The IBM Z is an amazing system comprised of several types of special purpose processors and which scales to extremely powerful capacity. Alebra’s Parallel Data Mover (PDM) is engineered to leverage the strengths and specialties of each of these processors, allowing data to be transferred with maximum speed and efficiency.

Following are the salient types of processors that exist in the IBM Z complex:

Integrated Information Processors (zIIP)

These are special purpose processors that applications can use (with IBM’s concurrence) to offload processing from the GPs. In order to use them, extensive programming changes by the application vendors are required. Usually, only a portion of an applications processing can be offloaded to these processors.

I/O Processors (IOP)

These are used exclusively for executing I/O operations to and from FICON channels. They manage both the initiation of the I/O and the processing of interrupts when the I/O completes. Their purpose is to ensure the maximum performance from attached I/O devices such as storage. The number of these is based on the number of FICON channels that are ordered.

Integrated Facility for Linux Processors (IFL)

These are used solely for running native Linux for Z or running z/VM with Linux for Z guests). They are very powerful engines and each one can do the equivalent work of many Intel Linux processors.

How does PDM use IBM Z Processors?

PDM makes very little use of the GPs. Extensive engineering work is behind its ability to offload virtually all processing to zIIPs. When PDM is used to send data over one of Alebra’s proprietary FICON-attached gateways, then there is no additional overhead because processing used by the operating system to access the data on the non-z system is performed by an IOP. When PDM is used to send data over standard network connections (TCP/IP), then overhead is incurred in the GPs because that’s where TCP/IP executes. Alebra uniquely offers the ability to offload this processing to an IFL, so that there still is no usage of GPs. PDM sends large blocks over HiperSockets to Linux on Z and forwards from there to target nodes, thus eliminating almost all processing by the TCP/IP stack on z/OS.

Because of the above, PDM is uniquely able to move data to and from an IBM Z platform with zero impact to production work executing in GPs.