Datrium DVX System With Flash End-to-End

Datrium, Inc. announced its next generation DVX system with Flash End-to-End (E2E), along with additional availability features for tier one apps.

While DVX was already nearly four times faster than all-flash arrays ⁽¹⁾, Flash E2E takes automatic VM and container performance to high levels, offering four times higher write throughput per terabyte while still enabling rich data management across snap replicas on flash, disk or cloud. The company also announced a high benchmark result in the history of IOMark with support of 8,000 IOMark-VMs, five times greater than the next closest result and ten times greater than the highest hyperconverged result.

According to Gartner, Inc., “A private cloud built atop an HCIS (HyperConverged Infrastructure System) will either be smaller in total or made up of many independent clusters of nodes. The former option is highly limiting, and the latter is cumbersome to manage.“ ⁽²⁾

Unlike Hyperconverged Infrastructure, DVX is Open Converged
Customers can run mixed workloads at scale as well as avoid both storage and hyperconverged cluster management entirely, since Open Convergence keeps flash-rich hosts stateless and performance-isolated. There is no need for users to manage QoS throttling utilities, because high performance is predictable and automatic. Open Convergence also eliminates the need for customers to purchase and manage separate backup silos because powerful VM data management is built-in, and data reduction efficiency is always on.

Extending virtual server uptime
DVX with Flash E2E improves host and VM uptime. In the case of any SSD failure on a host, customer VMs will continue to run with performance similar to that of an all-flash array, with random reads served from the persistent flash data pool. In the event of a complete local flash failure across multiple hosts, VMs continue to run unaffected with DVX Peer Cache. This software boosts host flash failure protection by leveraging a neighbour host’s flash for read speed without restarting VMs using VM HA. In addition, the company has certified DVX for use with VMware Fault Tolerance (FT) as a separate vehicle for continuous VM operation in the event of complete host failure.

10X higher VM performance than hyperconvergence
DVX has always been suited for performance-intensive workloads by leveraging local host CPU and flash for the fastest possible read IO, isolated from the workloads of other hosts. DVX with Flash E2E takes write performance to the next level for deployments such as multi-thousand seat VDI, real-time analytics and IoT. In a recent IOMark benchmark, DVX with Flash E2E shattered the previous record by five times and a high hyperconverged result by ten times, posting a record-breaking 8,000 IOMark-VMs. System-wide read performance is 200GB/s bandwidth and 18million IO/s, while system-wide write performance is 8GB/s for a disk-based data pool and 16GB/s for Flash E2E.

“Evaluator Group has performed hands on testing of many different storage and hyperconverged systems, and we were excited to see the new record IOmark-VM results that Datrium has achieved,” said Russ Fellows, senior partner and analyst, Evaluator Group, Inc. “These levels of performance and scalability enable customers to run thousands of VMs in a single Datrium hybrid cloud, scaling well beyond levels achieved by hyperconverged configurations.“

Data management retention across media types
Open Convergence is fundamentally different from hyperconvergence because it combines the fastest application performance, the best price-performance for backup and replication, and built-in VM – or container-centric policy and catalogue tools. DVX can store snaps economically with dedupe, compression, always-on wide erasure coding and encryption, across data pools on flash, disk or cloud. This frees customers to make the best price/recovery ROI decision for VM placement and retention policy by site, without having to involve multiple third-party, secondary storage solutions and management frameworks. In addition, if a user chooses to store their long-term snaps on disk with DVX, performance is still many times that of an all-flash array.

“Datrium DVX is the world’s simplest infrastructure for scaling VMs and containers, and with Flash E2E, it’s ready for Tier 1 apps,” said Brian Biles, CEO and co-founder, Datrium. “This is the last convergence choice you’ll ever need to make.“

Pricing, configuration and availability

• DVX F12X2 Data Nodes (all flash) offer up to 83TB of effective capacity, based on always-on wide erasure coding (2FT), global deduplication and compression. Up to ten flash Data Nodes can be pooled in a single DVX for up to 830TB. US list pricing for DVX D12X4B (disk-based) and F12X2 Data Nodes are $94K and $186K, respectively, and are available.

• DVX CN2100 Compute Nodes have been refreshed with Skylake processors and NVMe flash drives. Pricing starts at $17,500 list price and are available.

• All company’s nodes support a 25GbE option for additional performance.

• Peer Cache software is included in with DVX software at no additional charge.

⁽¹⁾ Based on internal lab testing of 70/30 R/W 8K I/O of DVX versus published Dell/EMC XtremIO results
⁽²⁾ Gartner: Deploying Hyperconverged Integrated Systems: Eight Great Use Cases, 29 November 2016

Ressources:
DVX specifications: Compute Node Spec Sheet, Data Node Spec Sheet
Blog: Could This Possibly Be The World’s Simplest Tier 1 Cloud Data Center Infrastructure?