May 13, 2014


Hitachi Data Systems announced a new global active device capability within the new Storage Virtualization Operating System (SVOS) and the new Virtual Storage Platform G1000 (VSP G1000) storage system.  This enables a virtual device to span two VSP G1000 storage systems that are separated by up to 100 KM apart for load balancing and active/active availability. This enables an application from either site A or site B to access a shared device through either VSP G1000, and continue operations if one of them should happen to become unavailable. A cluster application like Oracle RAC could access the device through either site as well as through either G1000 for greater availability.


A global active device requires the caches in both G1000s to be in synch. That requires a write to one G1000’s cache, to be retransmitted to the other G1000’s cache. The second G1000 must acknowledge the write back to the first G1000 before a write complete is sent back to the host server that initiated the write.  The overhead associated with the write updates, are minimal since they are done directly from one G1000 to the other G1000.

The primary way competitors try to provide similar global active device (active/active) capability is to do this with appliances that sit in front of the storage systems. In the slide below, you can see the difference between the Hitachi and competitive approaches to active/active. The competitive approach, through the use of appliances, adds overhead. A write to an appliance in site A must be replicated to the other appliance at site B. The site B appliance then writes the data to the cache in the site B storage system. The site B appliance must wait for acknowledgement from the site B storage system before sending an acknowledgement to the appliance at site A. The site A appliance must write the data to the cache in the site A storage system and wait for acknowledgement from the site A storage system and the site B appliance before it can return a write complete to the host server that initiated the write. This additional overhead amounts to an appliance tax for every write. There is also an appliance tax for every read since the reads must go through the appliance before it goes to the storage systems.


The use of appliances also adds a connectivity tax in terms of the cost and complexity in the connection between the host servers, the appliances, and the external storage systems. Global active devices need to be seen through all the nodes in the appliance cluster. That means each node in the appliance cluster needs an in and out port to its local host server and storage system  (2 ports) with alternate path redundancy (4 ports) and redundant connection to the other host in the cluster (+2 paths) as well as redundant connection to the other appliance in the cluster (+2 paths). Additional ports may be required for high performance applications. In addition to the multiplication of ports caused by the insertion of an appliance, there is locking and management overhead associated with the web of paths that connect all these ports. If the connection is through a FC SAN you have the additional cost of FC switch ports and cables and you now have a SAN zone between the servers and the appliances, a SAN zone between the appliances and the storage systems, and a SAN zone between the two appliances that need to be monitored and managed for buffer credits, LIP storms, state change notifications, etc. This additional complexity adds to the probability of application failures.

The complexity of inserting an appliance is further compounded since the operating systems in the appliance and the storage systems are different. The Hitachi approach is not to introduce a separate appliance but enable SVOS to provide the global active device functionality and also provide other storage virtualization functions within the G1000, including:

  • Universal Volume Manager enables virtualization of external heterogeneous storage.
  • Dynamic Provisioning provides thin provisioning for simplified provisioning operations, automatic performance optimization and storage space savings.
  • Device Manager provides single point management for all Hitachi physical and virtualized storage, and it acts as the interface for integration with other management systems.
  • Resource Partition Manager supports secure administrative partitions for multi-tenancy requirements
  • Cache Partitioning supports up to 32 cache partitions.
  • Dynamic Link Manager Advanced provides advanced SAN multipathing with centralized management.
  • Performance monitor feature provides an intuitive, graphical interface to assist with performance configuration planning, workload balancing, and analyzing and optimizing storage system performance.
  • Storage system-based utilities including: LUN Manager, Virtual LVI (Customized Volume Size), Data Retention Utility, Volume Port Security and Volume Security Port Options, Cache Residency Manager, Server Priority Manager, Audit Log, Command Control Interface (CCI), RAIDCOM, and Volume Shredder.
  • Standard management interface support including SMI-S provider, SNMP agent, Volume Shadow Copy Service (VSS) and REST

In addition to one integrated operating system, there is one integrated management system through Hitachi Command Suite v8. This enables the global active device functionality to be dynamically provisioned, monitored, and thin provisioned, snapped, cloned, and replicated just like any other device. With a separate appliance, separate management may be required for different functions. Additional appliances may even be required for functions like replication.

From a development perspective, the implementation of a global active device function through the use of appliances may be easy, but the integration of this function into an enterprise storage system requires more innovation in order to leverage the performance, availability, and functionality of that system. Putting a high performance enterprise storage system behind an appliance can only dumb down the storage system and add additional layers of complexity and cost.

The availability of Virtual Storage Machines and global active device functionality will be game changing for high availability applications.