- Product
- Solutions
Deployment TypeEnvironmentIndustriesApplications WorkloadsPartners
- Learn More
CompanyMenuResourcesMenuToolsMenu
- Support
- User Community Hub
Lightbits Labs Shift to Scale-Out Flash Storage: G2M Research Interviews Josh Goldenhar
This month’s interview was with Josh Goldenhar, Vice-President of Product Marketing at Lightbits Labs (you can hear the interview in its entirety here). Lightbits is well-known as one of the “champions” of NVMe™ over TCP (NVMe™/TCP)- Lightbits helped draft the specifications and they developed one of the first software drivers. However, Lightbits Labs real focus is on Scale-Out Flash Storage (SOFS) software.
Josh, when I first heard about Lightbits Labs, it was about your NVMe over TCP driver. Now you seem focused on what we call “scale-out flash storage” or SOFS, software solutions. Is this a pivot for Lightbits, or was this your planned approach from the beginning?
We certainly became known as the “NVMe over TCP” company, but scale-out was always the plan. Our founders have backgrounds in flash storage controllers and storage networking, so we have a lot of expertise in storage and scale-out architectures. Our focus on NVMe/TCP was to make adoption of our technology “frictionless” by universalizing our solution.
Your website talks about “cloud-scale data centers”, and the value of bringing hyperscale storage architecture advantages to “normal” datacenter operators. What do you see as the major benefits of hyperscale storage architectures for normal datacenters?
There are three things that enterprise datacenters can get with our solution: efficiencies, lower costs, and higher performance. We allow enterprise datacenters to normalize their server configurations and move from proprietary storage solutions to standard server-based hardware solutions. From a performance standpoint, we can provide 10X-20X more than storage arrays. Lower cost and higher performance – It’s like having your cake and eating it too!
If you cold talk to companies who were deciding between putting their data on a cloud service provider’s storage or keeping their existing datacenter and utilizing your software solution, what would you tell them are the reasons to choose Lightbits?
There is a place for public clouds, private clouds, and hybrid clouds – it really depends on the characteristics and the logistics of the data. In some cases, there is too much data to move to the cloud, or the latency of moving the data to the cloud and processing it can be too great. There is also the issue of cost – there are low-latency cloud solutions, but these are very expensive. Data privacy and locality issues can also be a driver to utilize private clouds over public clouds.
Lightbits Labs is almost synonymous with NVMe over TCP. Do you think you miss any significant parts of the market by not supporting other NVMe over Fabric protocols, or do you see NVMe over TCP eventually becoming the dominant NVMe over Fabric protocol?
The real question is what “significant” means. There are definitely some applications in FinTech and HPC that need extremely low latency. There are also legacy applications that utilize legacy transports like Fibre Channel. However, we think that these use cases constitute around 5% of the market, so we don’t think we are missing significant parts of the market. Additionally, Intel has come up with a new architecture called Application Delivery Queues (ADQ) that significantly reduces TCP latency, and may open up a good part of that 5% to TCP.
What impact has COVID-19 had on Lightbits Labs business?
We expected some of the impacts, such as the cancellation of tradeshows and marketing becoming virtual. Other impacts were less expected – installations and configuring IT equipment, for instance. Even in colocation facilities, time has to now be scheduled, and the windows are pretty constrained.
What are the next big challenges for Lightbits Labs as you continue to scale your company in the future?
The biggest challenge is adoption of NVMe/TCP, which we think will happen in the next year or so. the latest Linux distributions already include NVMe/TCP – it just takes time for enterprises to switch over.
Originally published by Mike Huemann, G2M Research’s Monthly Newsletter, April 2020.