Lascon Storage was founded in 2000 and provides hints and tips on how to manage your data, strategic advice and news items. A lot of this advice is about smarter, cheaper ways of working without compromising service.
Use the links above to enter the different sections of the site, or just select from the highlights further down the screen.

Storage Vendor News

IBM Storage

Hu Yoshidas blog

Provide “True” Enterprise-Class Data Protection for Your Apps The Advantages of Hitachi Global-Active Device
Tue, 11 Dec 2018

GAD Brains.png

The days when data recovery solutions were evaluated on how they could minimize the two R’s are over. The two R’s stand for Recovery Point Objective, RPO, how much new or changed data is lost because it hasn’t been backup yet, and Recovery Time Objective, RTO, how long it takes to resume operations. The reason I say this is because we can achieve zero RPO and zero RTO with today’s Global-Active Device on our family of Hybrid G Series and All Flash F Series Virtual Storage Platform, VSP, storage arrays.


The Global-Active Device (GAD) is another virtualization capability of the VSP storage platform that creates a virtual storage machine. A virtual storage machine virtualizes two separate VSP storage arrays and makes them appear as one storage array to a host server or a cluster of host servers. GAD is configured so that the primary and secondary storage systems use the actual information of the primary storage system, and the global-active device primary and secondary volumes are assigned the same virtual LDEV number in the virtual storage machine. This enables the host to see the pair volumes as a single volume on a single storage system, and both volumes receive the same data from the host. When a write is done to one of the GAD pair volumes, the data is replicated to the other pair and acknowledged before a write complete is returned to the initiator of the write. That keeps the volumes in synch and ensures zero RPO and zero RTO in the event of a storage system or site failure.


The virtual storage machine can span across physical storage systems that are separated by up to metro distances (500 kilometers). GAD provides non-disruptive, high availability (HA), disaster recovery (DR), and rapid data center migration services. In addition, it enables painless virtual machine storage motion where the location of the storage underlying virtual machines is moved between storage environments for load balancing or maintenance. GAD supports active-active processing of shared data, meaning that all interfaces to the storage system are always active, and the system synchronizes writes across the entire storage system. GAD can be configured in a single or clustered server configuration.


GAD Config.png


Some vendors might use their synchronous replication product to create a synchronized copy pair. This creates an Active/Passive pair on separate storage arrays. Processing occurs on the primary side of the pair and the passive side is only used for fail-over. This does not provide active-active processing of shared data. In order to achieve active-active processing, GAD provides three key features that distinguishes it from synchronous replication products.


  1. The first is a streamlined locking mechanism between the storage systems to synchronize the writes.
  2. The second is preservation of SCSI reserves, a control mechanism for avoiding conflicts or congestion with different host initiators.
  3. The third is a quorum disk to resolve conflicts between the active-active pair of storage systems.


The quorum disk is an external storage system that acts as a heartbeat for the GAD pair, with both storage systems accessing the quorum disk to check on each other. A communication failure between systems results in a series of checks with the quorum disk to identify the problem for the system to be able to receive host updates. The external quorum disk is accessed like a virtualized external disk which means that it can be a third party disk. It can also be on a virtual or physical server that has software to present itself as a disk. The quorum disk can also be in the cloud through the use of iSCSI. However, if GAD is used simply for non-disruptive migration, no quorum disk is needed.


GAD is a copy technology and can be one of multiple overlapping data copy technologies that are used in a production system. Backup and point in time copies are still required to protect against data corruption caused by errors or malicious attacks. While GAD provides protection in a metro area, an asynchronous, out of region replication with Hitachi Universal Replicator (HUR) might also be required in addition to GAD. Copy technologies include synchronous and asynchronous copies, clones, snap shots, and thin images or virtual copies. Copies are not only used for operational and disaster recovery, but also for archive, big data, analytics, audits, e-discovery, development and test, and other data repurposing requirements. IDC estimates that 60% 0f corporate data consists of copies of original data, and that on average, organizations have 13 copies of each data object. Managing the life cycle of copies is a major challenge. Unmanaged or orphan copies of data can be a liability. Hitachi Vantara’s Hitachi Data Instance Director provides a copy data management platform that simplifies creating and managing policy-based workflows that support business functions with controlled copies of data. Integration with Hitachi Data Instance Director is a key advantage of Hitachi Vantara’s GAD.


