Panasas has been around for quite a few years as a major proponent of scale-out storage. For years, it seemed the company was one of the lone voices in this arena. These days, many of the old-guard storage OEMs and several startups have cottoned onto the value of a scale-out architecture and are trying to muscle in on the Panasas beat.
But the company has a long head start and continues to innovate in this area, as evidenced its Panasas ActiveStor family of scale-out NAS storage appliances. These are best suited for demanding high performance compute (HPC) workloads in the bioscience, energy, finance, government, manufacturing and other research sectors.
They leverage the PanFS file system running on a modular hardware architecture, which scales from 40 terabytes to 6 petabytes of storage. At the same time, performance scales linearly from 1.5 GB/second to 150 GB/second as additional capacity is added to the system. Storage administrators can add storage to the global namespace from a single point of management in less than 10 minutes without disrupting workflows. ActiveStor systems also feature user quotas, snapshots and chargeback reporting so administrators can monitor and manage storage resources within a private cloud. Several versions are available.
ActiveStor 8 and 9
ActiveStor 8 is the entry-level HPC storage system from Panasas which is described as providing superior single client and aggregate bandwidth performance. ActiveStor 9 brings in solid state disks (SSD) that accelerate metadata operations, providing extreme levels of throughput for performance-sensitive applications, random I/O workloads and large and small files.
At 1.5 GB/s throughput per chassis, ActiveStor 12 is the fastest HPC storage system in the world according to benchmarks conducted by the Enterprise Strategy Group (ESG).
“It provides unrivaled performance per gigabyte for complex business-critical applications,” said Barbara Murphy, Chief Marketing Officer, Panasas.
Sitting somewhere between Versions 8/9 and 12, the newly announced ActiveStor 11 system provides high capacity and balanced performance for more cost-conscious HPC deployments.
“Panasas storage systems have the highest per-disk performance of any solution in the industry,” said Murphy. “While other scale-out NAS systems provide good scale-out capability, they are bottlenecked by filer heads and RAID controllers and cannot reach the performance levels of a Panasas system.”
For example, Isilon’s maximum quoted performance (85GB/sec), she said, is only a slightly more than 50 percent of what a Panasas system can achieve. That extreme parallelism coupled with an easy-to-manage scale out model is how Panasas differentiates itself.
“Panasas allows the HPC cluster nodes to directly access the storage appliance ensuring maximum performance of the system,” said Murphy. “In addition, performance scales as the system capacity grows, which is essential for HPC applications.”
While the ActiveStor 12 appliance is the company’s highest performing scale-out system, ActiveStor 11 delivers a combination of high capacity and good performance (not as high as 12) at a lower price point.
Blade Building Blocks
ActiveStor storage appliances are comprised of 4U chassis building blocks. Each chassis has 10 hot-swappable storage blades and one director blade, which directs traffic between the compute cluster and storage. Each storage blade manages two hard disks and has its own dedicated processor and memory. Both the ActiveStor 11 and 12 ship with 3 TB drives for a maximum capacity of 60 TB per chassis. They scale to 6 PB in a 10-rack system. The products use multiple dedicated embedded processors from the Intel Xeon product family. ActiveStor 11 has 48 GB of cache, while the higher performing ActiveStor 12 has 92 GB of cache. The system is 10Gb Ethernet enabled, and users can choose a router product that bridges across to InfiniBand if desired.
“High performance computing used to be relegated to government labs and core research, but is now much more prevalent in everyday business applications, so-called big data workloads,” said Murphy. “Traditional NAS systems, while providing good file level access, cannot handle the level of data throughput required to perform complex engineering and scientific research.”
Parallelism, she believes, is essential for the next generation of scientific research. What was once considered niche is now essential to everyday company research. In these environments, therefore, a storage system must be capable of scaling dynamically while continuing to scale performance as the data set grows. Parallel programming and parallel storage architectures like those used by Panasas are designed with that in mind.
“Computationally intensive workloads are best managed within a private cloud, allowing the system administrator to consolidate his very expensive compute clusters and efficiently schedule different departmental work projects,” said Murphy. “Panasas storage systems are optimized for the private cloud because it is essential that performance scales as capacity grows.”
The rise to prominence of the cloud, and the need for secure and scalable private clouds, then, are key trends that open the door to increased market share for Panasas.
“As private clouds become more pervasive, it is clear that increasing numbers of HPC users will require highly scalable parallel storage systems that are dependable, easy to manage and deliver the high throughput required for a wide range of technical computing applications,” said IDC analyst Earl Joseph. “The Panasas ActiveStor 11 appliance is well positioned to capitalize on this trend.”
Drew Robb is a freelance writer specializing in technology and engineering. Currently living in California, he is originally from Scotland, where he received a degree in geology and geography from the University of Strathclyde. He is the author of Server Disk Management in a Windows Environment (CRC Press).