In 2020, Dell created an entirely new storage paradigm with the launch of PowerStore, our most innovative platform in years. PowerStore is built from the ground up on the latest technologies, but it also leverages decades of storage development expertise across multiple product lines – and decades of first-hand customer knowledge based on our ubiquitous end-to-end data center presence. Data-centric, Intelligent and Adaptable, PowerStore is a complete game-changer that helps businesses innovate and thrive in an unpredictable and rapidly changing world. PowerStoreOS version 2.0 was released in April 2021 and improves the previous release with performance improvements and additional. New to PowerStoreOS 2.0:

Today we are officially releasing The PowerStore OS 2.1 software update which brings more enterprise innovation and improvements to the platform. PowerStoreOS 2.1 improves the previous release with even performance improvements and additional functionality to support current and future workloads. New to PowerStoreOS 2.1:

  • Portfolio-wide performance boost compared with the PowerStore 2.0 release
    • More IOPS for every PowerStore model – up to 34% at comparable response times (transactional latency)
  • Support for NVMe over Ethernet fabrics with NVMe/TCP and SmartFabric Storage Software (SFSS)
    • Provides advanced manageability and automation with infrastructure costs almost 25% lower than Fibre Channel
    • Performance up to 45% higher than iSCSI
    • Supported with SFSS in Dell and non-Dell switch environments
    • Utilizes existing 25Gb and 10Gb embedded and IO module ethernet ports
    • Support for VMware vSphere 7.0U3 (as soon as re-released by VMware) and SUSE Linux Enterprise Server (SLES)
  • Additional new features with PowerStoreOS 2.1
    • Improved AI for performance load balancing
    • Windows 2022 support for agentless import
    • Volume application tags
    • Customizable login banner
    • SSL Certificate import for management

NVMe/TCP


In Oct 2021, Dell announced that it was “ALL IN” on NVMe/TCP (NVMe over TCP/IP)
for Dell storage products. Not by chance, VMware had a similar announcement stating support for it. End-users continue to demand high application performance to perform their tasks in today’s rapidly changing business environment. Organizations initially addressed this issue by implementing NVMe SSD-based storage to mitigate storage performance as a potential bottleneck. However, scaling became an issue, as these performance gains required NVMe to be implemented as direct-attached storage (DAS) for every server host. Because NVMe acted as DAS, organizations potentially had unused storage that could not be easily shared by multiple hosts. PowerStoreOS 2.0 extended NVMe benefits across the network with NVMe-over-Fibre Channel support. PowerStoreOS 2.1 now extended NVMe benefits across the network with NVMe-over-TCP support. NVMe over TCP is much more efficient, parallel and scalable than SCSI. It’s designed to make an external networked array feel like direct attached storage to hosts. Starting in PowerStoreOS 2.1, ethernet interfaces can be used for iSCSI or NVMe/TCP host connectivity on PowerStore T model appliances. Ethernet interface creation is deployed in mirrored pairs to both PowerStore nodes since these interfaces do not fail over. This configuration ensures that the host has continuous access to block-level storage resources if one node becomes unavailable. NVMe/TCP is supported on 25 GbE (4-port) I/O module for front end host access 25 GbE speeds. For PowerStore T model appliances, storage networks are configured after the cluster is created. For a robust HA environment, create additional interfaces on other ports after the cluster has been created.

This is a picture that a lot of you are familiar with. We have an application that is running on a host that is accessing storage that is running on an external storage array with either Fibre Channel (FC) or iSCSI. With the introduction of NVMe drives, we primarily saw them first used on the host for caching and boot devices. We then wanted to improve performance on the storage array by adding NVMe drives to the storage array side. However, as you can see from this image, the NVMe-SSDs were still being accessed by the host through Fibre Channel and iSCSI through the SCSI stack. This adds some latency to the data path. That brings to where we are today, with NVMe-oF, users now have an end-to-end NVMe model that can run over Ethernet or Fibre Channel with low latency. NVM Subsystem is defined by the NVMe over Fabrics specification, and it is essentially any system that can serve NVMe storage over a fabric. Any storage system that is designed to handle NVMe/TCP or NVMe/FC traffic is also considered an NVM Subsystem. There is one NVM Subsystem defined by one NQN for the whole Cluster – this is true even for Multi-Appliance Clusters. PowerStore supports NVMe/TCP over all front end ethernet ports. It is supported on any port that supports iSCSI. The “Purpose” field was added to storage networks to define their support for either iSCSI and/or NVMe/TCP. Storage Networks can support both iSCSI and NVMe/TCP at the same time, or they can support only one or the other. The user selects this option using the Purpose option. When upgrading to PowerStoreOS 2.1, all existing Storage Networks will automatically be selected with NVMe/TCP for the Purpose. If the user does not want these networks to support NVMe/TCP targets, they can be removed after the upgrade using the Purpose option