HDID with GAD.png

Hitachi’s Global Active Device is the leader in active-active, continuous availability. Other storage vendors may claim similar replication capabilities. Since they all rely on synchronous replication over metro distances there is not much difference in performance. However, there is a great deal of difference in simplicity, scalability, and ROI.


  1. Simplicity. Hitachi’s implementation is simple in that it is an extension of the virtualization capability of our Storage Virtualization Operating System RF (SVOS RF). There are no additional appliances required to virtualize the pair of devices or additional software required in the host servers or virtual machines. No BIN files are required to set the configuration. Our virtualization also provides the most choices for selection of a quorum device. Integration with Hitachi Data Instance Director makes it easy to manage the interactions with other data copy technologies and application requirements.


  1. Scalability. The way we have stream-lined the synchronous replication of the GAD pairs and preserved the SCSI reserves provides greater scalability than competitive implementations. Competitive solutions that require the use of appliances scale by adding more appliances and this also adds more complexity. Hitachi GAD provides true active-active processing of shared data, across the storage arrays. This scales beyond other implementations where only one storage array is active while the other is used for standby. The controllers in each VSP storage array are also active-active which gives us greater scalability than other vendor’s controllers that are Active/Passive or ALUA (Asymmetric Logical Unit Access). (see my previous blog on the differences in storage array controllers)


  1. ROI. The return on your investment is higher with GAD due to our overall virtualization capability to virtualize external storage and create virtual storage machines. You can virtualize your existing storage behind a VSP array and create a virtual storage machine with another VSP over metro distances. Neither VSP in this scenario requires any capacity from a Hitachi storage system, all the capacity can be from virtualized 3rdparty storage systems. All VSP systems, from midrange to high-end enterprise support GAD, so you don’t need to have different solutions and different management tools for midrange, entry enterprise or high-end enterprise. No additional appliances or host software is required to support GAD. GAD also integrates well with other copy requirements such as HUR and thin images since it is a feature in SVSO RF. The use of Hitachi Instance director with GAD also reduces the cost of copy management and ensures that resources are not wasted on copies that are no longer needed.


Learn More


For more information on Hitachi Vantara’s GAD implementation

Enterprise Storage Systems in a Midrange package
Tue, 04 Dec 2018

High end enterprise storage systems are designed to scale to large capacities, with a large number of host connections while maintaining high performance and availability. That takes an architecture where a host can access any storage port and have direct access to a number of storage controllers which can service I/O requests and process their data services requirements. This takes a great deal of sophisticated technology and only a few vendors can provide such a high end storage system. Hitachi’s, high end, Hybrid Virtual Storage Platform (VSP) G1500 and All Flash VSP F1500 arrays provide this capability which is made possible through a switch architecture which dynamically connects front end ports with controllers and a global cache with backend device controllers.

VSP Family.png

For the midrange user where cost is a key factor and massive scalability is not required, the architecture has to be changed to trade off scalability for reduced cost. However, that trade off often means that some advanced enterprise functions like distance replication and zero data loss recovery would be compromised. With recent advances in Intel multicore technology and PCIe extensions, it is now possible to scale down these systems to reduce costs while still providing these advanced enterprise functions. The recent models of the Hitachi Vantara VSP Hybrid G and All Flash F series storage arrays, now provide, cost efficient, scaled down versions of the G1500 and F1500 Enterprise storage arrays, without any loss in enterprise functionality. This was done by consolidating the VSP controller, cache memory, and interconnects onto a cost optimized, two controller architecture which utilizes Intel’s latest multicore processors with high speed PCIe interconnects. These two controllers are configured as Active/Active controllers, which means that any I/O can access any device through either controller.


