I’m happy to share that we are releasing NetApp® StorageGRID® 11.7, and I could not be giddier.
The new release packs some very interesting features and improvements. It also sets us up for an exciting road ahead as it introduces two major changes to the StorageGRID DNA: Cross-Grid Replication and a new all-flash node that leverages commodity hardware. In this blog post I discuss these two major developments, as well as some other improvements in this release.
Improved usability
Let’s start with the refreshed GUI. As soon as you log in, you'll notice a new dashboard. The new home page is significantly enhanced with links to Performance, Storage, ILM, and Nodes.
I found myself glancing at the metrics in Performance and ILM panes while running a FabricPool tiering workload from ONTAP® to StorageGRID and appreciating how easy access made me use the StorageGRID metrics more often.
An even more powerful feature is that you can create your own dashboard by using the Actions dropdown button at the top right, and you can decide which dashboard should be the default you see when you log in (short overview video here).
Onboarding wizards
There's a new wizard for onboarding a FabricPool workload (short overview video here). The StorageGRID 11.7 FabricPool wizard is the best place to start when enabling ONTAP tiering. The wizard creates all the ingredients required for successfully tiering data from ONTAP to StorageGRID and setting them up according to best practices.
As a technical marketing engineer, I do have a soft spot for a well-written technical guide. Not having to read a technical guide in the first place – that is priceless (my imitation of a Mastercard tag line is purely coincidental). The upshot is that by running the wizard, you get a StorageGRID high-availability group, a load balancer endpoint, a tenant, a bucket, and a bucket-specific EC 2+1 ILM rule (best practice for FabricPool data) added to your active policy and a Traffic Classification policy for said bucket.
I saved the best for last. The wizard creates a text file with all the information required to configure ONTAP. This capability makes setting up ONTAP to tier to StorageGRID as simple as cut and paste from the text file into the ONTAP System Manager. In addition to the FabricPool wizard, there is a separate but similar wizard for other S3 applications (short overview video here).
Disaster recovery: Cross-Grid Replication and account cloning
Cross-Grid Replication is the marquee feature of this release. In a nutshell, Cross-Grid Replication (CGR) does what the name says – it replicates objects from one grid to another, physically separate, grid. In a previous blog post, I explained how the geo-distributed nature of StorageGRID, along with its powerful ILM policies, allows grid administrators to protect data across data centers while operating in a single grid. However, there are also situations when it's necessary to have a copy of the data in a separate grid instance. Prior to StorageGRID 11.7, that use case was covered by the CloudMirror replication service. Although CloudMirror could replicate objects, it was not suitable for a disaster recovery (DR) solution.
Cross-Grid Replication is vastly different from CloudMirror, and it creates a robust and scalable foundation for the future. StorageGRID 11.7 is the first step on this journey, and a solid step at that. For example, CGR doesn't just replicate object data; it can also clone tenant information. CGR can support active-active environments, meaning that the replication between buckets can be bidirectional. CGR relies on object versioning, which means that an object lifecycle (created, modified, and deleted) is tracked as object versions. Normal GET operations retrieve the latest version of the object, but each version is placed in the replication queue. The replication queue cannot be viewed or managed. However, it is large enough to allow temporary loss of connectivity between grids, and it can support multiple replication relationships. The replication relationship is set up at the bucket level as a one-to-one relationship – because we are not that kind of bucket. The main restriction is that the source and target buckets must have the same name and must be empty at the time of enabling CGR. The relationship between two buckets can be unidirectional or bidirectional. A grid can establish relationships with ten other grids, but a tenant can use only one CGR connection. A short video overview of the feature can be found here.
SGF6112: The next-generation all-flash appliance
Now let’s focus on the newest addition to our platform lineup: the SGF6112. As the name suggests (F for flash), it is an all-flash unit, and it can be configured with your choice of 1.9 TB, 3.8 TB, or 15.3 TB SED or non-SED drives. This unit addresses several use cases – first and foremost, the growing customer demand for all-flash object storage, driven by newer use cases such as artificial intelligence and machine learning, analytics workloads, and data lake architectures.
As an all-flash platform, the SGF6112 hits a sweet spot for workloads with small object ingest profiles. The main difference between the SGF6112 and the previous-generation all-flash appliance is the fact that the new appliance no longer leverages the EF disks and controllers. The corollary is that StorageGRID software is now in charge of local disk protection – and that is a major and intriguing departure from earlier versions. The SGF6112 uses a software-based RAID 5 (2 x 5+1) configuration for node-level data redundancy. This streamlined architecture – NVMe-based SSD drives with local software RAID in a 1U form – works really well in terms of performance, density, and price point.
Object Lock: Governance mode
S3 Object Lock now has a Governance mode to complement the existing Compliance mode. S3 clients can optionally set the following retention settings:
· Retention mode. Either Compliance or Governance.
· Retain-until-date. If an object version’s retain-until-date is in the future, the object can be retrieved, but it cannot be deleted.
· Legal hold. Applying a legal hold to an object version immediately locks that object. A legal hold has no expiration date and it is independent of the retain-until-date.
In Governance mode, users with special permissions can delete an object before its retain-until-date is reached or remove the retain-until-date setting altogether.
Security enhancements
StorageGRID software upgrades are digitally signed to prevent code tampering. Customers can optionally use FIPS-ready OpenSSL 3.0 on all load balancer endpoints, including SG100/1000 and software-defined gateways. The firewall controls are now more granular, enabling customers to restrict access to internal grid ports to grid source IP addresses only, and optionally a list of other trusted IP addresses.
Other improvements and creature comforts
There are several other improvements and creature comforts that I would be remiss not to mention. On new installations, a default site-level storage pool is created automatically. We have also streamlined the ILM module (short overview videos here and here).
Tenants can now be mapped to a single load balancer endpoint, which means that you can potentially restrict data access for a specific tenant to a single IP address and port.
New in StorageGRID 11.7, S3 Select supports Parquet files in addition to CSV files. Also, the performance of the S3 Select operation is significantly boosted. Offloading AI and ML queries to the grid hardware has never been easier. Also new in this release, S3 Select offloading is supported on both StorageGRID appliances and software-defined nodes.
Take the next step
The StorageGRID 11.7 release is a major milestone for us and a significant step forward in several new directions. I encourage everyone to download StorageGRID 11.7 and enjoy all the new enhancements we have worked so hard to make available. The software will be available for download soon!