NetApp – Not Just Surviving but Thriving

When you’re a company that has been around for over 25 years, some people might look at you like a dinosaur, slowly plodding along as the end of the world as you know it approaches. A lot of press has been made in recent years that puts NetApp in this light. Some have said that NetApp has just been plodding along, not in touch with the industry or its customers’ needs.

Yet in the last 6 months, this “dinosaur” has started to show its teeth. The stock price has gone from $37.43 in September to a high of $71.41, and with the announcements made yesterday, you can expect that to go higher.

With the newly announced AFF A800, NetApp is now able to provide sub 200 µs latency for workloads that have the most demanding data needs. That’s an order of magnitude better than previous generations!

Not only is the AFF A800 blazingly fast, it can handle huge amounts of traffic with 25GB/s throughput on an HA pair and the ability to have NVMe end to end from the server to the storage via NVMe over FC. If using 32 or 16Gb FC isn’t a requirement you can use ethernet speeds of 100Gbe, another industry first made by NetApp. With 12 pairs clustered together, you are talking 300GB/s throughput on a single management domain. That should meet the most demanding environments.

With a current run rate of $2.0B for their all flash business, having already shipped over 20PB of NVME, and with a 44% Petabyte year on year growth, NetApp’s flash business is not only going to increase in size in the future, but with numbers like this it will survive any extinction event.

But the announcements made yesterday are not just about end-to-end NVMe-accelerated performance. There were also more advanced cloud integration messages.

NetApp’s cloud strategy is geared towards enabling customers to deliver business outcomes for all IT workloads in cloud, multi-cloud, and hybrid cloud environments. To do this, you must modernise your data management from the edge, to the core, and to the cloud.

Fabric Pool is just one of the features designed to help you do just that. Fabric Pool enables automatic tiering of cold data, which means you can purchase a smaller system or get an even higher amount of consolidation on a single box. With the release of ONTAP 9.4, Fabric Pool has been improved to allow Azure as a capacity tier and ONTAP Select as a performance tier. It can now also tier from the active primary data set, which is something I am looking forward to testing soon.

So when you look at these and other announcements that NetApp made yesterday, if they are a “dinosaur,” I would put them in the meat-eating Velociraptor camp. And that’s one dinosaur you do not want to take your eye off.

Advertisements

Setting up FabricPool

Recently, I was lucky enough to get the chance to spend a bit of time configuring FabricPool on a NetApp AFF A300. FabricPool is a feature that was introduced with ONTAP 9.2 that gives you the ability to utilise an S3 bucket as an extension of an all-flash aggregate. It is categorised as a storage tier, but it also has some interesting features. You can add a storage bucket from either AWS’s S3 service or from NetApp’s StorageGRID Webscale (SGWS) content repository. An aggregate can only be connected to one bucket at a time, but one bucket can serve multiple aggregates. Just remember that once an aggregate is attached to an S3 bucket it cannot be detached.

This functionality doesn’t just work across the whole of the aggregate—it is more granularly configured, drawing from the heritage of technologies like Flash Cache and Flash Pool. You assign a policy to each volume on how it utilises this new feature. A volume can have one of three policies: Snapshot-only, which is the default, allows cold data to be tiered off of the performance tier (flash) to the capacity tier (S3); None, where no data is tiered; or Backup, which transfers all the user data within a data protection volume to the bucket. Cold data is user data within the snapshot copy that hasn’t existed within the active file system for more than 48 hours. A volume can have its storage tier policy changed at any time when it exists within a FabricPool aggregate, and you can assign a policy to a volume that is being moved into a FabricPool aggregate (if you don’t want the default).

AFF systems come with a 10TB FabricPool license for using AWS S3. Additional capacity can be purchased as required and applied to all nodes within cluster. If you want to use SGWS, no license is required. With this release, there are also some limitations as to what features and functionality you can use in conjunction with FabricPool. FlexArray, FlexGroup, MetroCluster, SnapLock, ONTAP Select, SyncMirror, SVM DR, Infinite Volumes, NDMP SMTape or dump backups, and the Auto Balance functionality are not supported.

FabricPool Setup

There is some pre-deployment work that needs to be done in AWS to enable FabricPool to tier to an AWS S3 bucket.

First, set up the S3 bucket.

Next, set up a user account that can connect to the bucket.

Make sure to save the credentials, otherwise you will need to create another account as the password cannot be obtained again.

Finally, make sure you have set up an intercluster LIF on a 10GbE port for the AFF to communicate to the cloud.

Now, it’s FabricPool time!

Install the NetApp License File (NLF) required to allow FabricPool to utilise AWS.

Now you’ll do the actual configuration of FabricPool. This is done on the aggregate via the Storage Tiers sub menu item from the ONTAP 9.3 System Manager as shown below. Click Add External Capacity Tier.

Next, you need to populate the fields relating to the S3 bucket with the ID key and bucket name as per the setup above.

Set up the volumes if required. As you can see, the default of Snapshot-Only is active on the four volumes. You could (if you wanted) select the individual or a group of volumes that you wanted to alter the policy on in a single bulk operation via the dropdown button on top of the volumes table.

Hit Save. If your routes to the outside world are configured correctly, then you are finished!

