Fusion-io is about high performance Flash – aka NAND – storage, though – as Mat Young, EMEA senior director at the company points out – it would be wrong to view it as a SSD player. IntelligentTradingTechnology.com asked him to explain …
Q: Can you begin by describing what Fusion-io does? What are your products?
A: Fusion-io offers a storage memory platform that improves capabilities within data centres by moving process critical data closer to the CPU where it is processed. This architectural shift helps to reduce the time you have to wait for data, or latency as it is otherwise known, increasing the productivity of the CPU significantly. Therefore Fusion-io products essentially help you to work faster and be more productive.
Fusion-io was founded back in 2005 by a team with experience in transaction processing, performance networking, storage, superclusters and computer and image processing. The team realised many storage functions in action were unable to keep up to speed with changes in digital processing. This is why Fusion-io created its server-deployed ioMemory technology to ensure that capabilities were not halted by speeds of storage infrastructures in place.
Q: How does your ioMemory technology differ from Solid State Disks? And how does it compare performance wise?
A: Solid State Disks or SSDs are used to store data with the intention of constant use – similar to that of a hard drive. These SSDs generally use disk-based protocols that introduce unnecessary latency into the system. Fusion’s ioMemory technology differs in that it doesn’t act as a hard drive. It performs as an extension of the memory hierarchy for servers. This means that they provide a tighter integration with host systems and applications, helping you to work more productively.
Fusion-io products offer the industry’s lowest latencies, which maximise performance and scalability, while delivering enterprise reliability.
Q: Can you provide some typical I/O performance figures for ioMemory compared to DRAM, and solid state disk?
A: With some generalisation, the order of memory, fastest first is as follows, DRAM with 100-300 nanosecond access, ioMemory with 15 microsecond access, NAND appliances with around 500 microsecond access and then SSD’s with around 1ms access. There are of course a number of factors that need to be considered in these times such as payload size, load, etc, however, in simple terms with all things equal and a well-designed product, latency is ultimately affected by the distance data must travel to get to where it is useful. So, the closer your technology resides in relation to the CPU the better the response time. That’s why that even though two products may use the same NAND chips and be connected on the PCI Express bus, you see markedly different latency characteristics.
Fusion ioMemory is accessed through our unique Virtual Storage Layer software. VSL eliminates unnecessary protocols and context switches which, when combined with Fusion’s well-designed hardware platform, delivers the lowest latency NAND access in the business that is also capable of performing under a high load without radical changes in performance.
Whilst latency is not the only metric, IOPS and bandwidth are rapidly becoming less relevant, as products like the ioMemory platform are now capable of delivering higher workloads that applications can readily use. Sustained, predictable latency combined with high IOP and bandwidth capability are the key factors, in the amount of useable density that can be brought to bear in a server. On these considerations, ioMemory delivers at the combined front of these factors.
Q: What kinds of tasks and processes within a trading system might benefit from ioMemory technology? How does ioMemory help reduce latency?
A: Tasks in a general computing environment use one of three resources internal to a server (CPU, RAM, Persistent Storage) and a network to receive and send data from the compute node. With low latency trading it can be broadly broken down into two main areas: Micro-second trading and nano-second trading. That is to say the scale of response or trade completion fall inside those scales of time management.
In nano-second trading, the requirement is so demanding that it is typical to utilise FPGA technology, where the device is connected to the PCI Express bus in the server. It is only housed in the server for power and maintenance, and the FPGA device having network and process all of its own.
For micro-second trading, intelligence can be executed on the server’s own CPU complex. Depending on the specific task and requirement for retention, it may be that RAM is a sufficient storage medium and logging occurs via a different method. Where micro-second trading becomes particularly challenging is when the data needs to be persistent. In this case even the most expensive storage arrays — SSD equipped or not — struggle to deliver anything below millisecond-range response. A large element of this is purely the “distance” the payload has to travel, i.e. where that process critical data resides in relation to the CPU. By introducing ioMemory into areas that require a persistent store, it is possible to reduce the overall transaction time by a considerable amount, and, interestingly, due to the heat, power, space and management costs association with traditional storage arrays, often at cheaper cost than external storage arrays.
Q: As well as I/O performance, what other attributes of ioMemory are finding favour among customers?
A: One of the things that our customers tell us provides a major benefit in addition to performance is the reliability of ioMemory and the cost savings generated from implementing Fusion-io solutions. Fusion-io products are uniquely reliable enough to be offered by all major OEM manufacturers, including Dell, HP and IBM.
We’re also committed to helping our customers succeed in every way, and our customers know we will go the extra mile to help. For example, sometimes our customers uncover new bottlenecks once they integrate ioMemory into their datacentres and accelerate their applications. Our support teams also help them work through those challenges to make sure they maximise the potential of their ioMemory solution.
Finally, many customers tell us that they save a lot of money on CapEx and OpEx, since ioMemory takes so much less power, cooling and real estate than traditional, scaled-out storage infrastructures.
Q: Can you point to any future directions and trends for storage technology in general, and where Fusion-io fits in.
A: As we continue to adopt new devices, and add connectivity to existing ones, we will continue to generate more and more data. To manage this growth, companies are already looking towards innovative new solutions to help meet these demands and turn data into the information that runs our economy.
Given that energy and space are finite resources, forward-thinking IT leaders are also factoring efficiency into their requirements for how to plan for current and future data demands. Fusion-io fits in very closely with these needs, as ioMemory offers a scaleable, energy efficient solution that helps IT teams deliver more with less.