...

QSAM Access Method for Data sharing with z/OS Barbara Mundle

by user

on
Category: Documents
7

views

Report

Comments

Transcript

QSAM Access Method for Data sharing with z/OS Barbara Mundle
Barbara Mundle
IBM Research and Development
last updated: 2012/03/09
QSAM Access Method for Data sharing
with z/OS
©Jan 20, 2015 IBM Corporation
Smarter Systems for a Smarter Planet – QSAM Access Method for Data sharing with z/OS
Trademarks
IBM, the IBM logo, and ibm.com are trademarks or registered trademarks of International
Business Machines Corp., registered in many jurisdictions worldwide. A current list of IBM
trademarks is available on the Web at “Copyright and trademark information” at
www.ibm.com/legal/copytrade.shtml.
Linux is a registered trademark of Linus Torvalds in the United States, other countries, or
both.
Other product and service names might be trademarks of IBM or other companies.
©Jan 20, 2015 IBM Corporation
Smarter Systems for a Smarter Planet – QSAM Access Method for Data sharing with z/OS
Our Hardware for Measurements
2817-789 H89 (zEC12)
0.18ns (5.5 GHz)
4 books with a total of 95 CPUs (89 configured)
L1 Cache 160kB (64kB Instr. +96kB Data) (each core)
2 MB L2 Cache (each core)
48 MB L3 Cache (per chip)
384 MB L4 Cache (per book)
2048GB RAM (1384GB available)
Driver 15F Bundle #7
8 FICON Express 8S
HiperSockets
OSA Express 4S 10 GiB and 1GiB
2422-961 (DS8870)
233.7 GB Cache
253.7 GB Processor memory total
8 GB NVS
464 * 300 GB disks
15.000 RPM
8 FCP (8 Gbps)
8 FICON (8 Gbps)
©Jan 20, 2015 IBM Corporation
Smarter Systems for a Smarter Planet – QSAM Access Method for Data sharing with z/OS
QSAM Access Method for Data sharing with z/OS
General remarks – Benchmark FIO
 Flexible I/O Testing Tool (FIO)
– Benchmarking and hardware verification / stress I/O devices
– Provides information regarding throughput, latencies, system utilization and many more parameters
– Easy to customize to individual needs
• Can be configured to run with and without page cache (direct I/O), operating modes:
Sequential write/rewrite/read + Random write/read
 Benchmark setup used here:
–
–
–
–
–
–
–
4
FIO with seq. read, direct I/O
Multiple processes – each process writes to or reads from a single file, volume or disk
Sync and drop caches prior to every invocation
zEC12 connected via FICON Express8/8S to a IBM DS8870
1536 MiB main memory, 8 processors
Large DASD with ~100 GB of data: 32 concurrent threads, 3 GB file each
For NFS: 10 GbE connection between z/OS and Linux on z Systems, MTU size 1492
©Jan 20, 2015 IBM Corporation
Smarter Systems for a Smarter Planet – QSAM Access Method for Data sharing with z/OS
QSAM Access Method for Data sharing with z/OS
General remarks – Technical setup
ZDSFS:
–
–
–
–
Method to directly access z/OS data from Linux on z Systems
Very easy to set up and use
Especially made for sequential read
Very fast compared to all “traditional” methods like FTP or NFS
 Technical details:
–
–
–
–
–
–
z/OS and Linux on z Systems in LPAR
ZDSFS: uses fuse file system (user space)
NFS: network connection to z/OS NFS server, direct access to data set
z/OS system: 4 processors, 8GB storage
z/OS V2R1, NFS version 3
No EBCDIC - ASCII conversion done in either case
 Link to development information:
– PDF: http://www.vm.ibm.com/education/lvc/LVC0730.pdf
– Live virtual classes for Linux: http://www.vm.ibm.com/education/lvc/zlinlvc.html
”zdsfs - Direct Linux access to z/OS data sets”
5
©Jan 20, 2015 IBM Corporation
Smarter Systems for a Smarter Planet – QSAM Access Method for Data sharing with z/OS
QSAM Access Method for Data sharing with z/OS
Comparison: Accessing z/OS data via NFS or using QSAM
Linux on z Systems
z/OS
Kernel space
NFS network connection
NFS-server
NFS-client
VFS
User space
Appl.
Fuse
DASD driver
Disk I/O
ZDSFS
Disk I/O
DASD
6
©Jan 20, 2015 IBM Corporation
Smarter Systems for a Smarter Planet – QSAM Access Method for Data sharing with z/OS
QSAM Access Method for Data sharing with z/OS
Throughput normalized to NFS
8.00
7.00
6.00
5.00
4.00
normalized throughput
3.00
2.00
1.00
0.00
nfs
zdsfs
– Accumulated throughput for 32 threads
– ZDSFS with local Linux DASD using 4 HyperPAV devices
– ZDSFS more than 7.5x faster than NFS
7
©Jan 20, 2015 IBM Corporation
Smarter Systems for a Smarter Planet – QSAM Access Method for Data sharing with z/OS
QSAM Access Method for Data sharing with z/OS
Processor consumption normalized to ZDSFS
1.20
1.00
0.80
0.60
Processor consumption
0.40
0.20
0.00
nfs
–
–
–
–
8
zdsfs
Processor consumption: Here Linux on z Systems part only
ZDSFS: Direct access to the DASD, not using z/OS at all
ZDSFS: Only 43% of NFS processor consumption
Z/OS Processor consumption for NFS not even included here!
©Jan 20, 2015 IBM Corporation
Smarter Systems for a Smarter Planet – QSAM Access Method for Data sharing with z/OS
QSAM Access Method for Data sharing with z/OS
General remarks - memory requirements
 ZDSFS:
– Each time a file is opened
• Up to 128 * 120 KB per file is allocated for the track buffer
• A seek buffer of up to 1 MiB is used
– For our case, this means a total of 32*120 KiB*128+1024 KiB*32=524 MiB
 Considerable memory requirements – but remarkable throughput
– Both buffers are tunable to either lower memory consumption or better performance
9
©Jan 20, 2015 IBM Corporation
Smarter Systems for a Smarter Planet – QSAM Access Method for Data sharing with z/OS
QSAM Access Method for Data sharing with z/OS
General remarks – restrictions
 ZDSFS:
Sequential
Random
Read
++
--
Write
not possible
not possible
–
–
–
–
Only reasonable usage: reading in sequential order
Random read extremely slow
Writing to z/OS data sets not possible
Right now limited to only special kinds of data sets:
• Physical sequential data sets (PS), represented as files
• Partitioned data sets (PDS), represented as directories
 NFS:
– complex to set up under z/OS
– Works for various kinds of data sets
– only alternative for write and random access
10
©Jan 20, 2015 IBM Corporation
Fly UP