You will probably want to monitor the space savings and tiering, and you can see from this image that the external capacity tier is showing up under Add-on Features Enabled (as this is just after setup, the information is still populating).

There you have it! You have successfully added a capacity tier to an AFF system. If the aggregate was over 50% full (otherwise why would you want to tier it off?), after 48 hours of no activity on snapshot data, it will start to filter out to the cloud. I have shown the steps here via the System Manager GUI, but it is also possible to complete this process via the CLI and probably even via API calls, but I have yet to look in to this.

One thing to note is that whilst this is a great way to get more out of an AFF investment, this is a tiering process, and your data should also be backed up as the metadata stays on the performance tier (remember the 3-2-1 rule). So, when you are next proposing an AFF or an all flash aggregate on a 9.2 or above ONTAP cluster; then consider using this pretty neat feature to get even more capacity out of your storage system or what I like to now call your data fabric platform.

The A700s Killing it in the Storage Market

On the 31st of Jan NetApp released a new box the A700s and it’s a game changer in more ways than you might know. On the same day they also released its SPC-1 benchmark results to show how pearly white its teeth are.

Hardware

Let’s start with the physical, this All Flash Array is a 4U chassis containing controllers and storage. NetApp has been producing integrated disk and controller systems for many years now but these have always been aimed at the SMB market. The A700s is the first designed with the enterprise market in mind and boy what a first. Housing 24 internal SSDs and expandability up to 216 per HA pair, with the ability to scale out to 12 SAN nodes or 24 NAS. This gives you some serious expandability options.

Each controller has four on-board 40Gbe ports with four PCIe slots for a vast array of additional functionality including more 40Gbe, maybe 32Gb FC, or even expanding to my personal favourite a MetroCluster.

Operating System

As expected this system is running the world’s number one branded operating system, ONTAP which provides a unified SAN and NAS architecture to meet the most demanding workloads. Consisting of anything from consolidated virtualisation, enterprise applications to design and engineering workloads, all delivered from a truly scale out architecture, ONTAP can manage up to 24 nodes as a single system with 13 PB of all flash storage.

SPC-1

The SPC-1 benchmark to quote their website – “consists of a single workload designed to demonstrate the performance of a storage subsystem while performing the typical functions of business critical applications. Those applications are characterized by predominately random I/O operations and require both queries as well as update operations. Examples of those types of applications include OLTP, database operations, and mail server implementations.”

This benchmark is a way to allow customers to compare various storage vendors. NetApp previously under took this benchmark back in April 2015 with the FAS8080AE and by way of comparison to how things have changed in just under two years I have put some of the more relevant results into a table. (For more detail see here and check out exec summary)

spc1table

SPC-1 Highlights

As you can see the FAS8080AE; which was before NetApp were required to adopt Gartner’s naming scheme for it to qualify as an All Flash Array; performed admirably giving us just over 685,000 IOPS at 1.2 milliseconds. This 8 node cluster was placed fifth in the top ten table but by the end of 2016 it had slowly been pushed out to 8th which is still very impressive. The most recent results published Jan 2017 show that the A700S delivered 2.4 million IOPS at roughly 0.7ms in a 12 node cluster. This huge number not only shows us the improvements in hardware that have occurred over the last couple of years but also the advances that NetApp has made with its ONTAP operating system. Even if you don’t need to scale out to 12 nodes a HA pair can deliver over 400,000 IOPS in under 0.3ms and when you consider you can stack it with 15.3TB SSDs giving you an effective capacity of a petabyte in a 4U enclosure delivering 650,000 (IOPS) in under a millisecond!

So what do these numbers actually mean? NetApp have reduced the physical capacity down from 2 full racks (84RU) to just over half a rack (26RU) whilst upping the node count by 50%. Yet by doing this they have greatly increased the throughput. So they’ve quartered your rack space requirements halved the latency and quadrupled the IOPS and this isn’t the box pushed to its max.

What’s in a name?

The S in the name has been said to represent slim but some have said it could well stand for sexy (and what a beast it is), or sport, but unlike a comparison made to hot hatchbacks, I would say this model is more the Aston Martin DB11 Intrepid Sport in Cinnabar Orange. This V12 monster is turning heads everywhere it goes, people are noticing her before they even get the chance to set eyes on her. Dropping into 3rd place in the SPC-1 is no mean feat but to do this with kit that occupies roughly half a rack is phenomenal!

The A700s is not trying to sneak around the corner no this AFA has all the capabilities we’ve come to love from ONTAP whether that be the data efficiencies with dedupe and compaction, SnapMirror, FlexClone, per Volume Encryption, Secure Multi tenancy, can form part of your Data Fabric solution, the list goes on. Remember this is an OS that keeps going from strength to strength as we can see from the addition of FlexGroup providing massively scalable next generation data containers (for more information see @NFSDudeAbides post here ) and this is hardware that marries the advances in technology beautifully.

Conclusion

All I can say is that if you are in the storage market this will have made you sit up and take note, and if you’re one of the many flash start-ups this has probably got you scared. No matter how you slice it this box delivers in all respects and is a deadly addition to any environment; just like Old MacHeath’s jack-knife.