Mountain View CA-headquartered startup Plexistor has announced the availability of its next-generation data center platform based on its new Software Defined Memory (SDM) architecture. It converges memory and storage technologies to deliver high capacity storage at near-memory speed for in-memory compute and data-intensive applications.
“Traditional storage was done with block-based technology, and on top of that, you put a file system,” said Sharon Azulai, Plexistor’s CEO. “We came to the conclusion that we needed a new technology that is not block oriented, where you can make changes on the storage itself. That’s our ‘storage class’ memory. It’s built from the ground up. We created a new layer that took everything out and created a direct interface from the application to the memory storage.”
The result, Azulai said, is managed completely different from traditional storage.
“We converged memory and storage together, on what can be viewed as either a very fast storage device or a very large memory device — and it’s managed by our software. My previous company was primary data, and we found it was easy to generate millions of IOPS — but that didn’t translate into performance improvement for applications, because the app didn’t know what to do with the IOPS. We realized then that latency is the issue, and we had to focus on that.”
Azulai said that this convergence of memory and storage technologies significantly improve performance in both new and traditional applications, including relational databases, in-memory databases, NoSQL, big data analytics, and complex event processing.
“We can create for the first time, extremely low latency for applications, because latency is three microseconds from the application level,” he stated. That means, for example, that SQL database latency is reduced by 24x, and that MongoDB performance improved by 450 per cent without sacrificing data persistency. In addition, Plexistor’s technology allows it to overcome the low capacity issues that limit RAM and cache products.
“Our long term vision is that 12 months from now, we will be able to offer customers 6 TB of storage class memory,” Azulai said.
Azulai indicated that they view storage class memory as compatible with the hyperconverged vendors.
“What we really will displace is the DRAM vendors,” he said. “You need less memory. That’s the bottom line. On the other hand, we can extend the potential of SSDs and hyperconverged vendors to get even further into the data center. We can fit in the same bucket as Nutanix.”
Azulai also emphasized that their go-to-market strategy by definition depends on working with other vendors.
“We have two ISV partners that we will be co-marketing and promoting the solution with, which can add much more value to their customers. We are also working with server vendors, Because we are container based, this brings more value to server vendors like Dell.”
The target market includes large enterprises, but it also could mean anything down to smaller business providing the business runs applications like MongoDB and has to deal with I/O bottleneck and latency issues. Digital advertising, log analysis, personalization, financial applications, cybersecurity, and DevOps are typical use cases.
“The market is very much app-specific,” Azulai said. “We are focusing very strongly on MongoDB, then Cassandra. Some applications will not benefit, like VDI. On the other hand, we have a customer with 5 people, who runs Mongo DB, and we allow them to have data on one machine instead of 10.”
The ability of smaller companies to access the SDM solution is facilitated by its being available on AWS out of the gate as a free community edition, as well as both on-premise and cloud-based deployments. The community edition can be downloaded here.
“We have done this free on AWS to engage the market,” Azulai said. “It also serves as a way to engage the on-prem market and test it. Will the AWS solution ever be revenue generating? At some point perhaps.”
Plexistor is also making use of the channel right out of the date. Azulai said that while they sold direct to some reference customers “because of the need to show the proof in the pudding in the beginning,” they plan a low-touch engagement strategy that relies on partners.
“We have a large customer in southern California who wanted a full solution, so they brought in an SI who did the entire engagement with the customer,” he said. “We provide the basic license and do the first line support, but we aren’t going to engage with customers with a high touch approach. It will be a low touch engagement, with ISVs, server vendors and system integrators.”