fotohansel - Fotolia
Dell EMC has unveiled the PowerMax series, the new flagship family of its SAN storage offerings. With a heritage in all-flash VMAX, PowerMax arrays offer improved performance as well as support for inline data deduplication.
At Dell EMC World last week in Las Vegas the company announced a refresh of its high-end storage range, with PowerMax unveiled as a successor to EMC’s VMAX and the latest in an evolution that originated in its historic Symmetrix family.
PowerMax takes much from the existing VMAX architecture but brings significant upgrades such as support for NVMe flash drives and inline data deduplication.
Like the current VMAX AF, PowerMax implements a multi-controller architecture based on so-called “PowerBricks”. Each PowerBrick comprises a 4U module with two active-active controllers and two shelves of 24 2.5” NVMe flash drives. The drives are dual-port to allow maximum availability in case one of the controllers fails.
PowerMax comes in two models, the 2000 and the 8000. The first is essentially an updated VMAX 250F. It can accommodate a maximum of two PowerBricks in the same rack, with 96 drives of 1.9TB, 3.8TB or 7.6TB for a maximum capacity of around 1PB. Like the VMAX 250F, the PowerMax 2000 only supports open systems (Windows, Linux, Unix) and not mainframe.
Topping off the series is the PowerMax 8000, which succeeds the VMAX 950F and can house up to eight PowerBricks. While its VMAX forebear was built on Intel Xeon E5v2 processors, the PowerMax 8000 uses the same Xeon E5v4 as the PowerMax 2000, which allows Dell EMC to boost the processing power of the array.
Like the VMAX 850F, the PowerMax 8000 supports open systems and the mainframe IBM Z-Series. PowerMax 8000 instances can be dispersed to different racks in the same datacentre to protect against electrical failure or other incidents within the same datacentre. As with its VMAX predecessor, arrays can be connected by a dual Infiniband fabric.
Read more about flash storage
- NVMe could boost flash storage performance, but controller-based storage architectures are a bottleneck. Does hyper-converged infrastructure give a clue to the solution?
- NVMe brings blistering performance gains over existing HDD-era disk protocols, and is a straight swap-in for PCIe server-side flash with array and hyper-converged products on the way.
Dell EMC claims the PowerMax 2000 can deliver up to 1.7 million IOPS (read, 8k blocks) with latency below 300µs, which is about 50% better than the VMAX equivalent. Meanwhile, the PowerMax 8000 tops out at 10 million IOPS with a similar latency figure, but with throughput of around 150GBps.
For PowerMax, Dell EMC has significantly re-worked the OS code that drives its high-end systems and has upgraded its file system to optimise data placement on its drives. Eventually, these arrays will support up to eight classes of service, but that depends on Dell EMC adding support for Intel Optane drives to the flash drives currently supported
Another new feature is native support for inline data deduplication. Dell EMC’s position historically was that the technology held no interest for it in its high-end arrays, but it has bowed to pressure from competitors. The company has recently added deduplication to its Unity arrays and decided to do the same for PowerMax.
According to those responsible in the company, the algorithm used allows for a maximum reduction ratio of 5:1 and an average of 3:1. In the high-end storage array market, support for deduplication returns the advantage to Dell EMC over the DS888x systems from IBM and allows the firm to go up against Hitachi Vantara (and HPE, which re-sells Hitachi systems re-badged as XP7).
Storage admins can select data duplication and compression functionality by volume without impact on existing data services.
The two new PowerMax models are available this week, with Dell EMC indicating that new configurations that incorporate Intel Optane drives are set for 2018 release.
Support for NVMe-over-fabrics is also part of the Dell EMC roadmap, but the firm has not explained whether the protocol will be used to link server-based storage or to support shelves of NVMe-over-fabrics compatible external storage.