SanDisk®’s announcement of its ULLtraDIMM flash SSD opens the door to acceleration of a wide variety of customer applications and workloads. Placing flash storage directly next to server processors dramatically accelerates application performance – and database performance.
With ULLtraDIMM, a broad range of applications and databases will be supported, for use in a variety of computing environments. This includes enterprise datacenters running mission-critical workloads – datacenters operated by hosters and CoLo (co-location) providers—and cloud service providers (CSPs).
ULLtraDIMM installations will provide multiple terabytes (TB) of capacity for data-intensive workloads. SanDisk, which worked with Diablo Technologies Inc. to connect the ULLtraDIMM to the DDR3 memory channel, is pioneering this DIMM-slot-compatible innovation for servers. Servers with this technology will start shipping into the marketplace in coming months.
Let’s take a closer look at some of these workloads:
Accelerating Data-Intensive Workloads
Data-intensive workloads, such as Big Data workloads and analytics software running against the Big Data repositories, will be able to rapidly pull in large “chunks” of data for processing directly into the flash storage – and the time-to-results for business will be greatly reduced. Analytics finds the “patterns in the data”, transforming transactional data into actionable data that leads to business decisions. For example, data scientists, working in business units, use analytics to identify credit-card fraud detection, to improve supply-chain communications and to make logistics/distribution changes.
Data transfer rates will be accelerated by ULLtraDIMM, reducing total processing time for large enterprise databases and data warehouses that are based on structured data. But organizations today know that analytics workloads also process large amounts of unstructured data, as well (e.g. audio, GPS/positional data, video and image data-types, including medical images for diagnostic purposes).
The high data-transfer rates will benefit high performance computing (HPC) workloads, such as computational fluid dynamics (CFD) simulations, visualizations of oil & gas fields; seismic research applications; genomic modelling and protein-folding apps for biotech workloads.
In-memory databases – defining a new market space – require high performance and low latency to speed time-to-results for database workloads that, quite literally, move significant portions of large databases into a single server for on-board, unified, processing. Examples include SAP HANA workloads – and the new in-memory database support built into Oracle’s flagship Oracle Database 12c database.
Adding Capacity to Cloud Computing Deployments
Cloud computing is characterized by high-performing computers – arranged in tiers that handle web-serving, application-serving and database-serving. Cloud workloads scale by scaling out, rather than by scaling up on a single server. High numbers of 1-socket and 2-socket servers are deployed in the cloud datacenter, numbering into the thousands, or even tens of thousands, of individual servers. This environment thrives on fast data transfers – and low-latency I/O.
Flash technologies enable these cloud computing servers by making them faster and more efficient in the way they support memory, and the way they handle transient data. That’s a good thing, because the sheer volume of data that arrives, in waves, from the Internet and the corporate intranet, threatens to become a data tsunami that has the potential to bring processing to a halt.
By adding capacity to each server, and by speeding performance in each server, ULLtraDIMM-enabled systems can be used as a strong building block for extensible – and scalable – cloud computing infrastructures. ULLtraDIMM is well-positioned to become an important part of that transformation.
Addressing Virtualized Environments
Today, enterprise applications (e.g. ERP, CRM) and databases run inside virtual machines (VMs) – but many customers would like to improve end-to-end performance for these enterprise solutions.
With fast data transfer rates – and high capacity – ULLtraDIMM-enabled servers will be able to tackle more challenging enterprise workloads, of the type that once ran only on mainframes and scalable Unix servers. That’s important, because most workloads running in the enterprise today (on physical or virtual servers) run on Microsoft Windows or Linux – or both – on top of a virtualized x86 infrastructure.
Virtual desktop infrastructure (VDI) allows hundreds of PC desktop “images” to be hosted on central-site servers, for better management and control – and for uniform back-up and replication procedures to preserve corporate data and security policies. But “boot-storms” have troubled these VDI deployments – as all of the employees start up their VDI sessions at once, when they arrive at work. ULLtraDIMM speeds application performance for this central-site service, improving the customer experience.
Summary: Flash That Extends Server Memory
Flash technologies are powering a revolution in improving workload performance, especially for data-intensive applications and databases—and for cloud-centric computing that depends on fast processing. Installing flash directly next to a server’s processors, without travelling through the storage software I/O stack, supports dramatic improvements in time-to-results.
This new technology will track well with evolving server system designs, which are built to support rapid technology refresh cycles. As the need for more onboard storage grows, driven by Big Data, Analytics and Cloud Computing, the ability to leverage ULLtraDIMM in servers will expand storage capability for data-driven workloads. It can be expected to play a strong role as a new generation of flash-enabled servers is deployed throughout the datacenter.