include CentOS and are working on adding RHEL support.” The RHEL support is planned for the late fall timeframe this year.

“With the broader OS support, we are basically a middleware layer,” Xie noted. “We kept the same value proposition, a very fast storage media for applications to run on. But we thought that this was a simpler and more effective way to present the solution to a customer. A lot bigger customer base can now consider the technology, not just Ubuntu users. It brings more piece of mind to a typical systems admin, because people didn’t want a proprietary system from a small startup.”

Another huge change is that while in 2.0 DRAM could be be provisioned as storage media, Forsa 3.0 adds Intel Optane DC Persistent Memory Support.

“With 3.0, we are extending capacity in terms of system types and system sizes,” Xie said. “We added support for the new Intel Cascade Lake family of CPUs, and with it, support for Intel Optane. So in addition to running on DRAM, Forsa now runs on Intel Optane persistent  memory. It’s not as fast as DRAM, but it has higher capacity and lower cost – and is something that some customers will view with more confidence.”

That reassurance that using Optane will provide customers is extremely significant, Xie stressed, even though Formulus Black had addressed that nightmare scenario created by DRAM’s volatility.

“We believe that the Optane support will be extremely significant,” he said. “Companies have been afraid to put mission critical in-memory, because of that .0000001 per cent chance that the data centre running the server could go dark. While we did address that, Optane removes that tradeoff, and it’s cheaper than DRAM. The admin can go home at night and sleep soundly, because nothing changes except that with us it will run on persistent storage in the memory channel itself, and we make that experience of using persistent memory very easy.”

As a result, Formulus Black and Intel are collaborating closely.

“We are now doing calls with the Intel Optane memory division every two weeks, and are doing joint testing together,” Xie indicated. “They are very impressed with the performance they are getting. Instead of creating overhead for them, we improve application performance and are negative overhead for them, because of our ability to pool physical Optane memory so that it’s not tied to any socket.”

Forsa 3.0 also makes some major capacity and memory enhancements over 2.0

“We can now support systems up to 64 CPUs through the pooled memory across sockets, up from two-socket server support in 2.0,” Xie said. “We can also now scale out a LEM to extend beyond the capacity of a single node, and can scale horizontally as well as vertically.”

The management changes include the ability to BLINK specific VMs and LEMs.

“Users now can BLINK much more selectively,” Xie stated. “Before, it used to have to be the whole system, We have also added an experimental feature, a two-node HA LEM. Before there was no LEM replication between nodes.” Docker container support has been added, and the GUI has been enhanced to permit multi-node management as well as single node.