The Latency Road Block
Last week I joined Storage Switzerland lead analyst, George Crump, for a joint webinar on one of the main storage system road blocks: latency.
When comparing flash performance, IOPS and MB/s are the most common metrics used to assess devices. But when it comes to data center architecture, it’s not only IOPS that stand in the way of high-performance, latency has become a top concern for IT managers today.
In sectors such as financial services, millisecond latency delays can end up costing millions of dollars. But not only transactions are dependent on speed. In today’s IT world, everyone from users, to customers and operators, have all come to expect application and web environments to deliver instantaneous responses. In order to achieve that, data center and application architects seek high-performance solutions that can provide consistent and predictable results when designing their systems.
Flash in the Data Center
Flash has seen a rapid adoption in the enterprise space mainly due to its ability to enable far greater performance and predictability (Quality of Service) than spinning drives. But flash is not only replacing hard drives where they traditionally were located in the system, it is also emerging in new forms and software solutions that help different points in the storage system benefit from its many advantages.
SanDisk®’s ULLtraDIMM is a storage device located in a new tier that blurs the lines between memory and storage by connecting flash directly to the memory channel using the DIMM form factor. Its location in the system architecture enables it to communicate directly with the memory controller without routing through the I/O hub. As a result, it can deliver extremely low and consistent latency, approaching DRAM performance, but at a much lower cost point.
The high performance of SSDs can often expose other bottlenecks in an existing system, such as bandwidth, contention or even software or protocol stack overheads. For this reason, the full potential of SSD performance is not always met, as it is often limited by server back-end piping. One of the unique features of ULLtraDIMM is that performance can scale linearly with every device added as the memory channel can support much larger bandwidth and there is no contention with any other device on the bus.
Reaching the Full Performance Potential of Flash-Based Storage by Leveraging the Memory Bus
To learn about how ultra low latency is impacting applications, virtualized environments, data center infrastructures or TCO, you can now watch the full webinar (55 mins) on BrightTalk.
Here is what you can expect to learn:
• What is latency and how does it differ from IOPS or throughput?
• How does ULLtraDIMM differ from other DIMM form factor storage devices such as SATA-DIMM or NV-DIMM?
• How flash can offer comparative performance to DRAM, in a more cost effective solution?
• How many IOPS could the ULLtraDIMM potentially support in a server?
• How can the ULLtraDIMM help reduce data center infrastructure costs?