What Makes Lightbits the Next Evolution Beyond Ceph

Insights learned in our webinar comparing Lightbits to Ceph Storage 

Our webinar last week,Lightbits: The Next Evolution Beyond Ceph, generated strong interest, making it clear that IT folks in a broad range of industries are looking at Lightbits as a Ceph alternative for block storage. The webinar ended with a lively Q&A discussion that we’re sharing with you in this blog, along with a few insights we got from our audience with our poll questions.

 

Which storage architectures were attendees looking to replace?

Given the webinar’s topic, the attendees were currently investigating software-defined storage solutions to eliminate issues with their current storage. Based on their poll responses, we were a little surprised that 60% reported they were already using a software-defined storage solution today. Perhaps Ceph? Twenty percent were looking to replace a SAN, and 10% were looking to replace DAS. That tracks with our current customers who have turned to Lightbits from all of these architectures, for example:

  • AI Cloud provider Crusoe switched from Ceph storage to Lightbits for their GPU-based services after extensive benchmarking. Lightbits gave them external, persistent block storage with performance that rivals local flash with greater scalability, plus built-in snapshotting.
  • An online retail giant in Asia deployed cloud storage for ecommerce from Lightbits to replace a DAS architecture for their real-time data platform after determining that Ceph didn’t have the scalable performance they needed for their massive customer transactions and intense analytics.
  • A Fortune 500 financial services company purchased Lightbits as part of modernizing their data centers to replace SAN storage. They chose Lightbits over Ceph because it offers the superior performance and data services their workloads require.
Lightbits ceph storage alternative performance comparison chart

Figure 1: Crusoe benchmarks showed Lightbits maintained significantly high bandwidth and lower latency than Ceph.

 

For more detail on performance benchmarks, including results from the Evaluator Group, watch the webinar on-demand or download the white paper.

 

NVMe/TCP is top-of-mind for IT shops investigating software-defined storage

The audience polls also showed strong interest in software-defined storage with NVMe/TCP support, with 50% either using NVMe/TCP today or having tested it in their data center. The webinar detailed how the NVMe/TCP implementation matters, and how a gateway approach such as Ceph’s can limit performance compared to the Lightbits architecture, which is built from the ground up for NVMe/TCP. After all, we invented the NVMe/TCP storage protocol.

Lightbits versus Ceph NVME storage topology

Figure 2: Unlike Ceph, Lightbits extends NVMe’s parallelism across the network without a gateway bottleneck.

 

Questions ran the gamut from networking to operations to hardware impacts

 

Here’s the expanded Q&A with Hemanth Kommuru, our director of product management.

Q: What is the difference between NVME over TCP and NVME over RDMA?

  1. NVMe/TCP doesn’t require any special hardware such as switches, or special configurations set for that hardware. NVMe/TCP is similar to iSCSI in its simplicity but delivers higher performance.

 

Q: For the Lightbits architecture, what kind of network would sufficiently support the NVMe/TCP speed?

  1. Lightbits uses standard Ethernet and TCP/IP, so it will work with whatever you already have. Of course, a network can be a bottleneck. If you have bigger, beefier switches that’s all you need.

 

Q: It looks to me that the advantage of Lightbits is the NVME/TCP protocol. Is that what makes Lightbits unique?

  1. The difference is how Ceph implements NVMe/TCP. With Lightbits, you don’t need any additional networking infrastructure. In contrast, Ceph uses a gateway, which in periods of high demand could introduce bottlenecks.

 

Q: You said Lightbits works on commodity hardware. How does it handle a mix of drives–NVMe, SSDs, and HDDs?

  1. The Lightbits architecture is built on NVMe, using NVMe drives and NVMe/TCP protocol which we invented. So we don’t support hard disk drives, just NVMe SSDs. But Lightbits can run on any commodity hardware server, so you can use your choice of server vendor with your choice of NVMe SSDs.

 

Q: Was the testing done with Ceph’s replicated rule?

  1. Three copies of data were used in the benchmark testing, both for Ceph and for Lightbits.

 

Q: How much downtime is required when adding nodes for scale?

  1. The advantage of Lightbits is being software-defined. The architecture is straightforward; you can add drives to scale up or add nodes to scale out. All of that is handled intelligently by the software so there’s zero downtime.

 

Q: What’s the upgrade process for Lightbits?

  1. Lightbits clusters auto-upgrade in a rolling fashion. Each node is upgraded one at a time with no downtime. Lightbits customers report they can successfully upgrade during normal business hours.

 

Q: How does Lightbits compare to DAOS storage servers?

  1. Lightbits and DAOS are designed for different storage needs and use cases. Lightbits is a high-performance disaggregated block storage solution, ideal for applications requiring fast access to raw storage blocks, such as databases and virtual machines. In contrast, DAOS (Distributed Asynchronous Object Storage) is an open-source, software-defined object store optimized for massively distributed environments, making it suitable for high-performance computing, large-scale data analytics, and parallel file systems. Therefore, they do not compete directly as they serve distinct purposes in the storage ecosystem

 

Q: Are you expecting any particular advances on the compute end of technologies (AMD/Intel chipsets and connectivity) to further optimize your NVMe/IP-based storage architecture?

  1. As a software-defined storage solution, our focus is on delivering exceptional performance on any hardware platform. Lightbits does not depend on hardware acceleration but is built to scale with the increased number of cores in newer processors, such as the newer Intel SPR processors that include embedded acceleration. We expect these new processors will improve our performance, as will the new DDR memory, bigger CPU caches, and faster PCIe (Gen4/Gen5) connectivity. Lightbits software performance generally improves hand-in-hand with hardware performance advancements.

 

Q: How does the Lightbits license work?

  1. Lightbits licensing is straightforward. It’s based on provisioned capacity, regardless of the number of nodes in the cluster. It’s a hybrid license too. The same license can run on-premises or in the public cloud. You’re charged on overall provisioned capacity.

 

Q: Is the license perpetual or a subscription?

Lightbits offers term-based licensing. You can purchase a one, three, or five-year license.

 

Do you have more questions or want to chat about how Lightbits would fit in your environment? Request a demo with Lightbits today.

 

Additional resources:

About the Writers: