This post was originally published on this siteSo one of our field engineers reached out to me because they had a power outage of some sort and their ...
Pure//Accelerate 2017 was a big moment for the storage industry. Pure announced 25+ new software features designed to redefine Tier1 storage for modern cloud workloads, deliver unprecedented performance and scale for big data analytics, revolutionize object store use cases, and unleash a new wave of data-driven innovation through a new AI engine called Meta.
The entire industry took notice, so it’s no surprise that Dell EMC® did as well. They just posted a blog highlighting Pure’s announcement of our expanded 75-blade scale FlashBlade™:
Check it out in the video below.
In the market for an electric car? You’d compare any offering against Tesla®. Looking for a new search engine? Google™ is the leading benchmark. Alternatives always measure against the leader in the industry by comparing against the solution that is in a unique position to set the bar. FlashBlade is an unprecedented scale-out file and object system purpose-built for modern, big data workloads. So it’s natural for EMC to measure the Isilon F800 against FlashBlade.
In their blog, EMC claims to deliver 9M IOPS with 144 Isilon F800 nodes. Hidden is the fact that it would take four physical data center racks of Isilon nodes to deliver on that promise, compared to just a half-rack for Pure’s 75-blade scale FlashBlade. Compare a 4U Isilon F800 chassis to a 4U FlashBlade; the EMC Isilon system offers 0.25M IOPS versus more than 1.5M IOPS delivered by FlashBlade.
The two systems have very different architectures. Isilon is a legacy architecture retrofitted with SSDs where huge amounts of data must be accessed through a narrow path into a spinning disk enclosure. Instead of using legacy software stack that’s optimized for the old media world, where data is often structured and sequentially accessed, Purity for FlashBlade is truly designed for unstructured data, accelerating modern analytics like machine learning and AI. In these modern workloads, data structures range from small, metadata-heavy to large files and access patterns can be random or sequential. FlashBlade dynamically handles this wide variety and delivers maximum performance for any data.
IOPS per TB is often cited as a key metric for modern analytics workloads. In this new era, where data is reported to be more valuable than oil, data is not meant to be simply stored but to be analyzed over and over again. While $ per TB is an important metric, an IOPS per TB metric more accurately represents the nature of big data workloads. EMC’s 4U Isilon F800 offers 250K IOPS to 924TB raw SSD capacity. Pure’s same 4U FlashBlade offers 1.5M IOPS to 780TB raw NAND capacity. FlashBlade delivers more than 7x higher performance density for modern workloads. $ per IOPS per TB metric offers a great baseline to compare cost, and FlashBlade has 7x headroom if both systems are priced the same.
At Pure, we believe performance and simplicity go hand-in-hand. Data scientists and engineers have a difficult job, building models and running analytics jobs that can consume hours to days of computing cycles in the data center. They want to rapidly find insight in their data, without having to manage complex storage infrastructure. While 75-blade scale FlashBlade delivers massive performance, it also offers the same simplicity that customers have come to love about Pure. Instead of managing hundreds of cables in four racks of storage and internetworking nodes, the 75-blade scale fits in half a rack. Customers get dramatically simplified deployment, with all networking built-in and all with a single IP address. With Pure1®, our SaaS-based management and support platform, FlashBlade is also effortless to manage. From any device, you can get a real-time view on the health of your storage.
Additionally, with Pure1 Meta, our global predictive intelligence, we deliver predictive support and fix issues before they impact any customer. Both FlashArray and FlashBlade continuously send rich telemetry to Pure1, and as we collect telemetry from systems we scan in real-time against a growing library of issue fingerprints.
We’re also thrilled to bring Pure’s Evergreen™ model to the world of unstructured data. Organizations tapping into big data are at the bleeding-edge of technology, yet their infrastructure is stuck in decades-old business models. With Evergreen, we are changing the way customers do their work:
That’s why some of the world’s leading organizations are pushing the limits of what’s possible with FlashBlade. Today, FlashBlade enables the design of the fastest race cars, underpins massive web-scale architectures, and helps design, simulate, and run next-generation automobiles and rockets.
In the fast-evolving world of AI, a web-scale company has deployed racks of NVIDIA® DGX-1 servers to build one of the world’s most powerful AI supercomputers. This impressive system is built on numerous FlashBlade systems, with the goal of keeping the powerful, data-hungry GPUs busy during deep learning training. Faster data ingest ultimately means smarter, more intelligent AI services.
It’s amusing that EMC took issue with our announcement of a (near) future product enhancement, yet EMC pre-announced an entire new product, the Isilon F800 in May 2016, under the name of “Nitro”, a year before it was publicly available. In October 2016, EMC made promises of reaching 25M IOPS in a single system, which appears to have been too optimistic in hindsight, given that they now tout much less than half that performance. In contrast, real world deployments of our single-chassis FlashBlade are today meeting or exceeding our initially claimed 1.5M IOPS. Here at Pure, when we say something, we mean it and we deliver it, and our customers tell us this aspect of our culture and our products really sets us apart.
Pure’s expanded 75-blade scale FlashBlade is just around the corner and will deliver unprecedented performance for your most challenging analytics workloads. We can’t wait to share it with you, and see what new possibilities you will achieve with it.
Please visit FlashBlade product page for more details.