Big Data – A Need for Speed – Hello PureStorage FlashBlade!

Fast does not always mean high IOPs. Sometimes hundreds of thousands of IOPs just is not enough. Big data applications such as machine learning, real time data processing, and backups of big data workloads needs high throughput and massive scaleout. Enter PureStorage’s FlashBlade!

Start from 7x8TB to 75x 52TB blades!

Have you ever asked yourself any of these questions?

  • How do you backup databases which are hundreds of TB in a timely manner?
  • How to you ingest and index Splunk, Spark or ELK logs from thousands of sources?
  • How do you ingest and process video from hundreds of video cameras?
  • How do you train machine learning systems like TensorFlow or Caffe with petabytes of source data?
  • How do you easily provide a scalable NFS, SMB, or S3-compatible storage solution to your researchers?

If so, PureStorage FlashBlade just may be for you!

 


PureStorage FlashBlade is winning and clever design on what others only hope to achieve: Multi-protocol, high-thoughput, scale-out storage.

Using Intel System-on-a-chip, mixed with FPGAs and Flash chips to store data, PureStorage has taken a step away from the old-fashioned controller-based architecture, to provide a solution which can scale from 7 to 75 Storage Blades.

Flashblade Chassis

Flashblade Blade

  • Each 4u Chassis provides:
    •  8x 40gb, or 32x 10gb network ports
    • Space for up to 15 blades (7-blade minimum)
    • ~17GB/sec Throughput on a fully populated chassis.
    • Under 1500 watts

 

 

With the latest FlashBlade PurityOS, up to 75 blades can be tied together in one fabric.  That is over 75GB/sec of read & 25GB/sec of write throughput, at just 6500 watts! 



 

 

Rise of the Machines [Learning]…

You may not have heard of machine learning, but you use it every day.

  • Yelp’s restaurant images are all sorted and categorized automatically
  • GMail’s “Smart Replies”, 12% of all replies, are all created by machine learning from all user’s past responses.
  • “Machine Vision” in self-driving cars
  • Email malware and virus detection
  • Automatic Translation on websites
  • Targeted advertising and geolocation

Machine Learning is all about turning Unstructured Data into Actionable Information

Machine Learning is exactly that.  You “train” the system with data, to make it learn patterns of recognition.  On many systems, these are massive data lakes or feeds.  Hundreds of thousands of images or documents.  Real time video from many sources. Millions of emails. This needs high throughput.    The ability to utilize NFS and S3 to store and feed your data, and ease of use PureStorage’s FlashBlade is the best solution to support your machine learning needs.

If you are not integrating one of the machine learning tools such as TensorFlow, Caffe, aiWare, and the many others available into your architecture and design, you may be missing out on utilizing all the data you have to the fullest. Your competitors may just be getting the edge on this.

 

 



 

Let’s talk databases and backups… 

I have had the {mis}pleasure  of dealing with Oracle, in one way or another, for over 20 years.  It was a great joy when I came across Oracle ACE Ron Ekins’s blog about his performance using Oracle, DirectNFS, and FlashBlade.  His testing showed 4GB/sec of single node performance, and 8.8GB/sec of two-node RAC performance on over Direct NFS with PureStorage FlashBlade.  With this much available throughput, not only is processing and table scans amazingly fast, but backups are as well.  How fast are your Oracle dataloads and backups?

 

RAC + FladeBlade – 8.8GB/sec

 

 

Another story:

A few months ago, we deployed a pair of PureStorage FlashBlade appliances for an enterprise customer for Splunk archive and replication tiers.  Their Hot Data and Indexers were running on

PureStorage FlashArrays.  Within a few hours, we had hundreds of TB of scalable storage, ready to be used.   The customer was trained on the new hardware in minutes, since it is intuitive and works much like Pure1, and their FlashArray.  PureStorage has continued the ease of use that they are known for.  The customer was so happy with the performance and ease, they are filling out their chassis with blades, with many more on the horizon!

 

 

 

Other customers are utilizing PureStorage FlashBlade for Veeam and Rubrik archive with great success!  Flashblade’s insanely fast throughput makes for insanely fast restore speed!

 

If you want to hear success stories for everything from medical imaging, to genomics research, to drone data collection, let me know!

 


 

 

 

 

 

 

Be Sociable, Share!
Comments are closed.