About a-team Marketing Services
The knowledge platform for the financial technology industry

A-Team Insight Blogs

Diablo’s Memory Channel Technology Reduces Flash Latency and Jitter

Subscribe to our newsletter

Memory technology specialist Diablo Technologies has introduced its TeraDIMM architecture, allowing Flash memory to be supported directly by the DRAM memory bus on servers, reducing latency and jitter for solid state disks. This approach from the Ottawa, Canada company is likely to be attractive for developers of trading applications, which leverage SSDs for such tasks as maintaining transaction state and audit trail data, and trade and quote time series, in a persistent manner.

To date, SSDs – which provide much higher performance than hard disks but still maintain data when power is switched off – have commonly been deployed via adding a PCIe card to a server. While PCIe is a high bandwidth bus, it is also fairly high latency, and prone to latency variances (jitter) as other PCIe devices – notable network I/O – typically share the PCIe bus and so contention can occur. Also, the number of PCIe card slots on server motherboards is generally limited, so that usually just a single PCIe SSD can be fitted into a server.

Working with SSD specialist Smart Storage Systems, Diablo has designed its Flash storage to work with the DDR3 memory bus that servers use to communicate between the CPU and DRAM. The Flash storage devices – 200GB to 400GB initially – are installed in the DIMM sockets on a server motherboard that might otherwise support DRAM. In a typical server, more than six terabytes of Flash can be supported, says Diablo’s vice president of marketing Kevin Wagner, though he notes that higher capacity is available in some servers with additional DIMM sockets.

Diablo reckons that compared to PCIe-based Flash, the memory channel approach reduces latency by more than 85%. It also provides very deterministic – virtually no jitter – communications, which improves the jitter performance characteristics of software applications – an important factor for low-latency trading systems.

Another benefit of the Diablo approach, says Wagner, is that blade servers – often deployed in low-latency co-lo data centres for reasons of compactness – can readily accept DIMM-format devices, whereas PCIe cards – if supported at all – generally need to be manufactured to a non-standard form factor, and so are expensive.

No operating system or application level changes are required to leverage the Diablo DIMM devices, since they are accessed like other SSDs as block disk storage. Wagner says that the company is working with a number of OEM and ISV vendors – he declines to name at this stage – to ensure widespread support for the technology, which is currently being tested by a number of customers, including Wall Street firms. When generally available, it will be available from Smart Storage.

As well as typical low-latency trading applications, the Diablo approach is also likely to find interest from those with big data analytics applications, since it supports several terabytes of fast, non volatile, storage per server.

Subscribe to our newsletter

Related content

WEBINAR

Recorded Webinar: Optimising cloud, marketplaces & managed data services

Financial institutions are under mounting pressure to rethink how they source, manage and distribute market data. Rising data volumes, multi-cloud adoption and the operational demands of regulations such as DORA are exposing the limits of legacy infrastructure, and driving firms toward managed services and cloud-native delivery models that can offer greater scalability, resilience and cost...

BLOG

LSEG Launches Model-as-a-Service, Extending Marketplace Strategy into Financial Models

LSEG has launched Model-as-a-Service (MaaS), expanding its marketplace strategy beyond data distribution into the hosting, commercialisation and integration of financial models. At launch, Societe Generale has joined as a provider, making fixed income, FX, ESG and equities analytics available through the platform. The move positions LSEG not just as a data vendor, but as an...

EVENT

Eagle Alpha Alternative Data Conference, London, hosted by A-Team Group

Now in its 8th year, the Eagle Alpha Alternative Data Conference managed by A-Team Group, is the premier content forum and networking event for investment firms and hedge funds.

GUIDE

AI in Capital Markets Handbook 2026

AI adoption in capital markets has moved into a more disciplined phase. The priority is now controlled deployment: where AI can be used safely, where it can deliver measurable value, and how outputs can be governed, monitored and evidenced. The 2026 edition of the AI in Capital Markets Handbook examines how AI is being applied...