Optimizing video data storage: Latency issues in write intensive applications

Latency issues in write-intensive applications

As virtualized storage solutions become more prevalent, ensuring efficient data management is critical. Virtualization often brings with it complexity and overhead that result in latency, which can negatively affect performance. When solutions are virtualized, data that is written/read from the disk interacts with both the physical and virtual hardware, and each hop that data makes before it reaches its destination increases overall latency. This latency, while often in the millisecond range, quickly adds up and can cause the data’s integrity to be degraded, or cause packet loss while writing to the disk.

Video data behaves much differently than traditional IT data, as it is bulky, highly write intensive, and requires hardware that is purpose-built to manage this data. BCDVideo’s Hyperconverged Infrastructure for Video Surveillance (HCI-VS) is engineered to be optimized for physical security and tailored to handle video surveillance data efficiently and effectively.

Powered by Scale Computing’s Reliable Independent Block Engine (SCRIBE), BCDVideo’s HCI-VS provides a virtualized environment that is optimized for large and complex video data. Due to how Scale Computing’s proprietary hypervisor software HC3 is designed, it allows data to have a maximum of three hops before it reaches rest. Compared to other software that achieves virtualization, hops can range from 10 to 13. This dramatic increase in hops for data introduces an extremely high amount of latency that can significantly impact a solution’s performance, especially for video data.

Accommodating this latency forces physical hardware changes, where a layer of flash storage is used as a buffer for data ingress in order to process video data. BCDVideo’s HCI-VS does not require a layer of flash storage to properly handle video data, enabling greater storage scalability and higher operational performance of the cluster because of how close the hypervisor operates in relation to the bare-metal hardware.