Details
Ready for the big + fast data of tomorrow
The past decade saw the rise of "Big data" - and storage solutions designed to allow PBs of data to be centralized and served for everything from data analytics to scientific discovery to movie rendering. These storage solutions were big but often slow, as they took advantage of bigger and bigger hard disk drives, the last mechanical relics in today's otherwise all silicon data center but with the rise of cloud-native applications, needs an expectations are changing. Big but slow isn't good enough anymore and the dramatic inefficiency of rows and rows of spinning disk in the data center just can't keep up with the efficiency demands of today. What if an all flash storage platform could deliver big and fast for tomorrow's data and be simple and efficient enough to be the platform for your cloud native applications?
BIG
- 10s of Petabytes
- 10s of Billions of files and objects
- 10s of Thousands of clients connected
- All in an unbelievably small all flash footprint
Fast
- Up to 15 GB/s of bandwidth per chassis and 1M NFS operations
- Fast metadata operations and instant metadata queries
- Consistent all-flash performance with no caching or tiering
Simple and efficient
- Scale out everything online and instantly by simply adding blades
- NFS, Object/S3 and more to come
- Centralized cloud management and support via Pure1
Flashblade components
Blade
Scale-out compute + flash
- The Blade is the scaling unit of FlashBlade - adding Blades linearly scales capacity, performance, metadata and clients.
- Each blade contains a low power Intel Xeon system on a chip processor and FPGA ARM cores to run the Elasticity Software
- Blades contain raw flash chips, DRAM, as well as integrated NV-RAM protection to guard write operations from sudden power loss
- Blades are available in 8 and 52TB capacities and can be both mixed and matched and hot added into the system
- Blades were designed with ultra low latency in mind, leveraging PCIe to connect flash chips to CPUs and the low latency Elastic Fabric to communicate with other blades in the system.
Elasticity
Scale-out storage software
- Elasticity is software the heart of FlashBlade, implementing its scale out storage core, scale out metadata, scale out multi protocol support and integrated software defined networking
- The Elastic Core is the base object store upon which all protocols are built and implements native data services (data reduction, snapshots, replication and encryption) as well as flash management services (N+2 erasure coding and LDPC error recovery).
- The Elastic Map is the scale out metadata service that underlies all layers of Elasticity. It implements an extensible, variable block metadata engine which allows for instant query of complex operations
- FlashBlade and Elasticity are managed by Pure1 - Pure storage's cloud based management and support platform - making it easy for a single administrator to manage multiple systems of any size globally
Elastic Fabric
Low-Latency, software-defined networking
- FlashBlade includes a built-in 40Gb Ethernet fabric, used for both communication between blades and to clients
- The Elastic Fabric implements low-latency proprietary protocols bypassing the TCP/IP stack, separating client, data and metadata traffic in the system via QoS.
- Each chassis includes 8x 40Gb/s Ethernet ports for both client connectivity and direct chassis interconnect, as well as 30x 10Gb/s internal ports for Blade connectivity via the passive mid-plane
FlashBlade differentiators
Elastic scale out
Scale out everything online and instantly by simply adding blades. Each blade adds capacity, performance, metadata scale, NV-RAM and client connections.
Consistent all-flash latency
100% flash architecture ensures that no operation is ever waiting for disk. FlashBlade accelerates large block and small block IO and both data and metadata are persisted on flash.
Metadata performance
75% of NFS operations are metadata only operations, like getAttrs, FlashBlade implements a scale out Metadata Map - used by all layers of the system and independently queryable.
N+2 Erasure coding resiliency
Blades are deployed in N+2 redundant fashion to protect against flash or full Blade loss. When Blades failm the system is automatically heals around them, returning FlashBlade to full resiliency.
100% Flash 0% SSD
FlashBlade uses 100% raw MLC NAND and all functions of flash management and error recovery are handled by Elasticity in software. This ensures the highest performance and lowest overhead.
No stacked protocols
Other arrays often start with a single protocol, then stack others on top, with each inheriting the limitations of the layers. Elasticity implements a common Object Core and then all protocols are peers on top.
FlashBlade simplicity
Simple to deploy
Flashblade is deployed as a single 4U appliance, with a minimum of 7 blades. All networking is internal and all software is included. No tuning or configuration is neccessary.
Simple to scale
FlashBlade is scaled by inserting new blades into the system. No pre-configuration or planning require and different capacity blades can be mixed and matched to ensure future expandability.
Simple to manage
FlashBlade includes a built in GUI and CLI, as well as Pure1, Pure storage's cloud based management and support platform. A single administrator can manage a FlashBlade deployment of any size
