Faster is, Indeed, Better! Introducing the EMC Isilon A100 Performance Accelerator Node


EMC’s new Isilon A100 Performance Accelerator Node is all about speed…so let’s think about what speed really means. You’ve likely seen a recent TV commercial where a man asks a group of young kids whether it’s better to be faster or slower. One kid responds, quite precociously, “Faster!” — and the way that he’d make his Grandma faster would be to “tape a cheetah to her back.”  I’m not sure how Grandma (or the cheetah) would feel about that, but when it comes to storage performance with the dynamic nature of today’s enterprise workflows and applications, being faster is, increasingly, a critically important point of differentiation. That’s why EMC has developed this new EMC Isilon product.

With enterprise “need for speed” in mind, EMC continues to push the frontiers of scale-out NAS with the introduction of our latest performance accelerator. The EMC Isilon A100 node is a next-generation performance accelerator designed for demanding environments that require maximum concurrent throughput and minimum latency. It will replace our existing Performance Accelerator node (Accelerator-X) and is expected to be generally available this July.

Isilon Accelerator Node 3

With 256GB of L1 cache per node (an eightfold increase over the 32GB of RAM available today), the Isilon A100 Performance Accelerator nodes can hold a huge, recently-accessed dataset in memory, enabling data to be read with extremely low latency. In addition, A100 nodes are expected to deliver aggregate throughput of 1100 MB/s per node, allowing enterprises to support high performance workflows that demand extreme levels of concurrent throughput.

In addition to providing customers the ability to scale performance independent of additional spindle capacity, A100 Performance Accelerator nodes also allow enterprises to significantly accelerate cluster operations, including disk and node rebuilds, file striping and file-based replication.

Despite the fact that EMC is introducing the Isilon A100 at the National Association of Broadcasters (NAB) annual event in Las Vegas, April 8-11, delivering the type of storage performance that can support highly parallel workloads is a “must have” for a range of industries beyond just media & entertainment. Whether it’s the genetic sequencing workflows of life science research organizations or the upstream energy exploration efforts of the oil & gas industry, extreme performance has become “table stakes” for storage vendors looking to seriously compete for the right to support next gen workflows and applications.

With Isilon solving customer needs by continuing to extend the limits of scale-out NAS performance, don’t be surprised to see our competitors get a little desperate and try to keep pace by resorting to product development efforts that involve a couple rolls of tape and one very angry cheetah…

Please feel free to share this post along with your thoughts on this new Isilon offering.

Continue Reading
Would you like to read more like this?

Related Posts

Click to Load More
All comments are moderated. Unrelated comments or requests for service will not be published, nor will any content deemed inappropriate, including but not limited to promotional and offensive comments. Please post your technical questions in the Support Forums or for customer service and technical support contact Dell EMC Support.

4 thoughts on “Faster is, Indeed, Better! Introducing the EMC Isilon A100 Performance Accelerator Node

    • Great question! The EMC Isilon A100 Performance Accelerator actually has no disks. Instead, it accelerates performance by serving cached data directly out of 256 GB of RAM. As a result, with the A100, organizations can scale performance without adding additional disks. In this way, the A100 allows you to scale performance independent of capacity.

  1. 1) Considering that there is already a multi-core CPU and sometimes over 100GB of RAM for every 2U or 4U of disks, why is it ever needed to add even more CPU and RAM? Isilon already uses more CPU cores, and more watts, per TB than any other product on the market.

    2) Knowing that Isilon clusters are currently limited to 144 nodes (based on the largest Infiniband single-cage chassis qualified with Isilon), that implies that each Accelerator node uses one of the available ports. In turn, this suggests each Accelerator node reduces the maximum theoretical cluster capacity by 36*4TB, or approximately 144 raw TB. It is possible that the cluster size may be increasing to 288, if a 288 port Infiniband is ever qualified, but regardless the Accelerator would still use 1 of those ports while adding no capacity. While this is not much per Accelerator Node, it isn\’t technically true that Accelerator nodes have no impact on capacity, because they do.

    • Thanks for your comments and questions mkulacz.

      An A100 Accelerator Node will not only add CPU and RAM to an Isilon cluster but it will also add network connectivity, scaling throughput independent of capacity. Accelerator Nodes benefit a cluster running mixed workloads – where some specialty apps may demand the highest throughput provided by the A100 Nodes while other apps such as traditional enterprise NAS workflows that are more file IOPS oriented as example, can be serviced from the S-series or X-series nodes. The additional cache of the A100s has the added benefit of accelerating all file operations of the cluster as its massive cache is additive to the cluster-wide, globally coherent cache.

      As to your concern on watts per TB, don\’t be. Isilon clusters offer leadership environmental efficiency (power, cooling and floor space). Compare these metrics for a minimum Isilon 3 node cluster to that of a minimally configured competitor\’s clustered scale-up storage product. The minimum configuration of a clustered scale-up system requires two dual RAID controller H/A systems (4 nodes) plus each RAID system\’s individual drive expansion modules – consuming more rack space and drawing more watts than the 3 node Isilon cluster. Now let\’s look at what\’s required to scale performance in the Isilon cluster and the competitor\’s clustered scale-up. To scale the Isilon cluster\’s performance, we simply add a bite sized (1U) A100 Accelerator node and and we just added an incremental 1,100 MB/sec of throughput to the cluster. In order to scale the performance of the clustered scale-up system we rack up another complete dual controller RAID system and its drive expansion modules – more rack space and more watts consumed again because we can\’t add performance without adding 2 complete controllers and additional capacity to house the datasets. I hope you\’re starting to see the value of adding CPU and RAM (and I/O) to an Isilon cluster in the form of the A100 Accelerator Node – a cheetah in 1U.

      Regarding your second point, you are correct, an Accelerator Node will use a port on the cluster\’s Infiniband switch thereby adding throughput performance without additional capacity. That\’s the intent. The A100 provides high aggregate throughput to workflows that don\’t require the maximum capacity of a 144 node cluster. Need the full capacity of a 144 node Isilon cluster and performance? That\’s covered too with over 100GB/sec of aggregate throughput.

      Performance, capacity or both. It’s a win-win for Isilon customers.

Comments are closed.