Here you can see the Purposes options for iSCSI and NVMe/TCP A Volume is not created as “SCSI” or “NVMe”. When a Volume is created, all the unique identifiers for SCSI and NVMe are created with it. Once the Volume is mapped to a host, then it is defined as an NVMe attached Volume or a SCSI attached Volume. Hosts can only be attached to NVMe OR to SCSI at the same time. You cannot have a Volume mapped to NVMe and SCSI hosts at the same time. If you map it to an NVMe host, you can unmap from all the NVMe hosts and then remap to a SCSI

host later if you wish. This can also be done in the opposite direction NVM subsystem can be a drive (SSD), could be an entire storage array Controller associated with 1 host NQN and provides access to an nvme namespace. In NVMe SSD the controller is associated with PCIe function, in NVMe-oF it’s a software driver that performs command processing. Relates to a specific port. Multiple controllers are supported Namespace is a volume. NSID in shared namespaces must be unique within subsystem Shared namespace is a namespace that can be accessed through two or more controllers (either same host or different hosts). Private namespace can be accessed through only a single controller. PowerStore doesn’t use private namespace, all volumes are shared namespaces and therefor support multipathing. Controller dynamically created; controller created when host connects to PowerStore

  • All common NVMe-oF Features applicable for NVMe/TCP
    • Support Multipathing capability
    • Support Asymmetric Namespace Accesses (ANA)
      • Equivalent to SCSI ALUA
    • Support NVMe Reservations (Host Cluster supported)
      • Equivalent to SCSI-3 Reservations
    • Support Offload commands
      • Dataset Management (deallocation, similar to trim)
      • Write Zeros
    • Support Fused operation – Compare and Write
    • Support Asynchronous Event Notification
  • Exceptions
    • Xcopy or equivalent is not standardized in NVMe-oF right now and not supported
  • The following hosts are supported for NVMe/TCP:
    • RHEL 8.2-8.4
    • SLES 15 SP2-SP3
  • Protection Policies supported on NVMe attached Volumes
    • Snapshots and Replication fully supported
  • Replication between Clusters for NVMe attached Volumes is still iSCSI
  • Migration supported for NVMe attached Volumes
    • PowerStore Manager will prompt user to rescan hosts since AEN/AER are optional
  • No vVol support for NVMe/TCP

You can read more about PowerStore NVMe/TCP integration with VMware vSphere 7.0U3 in this previous post Below you can see a demo showing how it all works with VMware vSphere 7.0U3: 

SmartFabric Storage Software

SmartFabric Storage Software (SFSS) is an End-to-End NVMe IP SAN solution from Dell Technologies. In partnership with VMware, Dell Technologies has engineered a complete end-to-end NVMe/TCP solution to help provide enterprise organizations with the agility to stay ahead of growing workload demands with modern, automated, and secure storage connectivity both today and as they migrate to extended hybrid clouds. The Dell Technologies SmartFabric Storage Software (SFSS) product is a standards-based centralized discovery controller for NVMe/TCP hosts (servers) and subsystems (storage). In this initial release, SFSS is a stand-alone software solution packaged as a containerized application enabling an end-to-end automated and integrated NVMe-oF solution running TCP over an Ethernet fabric. The key functions of SFSS are policy-driven to help automate NVMe-oF storage service discovery, end-point registration, connectivity, and zoning. For more information visit https://www.delltechnologies.com/en-us/storage/smartfabric-storage.htm

