Tech ONTAP Articles

Flash Forward: Get Ready to Deliver Flash as a Service

Tech_OnTap
18,783 Views
pdf-icon.gif

August 2015

author_andrew_grimes_86x86.jpg
Andrew Grimes
Principal Architect for Flash
NetApp
 

You've probably seen the headlines announcing that "flash changes everything" in the storage industry. The truth is that the fundamentals of storage remain unchanged. You still need to serve data with reliability, to protect your data, and to deliver IT services across a range of requirements, regardless of the media under the data. But as flash technology matures, it lets us change the way we deliver those services: with greater efficiency and performance, and with the potential for major savings for your organization.

 

The evolution of flash goes hand in hand with other changes sweeping through our industry, most obviously the migration of data to the cloud. Any storage solution using flash shouldn't isolate you from that change. Today, most flash solutions are silos with a few features that take you back to the days of stand-alone servers and external backup software. To drive business value, flash must support your enterprise data management, both on premises and in the cloud—without compromises.

 

At this October's NetApp® Insight™ conference in Las Vegas, I'll be sharing some ideas for how to make that happen.

Flash Matures Beyond Silos

Initially, flash was all about raw speed. If you had an application that couldn't tolerate high latency, it was worth the high price and single-purpose performance focus to use flash media. But disk prices are reaching a tipping point. NetApp estimates that by 2016, flash solid state disk (SSD) prices will drop below prices for serial-attached SCSI (SAS) hard disk drives (HDDs).

 

As a result, flash technology is rapidly replacing traditional 10k spinning HDDs. The problem with the current iterations of these new arrays, however, is that they are just silos of flash with limited features and tools for data management.

 

Early adopters of flash were willing to overlook the inadequate data management and data protection features of flash arrays to gain the benefits for database acceleration, virtual desktop infrastructure (VDI), real-time apps, and other performance-driven applications. But as more enterprise operations move to cloud or hybrid cloud environments, it's no longer realistic to expect enterprises to work around flash arrays' limitations.

 

The era of the shared flash storage array is here. Yes, this new class of flash array is fast. But it doesn't make you go without the features you expect from enterprise storage.

 

 

Figure 1) Enterprises use flash technology for a variety of applications (customer use cases).

 

feat_fig1_lg.jpg

 

Source: 451 Research LLC, July 2015

What "Enterprise-Grade" Flash Really Means

Flash arrays that run a range of applications are becoming the norm. In fact, they are considered table stakes for the flash space. Shared multi-tenant flash arrays are going to be the standard, and if a vendor can't innovate to that standard, its days are numbered.

 

To support these changes, IT consumers of flash will need all the tools and features of traditional HDD arrays merged with all the benefits of flash. The performance is already in place. Now it's a race to give the enterprise the rest of what it needs for enterprise-grade flash.

 

But what, exactly, is "enterprise-grade flash"? In addition to the assumed capability of superb storage efficiencies, it must also deliver:

 

  • A unified architecture that supports both networked-attached storage (NAS) and storage-area network (SAN) protocols
  • Scale-out storage with dynamic capacity additions
  • Built-in backups
  • Cloud-enabled operations for the complete data life cycle
  • Quality of Service (QoS) to manage and enforce service levels
  • Automation and orchestration
  • Application integration
  • Secure multi-tenancy
  • Snapshot copies for on-array backups
  • Storage-reduction techniques, such as deduplication and thin provisioning
  • Synchronous and asynchronous replication
  • Storage encryption

 

All of these capabilities are available today in NetApp All Flash FAS systems, thanks to the powerful data management capabilities of our Data ONTAP® storage operating system (OS). Because our AFF8000 All Flash FAS systems run the Data ONTAP OS, they automatically inherit the full complement of enterprise-grade capabilities within their all-flash architecture.

 

"Our dealership customers expect a live market view of vehicle inventory and pricing. To provide dealers with real-time data, we've moved a 15 terabyte Oracle database to a NetApp FAS8080 EX all-flash array with outstanding results. Database latency now averages a mere 370 microseconds, helping us provide consistently excellent performance and maintain the attention of the dealer community. And our IT operations team can use all the NetApp software features they have come to rely on, such as deduplication, replication, snapshot copies, and thin clones."

 

— Chris Prosceno, Manager of Enterprise Data Storage, Cox Automotive (parent company of Autotrader.com, Kelley Blue Book, vAuto)

Flash as a Service

As flash adoption becomes mainstream, flash will become a service level of your infrastructure—a way to deliver multi-tier performance, all from the same media types.

 

Enterprise IT is now about delivering services to the organization. Those services can be co-located off site or delivered on premises, or by a service provider, or in the cloud. Modern IT has to support all methods of service delivery so that your data is not stranded.

