May 19, 2014
As servers increase their number of cores and memory performance, VMware is able to scale up additional virtual machines and drive more applications almost indefinitely. Unfortunately, each virtual machine adds additional random I/O workload, and the storage systems become a bottleneck. While VMware has tools like vMotion and Storage vMotion to rebalance the workload and SRM (Site Recovery Manager) to recover from site failures, the movement of data between storage systems and the reconnection of servers and application to servers can be very disruptive and unpredictable in large VMware configurations. In many cases, this is because the storage technologies have not advanced at the same pace as server technologies, or have different implementation types and scalability and processing requirements.
Wouldn’t it be fantastic if the storage industry had a storage hypervisor that could virtualize storage devices, scale up additional data stores almost indefinitely, and had tools to rebalance workloads, including across datacenters, and still manage site failures and movement of data? Wouldn’t it be even more amazing if this type of storage virtualization would seamlessly integrate with VMware in a way that not only removed the bottlenecks but also increased the number of virtual machines that could be run on VMware hosts?
These problems are solved with the Hitachi VSP G1000 and Hitachi Storage Virtualization Operating System (SVOS)
First in terms of performance, the G1000 with the Hitachi deigned flash modules can scale up to 3.9M random read IOPs. This is 3 times more than any flash storage system available today. It is important to realize that VMware is a consolidation play that requires a scale up storage performance solution rather than a scale out solution that is designed for distributed compute nodes. You may not need this many IOPs today, in which case you can start with a small configuration of the G1000 and non disruptively add storage processors, cache, and ports as your storage demands increase over the next 5 to 7 years. The G1000 will scale with your future VMware growth.
Second, the G1000 with SVOS provides a new capability that supports vMotion and vStorage and eliminates the performance issues if physical volumes must be moved for data mobility. vMotion and Storage vMotion can shift apps and data to less busy systems to balance workloads or for failures. While this can be manually invoked or automated by VMware Distributed Resource Scheduler (DRS) and Storage DRS, the movement of the data takes time and the applications, while remaining online, can experience dramatically increased I/O delays, which can be impactful to today’s need for realtime access. Additionally, today’s always-on applications and limited resources often demand active/active data centers that are equally accessed and highly utilized. While VMware does a fantastic job with being able to vMotion the VM and active memory, migrating the storage data (easily 10x the size) is much more challenging, and almost impossible without application I/O delays and risk of timeouts. While VMware provides Metro Storage Cluster (vMSC) certification to help offload the storage transport to storage array replication, the implementation often requires severe restrictions to DRS policies to help keep create VM “affinity” within physical sites to avoid large numbers of these site migration delays.
SVOS provides a Global Active Device feature on the G1000, which provides an active-active stretch cluster (up to 100KM) between servers. Think of this as a volume that stretches across the G1000s located at each server site. Multi-path software allows access to data from the shortest path. No data needs to be migrated and there is no application interruption or administrative action required. vMotion only moves the VMs (compute) and the applications can access the storage global active device from either site equally. The overhead and delays for storage vMotion data movement are eliminated and the service level is continuous from either location.
Keep in mind that this is over and above the full VMware integration and offload support of Hitachi SVOS with the full suite of vStorage API’s (VAAI, etc.) Also, it’s important to mention that with storage virtualization at the core of Hitachi SVOS, all of these features can be extended to any storage device virtualized by the G1000, allowing an extended life to your existing storage capacity and assets.
Third, this same Global Active Device feature can provide non-disruptive operations in the event of infrastructure problems (power, services, etc.) that cause access failure to primary storage. These problems require failover to remote copies of the data stores at a remote site through VMware Metro Storage Cluster (vMSC) configurations or VMware SRM software.
This failover causes interruptions, which cannot be tolerated by SLAs in mission critical environments. In addition to active/active access, the global active device feature allows automatic site switch over to the available volume up to metro distances (100KM) delivering the highest level of continuous operations with no SRM invocation or application interruption.
The configuration of global devices, monitoring, and SLA management is done through our Hitachi Command Suite. Hitachi Command Suite 8 also provides VMware vCenter Operations Suite (vCOPS) integration capability for real time health and monitoring of VM’s and storage. Hitachi has enabled taking VMware to the next level by taking storage to the next level with SVOS in the VSP G1000 and Global Active Devices. Finally, a storage hypervisor that can virtualize storage devices, scale up additional Datastores almost indefinitely and rebalance workloads, including across datacenters, while still protecting from site failures and seamlessly managing movement of data.
A converged solution with G1000, SVOS, blade servers, networks and VMware, preconfigured and pre-validated, and managed through vCenter is available as part of the Hitachi Unified Compute Platform with UCP Director 3.5. UCP Director also includes advanced automation, orchestration, and operations with VMware vCOPs, vCO, and SRM. A converged solution with Oracle RAC and global active device is also available for high performance, high availability Oracle environments.
The VSP G1000 with SVOS is available now. The global active device feature will be generally available later this year.