There are two different ways to connect and use NVMe/TCP. There is the Direct Discovery option and the Centralized Discovery option. With the Direct Discovery option, it is more suited for smaller types of deployments where you may only have a few hosts and a single target. From the host, you need to perform a discovery to each of the subsystems (targets) which will return a list of subsystems that the host connects to. This is fine for a small environment but once there are many hosts and targets, this can be error prone and difficult to scale. That is where a Centralized Discovery option comes into play. The Centralized Discovery Controller (CDC) is something that advertises itself on the fabric such that NVMe/TCP endpoints such as hosts and subsystems can log into it. This Centralized Discovery Controller will be able to keep track of every device that logs into it by storing the entries inside a database such as a “nameserver”. Users will then be able to have access control through methods like zoning to provide access in between hosts and subsystems. SmartFabric Storage Software (SFSS) has many aspects that are like Fibre Channel. Before we dive into the similarities, let’s discuss one thing that is different which is the Discovery Service. The Discovery Service enables NVMe/TCP endpoints to dynamically discover the CDC (if supported). The CDC / SFSS supports mDNS queries from end points in this first release. However, we need to wait for these end points to broadcast themselves with mDNS which will come in future releases. Now let’s talk about some of the more Fibre Channel-like similarities: End Point Registration Service SFSS has the capability of having an End Point Registration Service. This allows NVMe/TCP end points such as hosts or subsystems to register with the CDC. This is very similar to a host in Fibre Channel logging in with a Fibre Channel switch and populating the host WWN with the FC switch name server database. In the case of SFSS, we have a NVMe/TCP host that has a host NQN that needs to register and populate the name server. We also have NVMe/TCP subsystems that also need to register and populate their information into the name server as well. End Point Query Service Once the NVMe/TCP hosts and subsystems are logged in to the name server database, each of these endpoints can communicate with the CDC to figure out who they can communicate with. For example, a host will query the CDC to find information about which subsystems they can communicate with. Similarly, a subsystem will also query the CDC to find information about which NVMe/TCP hosts it can communicate with. Zone Service If we want to provide access control and zoning, CDCs and SFSS support this capability. We can group hosts and subsystems into zones via soft zoning. This is very similar to how Fibre Channel does their zoning today as well. Asynchronous Notifications CDCs and SFSS have a capability of supporting AERs and AENs if there are any changes on the fabric. This is very similar to a Fibre Channel Registered State Change Notification (RSCN) We’ve talked a lot about NVMe/TCP end points. Let’s now talk about which end points are supported with SFSS 1.0.0. Here on the screen, you can see that from the host side, we are supporting VMware 7.0U3 which introduces support for NVMe/TCP. This release from VMware will allow us to discover a SFSS that is broadcasting itself on the fabric. We can discover and register to the CDC with what we call a “Push Registration”. This Push Registration mechanism will be covered later in the presentation. If there isn’t a CDC or SFSS that is deployed on the fabric, a VMware 7.0U3 host also supports the capability to directly connect to a subsystem as well. Next on the subsystem side, we are supporting PowerStore OS 2.1. This release introduces NVMe/TCP for the first time. We support this PowerStore array discovering and registering to a SFSS with what we call a “Pull Registration”. Both Push and Pull registrations methods will be covered later in this presentation. Next, both PowerStore T and PowerStore X model appliances are capable subsystems, but we only support block volumes in this release.

DC Support

The PowerStore 500 platform offers either AC- or DC-powered power supply units (PSUs). The base enclosure comes with two PSUs, one per node, and the PSU models must match between nodes. The 1450 W PSU supports high-line and low-line AC power as well as C13/C20 or C13/C14 connectors. Figure 2 shows the supported connector types. The 1100 W PSU supports -48V DC power using a Positronic connector and requires PowerStoreOS 2.1 or greater. Either AC or DC power is selected at the time of ordering and a PowerStore 500 system cannot be converted from using AC to DC or from DC to AC PSUs. A

single power supply can power the entire base enclosure during a power-supply failure. You can replace the power supplies without removing the node.

Performance Improvements

Latency and IOPS improvements across the board, here are some examples:

Customizable Login Message


A Login Message provides the ability for storage administrators to create, enable, and disable a customizable login message starting in PowerStoreOS 2.1. The Login Message is displayed whenever a user navigates to the PowerStore Manager login page from their browser. Login Messages can be used for a wide variety of use cases, such as informing the user what system they are logged into or even provide the user with a security warning. The Login message can easily be configured from PowerStore Manager via Settings > Security > Login Message by enabling the message, typing in your text up to 2000 characters, and then clicking the. Once the message is enabled, it will be seen by all users from the PowerStore Manager login page. Login Messages can also be enabled and configured through REST API and PowerStoreCLI. Login Messages can only be edited, enabled, and disabled by users with the Storage Administrator or Security Administrator roles.

