EMC’s XtremIO X-Bricks are all solid-state systems, providing both FC and iSCSI block storage access. XtremIO systems are designed as a multi-controller scale-out architecture, adding additional controllers and capacity ranging from 1 to 8 nodes. X2 system was announced in May 2017 with new packaging, higher capacity SSDs and faster processors.
All data is deduplicated and compressed (inline) prior to being written. Dell EMC states there is no performance penalty for deduplicating or compressing data in-line. Today, an X-Brick can be purchased in different capacity configurations – 7.2TB and 138 TB. With 1.92TB SSDs in the X-Bricks, the physical capacity is 138TB per 2-controller X-Brick and up to 1.1 PB of raw capacity with an 8-node, 16-controller cluster.
The data protection of SSDs is called XDP, which distributes data in a 23+2 arrangement and is designed to not cause a performance impact for the protection updating. No spare is included in the XtremIO system. EMC notes performance acceleration with no system level garbage collection by using an in-memory metadata processing that is implemented cluster wide. Data is automatically distributed for load balancing by using part of the content-aware address (digital fingerprint) of the individual 4K blocks to select the node for storage. Access to data on the different nodes is accomplished using RDMA transfers over an InfiniBand interconnect.
EMC XtremIO system is a general-purpose block storage system designed for use in environments with high performance requirements or multiple workloads running on the same storage system. A single X-Brick contains two controllers (servers) with a drawer of 72 solid-state devices.
The XtremIO system is scalable from one to eight X-Bricks currently (2 to 16 controllers). The system comes with a number of standard features identified:
- Thin provisioning with chunk allocation in 4K increments
- Scale-out architecture and data distribution
- Read only and writeable Snapshots
- Inline deduplication and compression
- Device-based AES-256 bit encryption of data at rest
- Data protection from component failures with algorithm to optimize use of SSDs.
Their dual-stage metadata management allows for real-time data placement, balanced resource utilization, and maintains levels of performance during controller outage and rebuilds. Since the metadata and its management is done cluster wide, data is evenly distributed throughout the cluster.
The data protection, XDP has no system level garbage collection and uses back-end processes to free up empty blocks to minimize negative impacts on performance. XDP is designed to handle overwrites without the need for log structured file storage and is partial stripe optimized. Dell EMC has measured capacity overhead at 12% on average with a best case of 8%, which is low compared to RAID 1 at 50%, RAID 5 at 25%, and RAID 6 at 20%. There is no spare SSD.
The shared in-memory performance is independent from I/O access patterns. In VMware virtual environments using the VAAI X-Copy command, the copies are created at the metadata level. This method is faster than waiting until all writes are complete for the command to report copy completion from the storage. There are similar efficiencies for Hyper-V environments using ODX.
Download the full Product Brief now!