Midrange storage arrays are configured with dual controllers for availability. However, there is a great deal of difference in how the two controllers are configured. Most are configured as Active/Passive or Asymmetric Logic Unit Access Asymmetric (ALUA). Very few are Active/Active. The configuration of the storage controllers is a key differentiator when it comes to the performance and functionality of the storage system. Here are the differences between the different types of controller configurations.

Active Passive.png




Active Active.png


The VSP’s Active/Active configuration is made possible through the concatenation of the cache that is attached to each controller. Both controllers work with the same cache image of the storage LUN. This configuration is also known as Active/Active Symmetric since both controllers can process the I/O request versus the prior ALUA asymmetric configuration above. This requires additional routing information that is provided by the Storage Virtualization Operating System that is available in all models of the VSP series. This Active/Active capability provides many benefits. We don’t need to worry about LUN ownership as in the case of asymmetric controllers. This provides the ability to use VMotion for live migration of running virtual machines between servers that don’t have identical primary failover controller assignments, with zero downtime, continuous service availability, and complete transaction integrity. This also provides the ability to load balance across a SAN without the worry of creating a performance issue on the storage. Since the VSP can virtualize external storage, this also makes it possible to process a cache image of a LUN from an externally attached storage system. This ability to virtualize external storage enables the extension of the useful life of legacy storage systems and the non-disruptive migration of data to and from an external systems as I posted in a previous blog.


The latest versions of our midrange offerings, the Hybrid VSP G350/G370 and All Flash VSP F350/F370 come in a cost effective 2U of rack space with the enterprise features of our high end and over 1 million IOPS for the VSP F370. All priced and packaged for the midrange.


On the higher end, which many refer to as the entry enterprise, we offer the Hybrid VSP G700/G900 and All Flash VSP F700/F900 that can scale to 2.4 million IOPS and comes in 4U of rack space. The differences in the model numbers is based on the number of Intel processors and cores. Both Hybrid and All Flash entry enterprise models have the same dual controller Active/Active design as the midrange models but with many more Intel cores, memory, front end ports, back end ports, and higher internal bandwidths. Here are the specifications for the Entry Enterprise Models.


VSP Data Sheet.png

The Hybrid VSP G1500 and All Flash VSP F1500 are included to show the full scale out capability of their switch architecture. The VSP G/F1500 is a multi-controller architecture in which the processors, memory modules, front end ports and back end ports are all connected through an internal non-blocking, cross bar, switch, which enables a LUN to be accessed directly through any port without the overhead of passing control between controllers.

Cross bar switch.png


There are other architectures which support more than two controllers, but those types of architectures involve a great deal of overhead and complexity. Without an internal switch architecture, no matter how many controllers you hook together, I/O requests for a LUN still have to be processed through the controller that owns the LUN. So If your I/O request comes in from another controller port, the I/O has to be passed to the controller that owns the LUN, creating more overhead and complexity with each additional controller that is added. Having two controllers adds redundancy but adding more controllers to a two controller architecture, can create less availability since the cumulative failure rate increase with each added controller. Having two controllers fail independently is rare, and a two controller failure is usually due to a base failure which would affect all the controllers no matter how many you have.

Four controller.png



In the case of the VSP G/F1500 the switch architecture would allow the controllers, cache, and LUNs to be assigned dynamically and fail independently.


The VSP family of storage arrays, provides a choice of cost optimized configurations from midrange to high end scalable enterprise systems, all running the same software and providing the same functionality to help our customers preserve their investments in policies and procedures and leverage their ecosystem-related investments. Although there is a difference in architectures we are able to simulate the architectural differences in software so that all the models have the same functionality even when it is scaled down to the midrange price and packaging. Our dual controller architectures are fully Active/Active which differentiates us from many other midrange and entry enterprise systems.