Application Tags


PowerStoreOS 2.1 introduces Volume Application Tags. A Volume Application Tag is specified during volume creation, allowing users to label their volumes with a specific category and application type based on the use case for those volumes. Users can then use application-centric management to view and sort through their volumes by application type by adding the new “Application” column in the list view. There are six different predefined categories, each with several different application types. If the application type is not listed, the user can then manually type in their own application name. Finally, if the five categories do not match up with the application you are looking for, you can select the “Other” category and optionally type in the application type, up to 32 characters in length.

  • Relational Databases – Oracle, SQL Server, PostgresSQL, MySQL, IBM DB2
  • Big Data & Analytics – MongoDB, Cassandra, SAP HANA, Spark, Splunk, ElasticSearch
  • Business Applications – ERP / SAP, CRM, Exchange, Sharepoint
  • Healthcare – EPIC, Meditech, Allscripts, Cerner
  • Virtualization – Virtual Servers (VSI), Containers/Kubernetes, Virtual Desktop (VDI)
  • Other – User specified application type


Application tags can be changed after a volume is created by modifying the volume properties and specifying a different application category and type. Any existing volumes created prior to PowerStoreOS 2.1 will not have any Application Tags configured. After upgrading to PowerStoreOS 2.1, the user can then manually apply Application Tags to existing volumes. Once an Application Tag is applied to a volume, it cannot be removed, however it can be modified as needed. Application Tags can also be set and managed via REST API or PowerStore CLI (pstcli). Designed with interoperability in mind, Application Tags are applied at the volume level and cannot be applied to the Volume Group itself. However, Volume Groups support member volumes with Application Tags. If there are multiple member volumes with different Application Tags, then the Volume Group will show as a “Mixed” application category. Application Tags have also been designed with backwards compatibility in mind. When creating a new Volume through REST API or pstcli, the Application Tag fields are not a required field. Where as in PowerStore Manager, Application Tags must be specified to continue the Create Volume Wizard. This is to prevent issues with pre-existing automation scripts or applications after upgrading to PowerStoreOS 2.1. The following considerations should be made when applying and managing Application Tags

  • Cloning Volumes – Clones of a Volume with an application tag will inherit the tag of the parent volume
  • Snapshots of Volumes – Snapshots do not inherit Application Tags of the source volume therefore operations like refresh and restore do not affect the application tag of the source volume
  • Internal Migration of Volumes – The Application Tag will be migrated and retain the same value as the source volume
  • Importing Volumes – Imported Volumes will not have an Application Tag set and can be modified post-import to add an application tag
  • Replication – The Application Tag will always be replicated to the destination volume. It is recommended to pause the associated replication session prior to modifying the source volume Application Tag and then resume the session to allow the new tag to sync to the destination volume

Thin Packages / Upgrades

From the Settings menu > Cluster > Upgrades page, users can manage, upload, and deploy various Non-Disruptive Upgrade (NDU) packages for the PowerStore cluster. Generally, NDU packages consist of two categories: Software Releases and Thin Packages. Software Releases are PowerStoreOS upgrades which contain the full OS image or patch/hotfix images for a specific OS version. Thin Packages contain a smaller and more targeted amount of functionality than regular PowerStoreOS packages. These may also consist of different package types such as Disk Firmware or Health Check updates. Thin Packages are generally smaller in size, take less time to apply, and in many cases do not require node reboots. In PowerStoreOS 2.1 users must run the prerequisite UPGRADE EXTENSIONS System Check within 60 minutes prior to executing the UPGRADE button. If the user has not run the UPGRADE EXTENSIONS System check or the 60-minute window has expired, the UPGRADE button will be greyed out with a tooltip providing instructions to run the UPGRADE EXTENSIONS from Monitoring > System Checks.
Telemetry notices Notification displayed after the EULA which provides information about the Dell Telemetry collector as well as privacy policy information.

Download

You can download the 2.1 version, by clicking the screenshot below, or, just click the ‘update’ button



Additional resources You can view the release notes by clicking the screenshot below

Leave a Reply