[Products] [Services] [Training] [Recommendations] [Careers] [Links]

Performance Service for Tandem Systems

Our performance service consists of a detailed system analysis from Measure Data, Pathway statistics, File statistics, ODBC Metrics, SQL Catalog information, TCP/IP or other Comm Stats, DSAP, PEEK, and Disk cache analysis. A detailed written report is then produced from these statistics, with charts graphs and lists highlighting the performance trends and potential bottlenecks, followed by a prioritized list of “action items” to be address to increase performance and plan ahead for future growth.

Partial Performance Report Examples…

Note: The complete report is too detailed and lengthy to be shown here, so this is a sampling only.

Performance Report for Tandem System - \MYSYS

CPU Busy

CPU Busy Pie Chart

Logical IOs Per CPU - Pie Chart

Disk Statistics

Disk Reads - Percentage of Total - Pie Chart

Recommendations.

· Move the Spooler data file from disk $DATA03 to $DSV1.

· Balance the Disc IO across all of the CPUs…

o Primary (switch) $DATA04 and $DATA05 from CPU 2 to CPU 3.

o Primary (switch) $DATA07 and $DATA08 from CPU 4 to 5.

o Primary (switch) $DISK02 and $DISK03 from CPU 8 to 9.

· Rebalance the NSKSWAP files evenly across the available discs.
No disc should have more than one NSK swap file. Once the disks are balanced throughout each CPU, then a swap file should be placed on a disc primaried in the CPU it is swapping. CPUs 0 and 1 have 2 Swap files, and should be on different discs.

Automate FUP Reloads with the AutoDBA product.
Many files and indexes have index levels >= 3 and large amounts of slack space and free blocks.
Automate Partition splits with the AutoDBA product.
Some of the large SQL tables have a date-based key. These should have automated partition splits every month. Other SQL tables have a account-type key and the table growth is unpredicatable – use the Escort Ranger enhanced Repart component of AutoDBA to automate repartitioning of these tables.

· Consider – or Plan/Budget for a Memory Upgrade.
Currently, 88.76% of the available memory is in use. However, this memory may have been used and remained dormant for some time, waiting to be swapped. The cache and swap rate can be a good example of memory pressure, which was an acceptable 90%. However, if more disks (or disk controllers), more LAN controllers, or Pathways are added - or even an operating system upgrade - then more memory would be a good idea, and perhaps a necessity.