New Requirements for Mainframe Storage

By Hubert Yoshida posted 02-12-2020 00:52


When it comes to processing massive transactions per second, nothing beats a Mainframe. According to a recent SHARE (IBM user group) blog:


  • Mainframes run 30 billion transactions per day, hold 80 percent of the world’s business data and handle 90 percent of all credit card transactions.
  • Mainframes host more transactions daily than Google (1.3 million/second on CICS vs. 68,542/second on Google), including 55 percent of all enterprise transactions.


Mainframes are powered by IBM Power microprocessors which compete directly with Intel’s highest performance Xeon Scalable Processors but offer up to twice the performance per core and nearly twice the memory bandwidth performance. Mainframes are designed for data intensive workloads and transaction processing. Banks depend on mainframes for instant response for ATM operations, credit card usage, management of investment capital, fraud detection and much more. Mainframes, with the highest clock speed of 5.5 GHz, support the time-critical needs of banks with the highest availability, reliability, and security.


When you consider that each transaction will involve multiple reads and writes to storage, you realize that you need a “mainframe class” storage system with the capability to match the I/O performance demands of mainframe processors and keep them saturated with data so they can operate at peak efficiency. You don’t want your high powered mainframe idling while you wait for your storage to catchup.


Mainframe storage systems have some unique requirements. First, they require special FICON, Fibre channel connections. FICON (Fibre Channel Connection) is an upper-level protocol supported by mainframe servers and attached enterprise-class storage controllers that utilize Fibre Channel as the underlying transport. Mainframes are built to provide a robust and resilient IT infrastructure, and FICON is a key element of their ability to meet the increasing demands placed on reliable and efficient access to data. Mainframe I/O architecture works in conjunction with the FICON protocol to allow an application to pass a series of commands (possibly thousands) that include a mixture of read and write operations to the HBA with a single invocation from the host. In addition, FICON allows a host to access non-contiguous portions of a disk in a single I/O operation. These characteristics of FICON improve application work efficiency while demanding high storage performance. Support for FICON is only available from IBM, Dell, and Hitachi Vantara Enterprise storage arrays.


Secondly, you need a scalable high performance storage array, one that can support millions of IOPs with latencies in the microsecond range. The single most important metric of any storage system is latency.  Latency is the amount of time it takes to service a read or write request from a storage array and directly impacts the performance of the workloads requiring that data. A major new enhancement to reducing latency is the use of NVMe with SSD devices. While SSDs removed the mechanical latencies associated with disk drives, up to now, connectivity to the storage controller was still done through the SCSI protocol which is a serial protocol that only talks to one device at a time. NVMe is a brand new approach that lets you have up to 64,000 simultaneous outstanding requests, so you can get a lot more devices on the back end and reduce the software overhead between applications and storage in flash systems. While many storage systems have implemented NVMe, the full potential of NVMe cannot be realized by just replacing SCSI with NVMe. The higher performance of NVMe will expose other bottlenecks in the I/O system unless they are addressed as well.



The performance leader in enterprise storage systems is the Hitachi VSP 5000 series storage array that has been rearchitected to leverage the technology advances in NVMe. The Hitachi VSP 5000 series delivers up to 21 million IOPs, twice as many as any competitive enterprise systems with best in class latency of 70 microseconds. When VSP 5000 was announced with 21 million IOPs last year, some critics called this over kill. However, when one considers the growing demand for transaction processing, AI/ML, IoT, and future applications of distributed ledgers in mainframe applications, 21 million IOPs will be the new table stakes for mainframe storage.


Hitachi has redesigned the VSP 5000 Series storage controller for NVMe and NVMe-oF starting with the Storage Virtualization Operating System RF (Resilient Flash) where we off load some of the software into FPGAs. Hitachi also introduced a new high performance internal switch fabric which is implemented with PCIe. This patented Hitachi Accelerated Fabric allows the Hitachi Storage Virtualization Operating System RF to offload I/O traffic between blocks. It uses an architecture that provides immediate processing power without wait time or interruption to maximize I/O throughput. As a result, your applications suffer no latency increases since access to data is accelerated between nodes even when you scale your system out. Hitachi also redesigned the shared memory and data cache to streamline the movement of data through the controller for increased performance and resiliency. This not only improves the performance of NVMe flash but also the performance of lower cost SAS flash devices, future SCM devices and other data services, such as data reduction, automation and metro-clustering, which are available with the VSP 5000 series. 


The third requirement for mainframe storage is for the highest level of Reliability, Availability, and Serviceability that matches or exceeds that of the latest mainframe systems.  The IBM Z mainframe architecture has built-in self-detection, error correction and redundancy to eliminate single points of failure, delivering the best reliability of any enterprise system in the industry, along with 5x9s (99.999%) availability. The Hitachi VSP 5000 Series Storage System exceeds the availability of the mainframe, providing 8x9s (999.999999%) availability with a 100% data availability guarantee and supports IBM mainframe storage features like GDPS and HyperSwap for zero RPO and RTO recovery as well as unique Hitachi Universal Replicator Asynchronous Replication delivering long distance protection with the lowest RPO. While IBM also provides mainframe storage, like the latest IBM DS8900F, this storage system delivers 6x9s availability and is limited to two active/active controllers with 3.45 million IOPS, compared to the VSP which can scale to 12 controllers with 21 million IOPS.


For more information on the mainframe advantages of the Hitachi VSP 5000 series storage systems see the following data sheet.


Now is the time to consolidate your advantage with the next generation of data storage for mainframe workloads. Visit the Hitachi Vantara booth 321 at SHARE Fort Worth 2020 from February 23 to February 28 in Ft. Worth, TX and find out more about the new VSP 5000 Series for mainframes.

1 comment



04-25-2022 08:27

Great Write-up