 

Flash must be a part of this service delivery model, providing tier 0, tier 1, tier 2, tier 3, data recovery (DR), archive, and cloud data support. Your flash storage can't imprison your data. Your shared multi-tenant flash arrays must work with everything in your IT environment. For example, you need:

 

  • Replication from flash to capacity drives for cost savings, development/test, DR, and archive
  • Replication from flash to cloud for cost savings, development/test, DR, and archive
  • Efficient flash storage management across both sites and platform sizes, and as your IT environment grows, evolves, and scales

Benefits of Enterprise-Grade Flash

Is enterprise-grade flash really that important? Does it deliver real benefits? Our answer is a resounding YES. Here are a few examples:

 

  • Increased storage efficiency. Achieve higher petabyte (PB) per full-time equivalent (FTE) data administrator, up to 2PB to 3PB per admin.
  • Power and space-cooling savings. Replace three floor tiles by 8U of rack space, or more.
  • Improved organizational efficiency. Use fewer technologies, and have fewer people to manage.
  • Scalability. Migrate easily between underlying technologies as your IT environment evolves.
  • Reduced costs. Reuse your expensive assets for years to come, with nondisruptive operation (NDO) and scale-out capabilities. NetApp customers report an average disk age of seven years with our technology. This longevity allows you to reuse and keep your disk even as you upgrade.

 

Figure 2) Oracle Database performance excels on NetApp AFF8080 EX All Flash FAS systems.

 

feat_fig1_lg.jpg

 

Source: "NetApp AFF8080 EX Performance and Server Consolidation with Oracle Database, TR-4415"; June 2015

 

 

Figure 3) Microsoft SQL Server performance excels on NetApp AFF8080 EX All Flash FAS systems.

 

feat_fig1_lg.jpg

 

Source: "NetApp AFF8080 EX Performance and Server Consolidation with Microsoft SQL Server 2014, TR-4403"; June 2015

Let's Talk More

By now, I hope you're convinced that the flash revolution is real and that there's much to be gained by putting enterprise-grade flash to work for you.

 

Unfortunately, however, the revolution has caught many storage vendors off guard. They're scrambling to invent the features NetApp has been providing all along. Our All Flash FAS systems prove that enterprise-grade flash is available today, not as some future promise.

 

The Data Fabric, enabled by NetApp, is another big part of the story of our evolving industry. The Data Fabric works with our flash solutions to give you cradle-to-grave, on-premises-to-cloud life-cycle support for your data.

 

I hope you'll join us at the NetApp Insight conference in Las Vegas in October. We can talk about how NetApp enterprise-grade flash, with cloud enablement, can help address your current and future data issues.

 

The market is changing. Are you ready?

 

 

Andy Grimes has been in the IT industry for 15 years. He worked as a business systems architect, systems engineer, and chief technology officer for healthcare at NetApp before becoming NetApp's principal architect for flash. Previously, he worked for IBM developing and testing storage technology, generating four storage technology patents.

 

Please Note:

All content posted on the NetApp Community is publicly searchable and viewable. Participation in the NetApp Community is voluntary.

In accordance with our Code of Conduct and Community Terms of Use, DO NOT post or attach the following:

  • Software files (compressed or uncompressed)
  • Files that require an End User License Agreement (EULA)
  • Confidential information
  • Personal data you do not want publicly available
  • Another’s personally identifiable information (PII)
  • Copyrighted materials without the permission of the copyright owner

Continued non-compliance may result in NetApp Community account restrictions or termination.

Replies

Saluting Andrew,

 

Would you please alsoa dvice the top 3 reason AFF8080 is better than M Series from Pure Storage?

 

Tks from Henry PAN

The fundamental differences between Pure and NetApp make such a comparison difficult. To compare apples to apples, you have to reduce the functionality and the requirements to an incredibly minimal level. The Pure system (like many startups) is simple because it does so little. Dedupe and Compression are table-stakes and in the AFA space, but scale-out is also a requirement and Pure can’t even do that now. Today they focus on SAN, and only FCP is used, replication that is rarely if ever used, and clones/snapshots that are very hard to utilize. Lack of scale-out is a fatal handicap for Flash, so it’s important to factor that in. Most AFA implementations are just performance with efficiencies, and that gives them enough of a story for a “silo”. But, shared multi-tenant flash is here and we need more than just a fast, efficient silo. Low Latency is also typically asked for and there are many AFAs on the market that are not spectacular. So it really comes down to what are the requirements. 

 

  • An m70 is probably similar in throughput to an 8080 controller, but apples and oranges in functionality. 
  • An m50 is probably about an 8060 in throughput/performance, but is missing all of the key features of a true SAN
  • An m20 is around an 8020/8040 in performance but is missing the features to deliver a complete solution. 

 

If you need features, then take this example: Do you really need to buy Commvault to get a consistent backup of SQL on Pure storage? What about the others, Kaminario, SolidFire, Tintri, Nimble, and a long list of new startups.

 

It is not hard to be easy to use when you do one thing, but what is “easy” if you have to layer a Commvault, TSM or NetBackup on top. If you reduce your requirements to just what they do, are you really solving the business problem? Or, are you just creating a dozen more further up the stack. Add to the lack of scale-out, protocols, backup and recovery tools, and integration with monitoring and management tools at the Enterprise, and most of the startup AFAs become just dumb very expensive disk. 

 

Welcome to the world of Enterprise Grade Flash. It’s more than about IOPS.

 

 

Public