June 23, 2011
High performance parallel storage vendor Panasas is once again eyeing the technical computing and big data markets with the release of its ActiveStore 11 parallel storage system appliance--this time with keen emphasis on customer needs to consider storage in the face of budget constraints versus simply no-holds-barred performance.
Last year Panasas launched its PAS 12 high performance NAS storage line, which represented the fourth generation of the flagship product, which they dubbed as the world’s “fastest parallel storage system.” Notice that they are moving backwards in their numbering system with ActiveStor 11, a fact that represents going in reverse on pricing and compute-side features.
According Geoffrey Noer, Senior Director of Product Management at Pansas this backpedalling was intentional. He claims that when they built out to ActiveStor 12, they left some room in the middle to add value and create a line that was intended to offer more balanced storage capacity that is tailored to a wider range of storage budgets.
While ActiveStor 12 was made for extreme performance, Panasas has tried to find middle ground with 11. The company scaled back to make more room on the hardware end that could be spent on drives, allowing Panasas offer 11 with what Noer described as “a more pleasing dollar per terabyte package.”
The degree to which Panasas is offering something far below ActiveStor 12 can be debated since both 11 and 12 are available with 3 terabyte drives. As Noer said, “whereas before we were talking about 40 terabytes per chassis scaling up to 4 petabytes, we’re now talking about 60 terabytes per chassis, scaling up to 6 petabytes in a single file system.” The real news is contained in the fact that there is no price premium for the bump, customers who bought 40 terabyte ActiveStor 12 systems can but the same system in essence with the 3 terabyte drives and get 50% more capacity.
Noer says that moving “back” to ActiveStor 11 is not necessarily much of a downgrade from 12, given the system’s ability to scale to 6 petabytes and 115 GB/s of throughput via a single global namespace. He claims that the fully parallel performance driven by a blade design that meshes capacity and speed can scale linearly due to the shedding of filer heads and hardware RAID controllers that can buffer performance. These distinctive discards of filer heads and controllers are a hallmark of the Panasas line of storage products and drive their reputation as a suitable HPC storage option, Noer explained.
Noer also touted the deployment, use and manageability advantages that can come from a tightly integrated system versus software-only approaches. He said that Panasas is always being compared to Lustre or IBM GPSS running on SAN storage but with that taking a software-driven angle means you’re trying to marry to a hardware architecture that wasn’t explicitly designed to handle such a system. Noer says by doing this, you introduce manageability issues that Panasas is freed from due to their Object RAID model wherein are stored on the blades. Since these blades have been designed at ground level to deliver maximum performance all elements of the hardware architecture are driven to mirror advantages on the storage OS side.
Panasas is pulling in 3 terabyte SATA drives with the introduction of ActiveStor 11, which is another density bump that Noer says stretch traditional hardware RAID controllers to the limit. He claims that with each density upgrade the reconstruction times haven’t been going up as fast as the drive capacity, which is a problem since it can take several days to rebuild a single drive after failure in many hardware environments. He says that under the object-oriented paradigm, it is possible to take advantage of the massive parallelism and throw a slew of processors at the problem to reconstruct in the tens of minutes. Additionally, although it might seem a bit illogical, Noer argued that the larger the scale of the single file system the faster a rebuild takes place.
He said that when all is said and done, “we still have the software and hardware redundancy on the hardware and software services side, but in many ways our object storage architecutre enables the use of these high speed drives--and the high performance we’re getting for disk allows us to take advantage of them without becoming an archival solution.”
On a side note, if you’re off looking for something that falls just below ActiveStor 11 and find yourself on a fruitless chase for ActiveStor 10 there isn’t one. According to Noer, plans for 10 were on the table but we scrapped. He says the performance and price appeal of 11 provides a balance between previous NAS solutions and the more expensive, performance-driven ActiveStor 12.
10/30/2013 | Cray, DDN, Mellanox, NetApp, ScaleMP, Supermicro, Xyratex | Creating data is easy… the challenge is getting it to the right place to make use of it. This paper discusses fresh solutions that can directly increase I/O efficiency, and the applications of these solutions to current, and new technology infrastructures.
10/01/2013 | IBM | A new trend is developing in the HPC space that is also affecting enterprise computing productivity with the arrival of “ultra-dense” hyper-scale servers.
Ken Claffey, SVP and General Manager at Xyratex, presents ClusterStor at the Vendor Showdown at ISC13 in Leipzig, Germany.
Join HPCwire Editor Nicole Hemsoth and Dr. David Bader from Georgia Tech as they take center stage on opening night at Atlanta's first Big Data Kick Off Week, filmed in front of a live audience. Nicole and David look at the evolution of HPC, today's big data challenges, discuss real world solutions, and reveal their predictions. Exactly what does the future holds for HPC?