Data Sharing Software Offloads Mainframe Sorts to CoSORT

At a Glance:
Overview
Introduction
Description
Benchmarks and Benefits
Platform Availability
Licensing and Support


Overview

so you can:

  • Execute your existing JCL parms (sort cards)
  • Use/get the same SORTIN/SORTOUT data/results
  • Speed sort execution and I/O throughput
  • Eliminate costly mainframe sort software and CPU cycles

 
Introduction

In the 1970's, 50% of mainframes were sorting 50% of the time. Today, mainframe sorts consume between 20-40% of mainframe resources after peak work hours, using CPU cycles and rented sort licenses. While it remains a reliable, mission-critical combination, annual 5- and 6-figure price tags make native mainframe sorting unduly expensive.

Current mainframe sort processing is also inefficient in terms of data movement and data availability. In a mix of mainframe data and open systems hardware, your access to data via the mainframe is done as part of an application-specific request, or by data movement (ETL) processes run during off-peak hours that replicate the mainframe data onto open system databases:

Today's unsorted data to (and sorted data from) the mainframe transfer at slow ESCON/FICON rates. Data are needlessly replicated and stored on both sides (redundancy). Mainframe sorts are expensive to rent and upgrade, as are the CPUs needed to run them. Finally, when sorts run on the mainframe, other jobs have to wait (latency). Fortunately, there are alternatives.

The first and most obvious alternative has been migrating or rehosting entire mainframe sort and related data processing operations onto faster and less expensive open systems hardware. This is the traditional server-centric model. CoSORT has provided standalone and ISV- integrated sorting solutions for mainframe migrants to UNIX since 1985. Free MVS and VSE JCL sort parm (and MF COBOL copybook) migration tools come with the CoSORT package, which both replicates and expands on mainframe sort/report tools.

The second and new alternative is to mix both worlds by allowing mainframe sort and related legacy operations to continue while offloading the costliest work items to CoSORT on faster, less expensive, and more available open systems hardware. Data reside centrally, so that both mainframe and open systems users have fast and simultaneous application access to it. This new storage-centric paradigm shift is called data sharing.

 
 
Description

In order to offload your existing mainframe sort steps to open systems, you need a CoSORT license on an available UNIX or Windows server, and Corigin Corp's data sharing software components on your mainframe ("SortWize" program) and open systems platform(s). The data sharing components communicate between the mainframe, central EMC or HDS storage, and CoSORT for UNIX or Windows. Click here for supported computing and storage platforms.

First, you link NewFrame’s SortWize program into the existing JCL as the sort engine. This is accompanied by a configuration file describing the open systems machine(s) available for sorting. When your JCL sort steps run, CoSORT's sort control language (SortCL) program on UNIX or Windows is instructed to execute equivalent specifications on the shared (sortin) data. Because you are sorting elsewhere, your mainframe CPU cycles and I/O capacity remain available for other jobs.

When the sort completes, JCL steps continue automatically with the newly available sorted data. Since the data are shared, it need not move to an open systems disk environment. CoSORT has read the input data (sortin) directly from central storage, and put the results (sortout) on central storage. Thus there are no extra I/O costs to sort the (same) data in the same way.



Benchmarks and Benefits

It is faster to access and process centralized/shared mainframe data using open systems hardware. A simple benchmark proves the case:

Mainframe specs:
Machine: Type 7060 Model P30A Single processor.
This system can deliver 5519.1445 service units per second
SRM MIPS: 113.796 (SU/sec) / 48.5
Processor Speed: 102.9 Million BCTR instructions per second
Real Storage: 921600K
Operating system: OS/390 02.08.00 DFSMS/MVS
 Sort: DF-SORT, Version 1.5.0

PC specs:
Machine: Intel Pentium 3 800 MHz, dual processor, 512 MB memory
Operating system: Windows 2000 Advanced Server
Sort: Windows CoSORT sortcl, Version 7.5.3

Sorted file specs:
RECFM: FIXED Record length: 900
Records: 638,230
Size (bytes): 574,407,000

SortCL script:

/INFILES=(/home/hana/smf/blsmfhds.new208)
   /LENGTH=900
   /FIELD=(field_0, position=15, size=8, ebcdic)
   /FIELD=(field_1, position=31, size=2, mf_comp)
   /FIELD=(field_2, position=11, size=4, ebcdic)

/SORT
   /KEY=(field_0, ascending)
   /KEY=(field_1, ascending)
   /KEY=(field_2, ascending)

/OUTFILE=(blsmfhds.out208)
/statistics

Results:
MVS DF-SORT: 7.00 min.
sortcl, IDE-based storage 1.34 min.
sortcl, EMC-Symmetrix storage (work on IDE): 3.04 min.

Offloading can be faster than running natively because:

  • The connective bandwidth between UNIX/Windows platforms and the stored data runs at SCSI/Fiber rates (versus ESCON/FICON to the mainframe);
  • Newer, high-speed CPUs on encumbered UNIX and Wintel servers now run at faster speeds than many mainframes; and,
  • CoSORT's parallel-processing (SMP) sort engine leverages multiple CPUs aboard these open systems servers

By using NewFrame's data sharing software and CoSORT to offload mainframe sorts, you can simultaneously:

  • Reduce and balance your mainframe loads
  • Improve your mainframe's response times
  • Cut your dependence on mainframe capacity
  • Use available open systems automatically
  • Transfer data at faster FibreChannel I/O rates
  • Save on sort software and mainframe hardware costs

Last but not least, with CoSORT's sortcl licensed on open systems, you can do more than just sort centralized data. You can speed database loads, write complex reports, and perform several flat file ETL functions at once.

 
Platform Availability

Click here to see the requirements schematic. To directly offload mainframe sorts to CoSORT, your configuration must include an IBM-compatible mainframe running either:

  • MVS
  • OS/390
  • z-OS

connected to one of these storage controllers:

  • EMC Symmetrix ESP, model 5230 and above
  • HDS models 7700, 7700E, 99xx, or 99xxV

with a UNIX or Windows server connected to the same storage controller (via SCSI or Fiber or SAN) and running either:

  • HP-UX 11.0 or above
  • Sun Solaris 2.6 or above
  • IBM AIX 4.3 or above
  • Windows NT 4.0 SP3 or above
  • Windows 2000 SP2 or above

Additional server software required:

  • CoSORT v7.5.x or above; and,
  • EMC SymmAPI-Access Runtime; or,
  • Hitachi RapidXchange (formerly HMDE)

 
Licensing and Support

IRI licenses and supports CoSORT and will refer you to Corigin Corp. for the testing, licensing and supportof the mainframe data sharing (SortWize) components as well as the EMC or HDS runtime libraries required.

Click here to obtain more information and/or to arrange your free evaluation.



Despite its international reputation as 'the open systems standard' in sorting software, CoSORT actually has its roots in the proprietary mainframe world. CoSORT's coroutine architecture is derived from COBOL procedure logic, and most of CoSORT's users migrate and upgrade from mainframe sorts when they are 'rightsizing' into open systems. The mainframe-familiar CoSORT sort interface, sortcl, also plays a key role as an "ETL engine" in the world's largest operational data stores (ODS) and data warehouses and webhouses because of its unique single-pass transformations of VLDB extracts and mainframe flat files, and fast pre-sorts that speed loads. CoSORT's fast parallel sorts and joins also occur outside of databases and ETL tools like Informatica's PowerCenter and Ascential's DataStage.