Virtel/TN3270 overhead comparison
SysperTec benchmarked similar Virtel and TN3270 workloads on a z/OS system to compare the respective CPU footprints of both web access solutions. The following table provides
estimates of Virtel overhead for small, medium, and large applications:
| Virtel application |
Small |
Medium |
Large |
| Partition (LPAR) size |
50 MSU |
150 MSU |
350 MSU |
| Average concurrent users |
800 to 1,000 |
2,000 to 3,000 |
> 10,000 |
| Traffic volume |
Small |
Medium |
Large |
Type of user
interface |
3270 screen
or rich GUI |
3270 screen
or rich GUI |
3270 screen
or rich GUI |
Connections per
90-second period |
1 |
1 |
1 |
Transactions per
90-second period |
18 |
18 |
18 |
Transactions
per connection |
18 |
18 |
18 |
| Bytes per transaction |
250 |
500 |
1,000 |
| Bytes per connection |
4,500 |
9,000 |
18,000 |
| Transactions per day |
800,000 |
1,600,000 |
6,400,000 |
| Virtel overhead |
Small |
Medium |
Large |
| In percentage of partition size |
1 / thousand (about 0,102%) |
1.3 / thousand (about 0.129%) |
4.3 / thousand (about 0.403%) |
Traffic volume
It doesn't make any difference in terms of CPU consumption whether Virtel is configured to:
- Replace 3270 emulation using the default presentation template for classic 3270 screen experience
- Create user-friendly web pages (GUIs) with JavaScript widgets (dropdown lists, graphical calendars, check boxes, etc) and new AJAX application features using
custom-developed presentation templates
In both cases Virtel inserts the data sent by the 3270 application into an HTML/JavaScript template to create a web page file that it sends to the client browser for rendering.
The GUI smarts coded in the template do not run on the host and do not consume host resources: they run on the client browser and consume client resources. With Virtel, a rich GUI
doesn't cost more MSU than a web page emulating a 3270 screen.
Traffic volumes have been calculated conservatively so that most customers fall within or below our estimates.
Virtel overhead
It is unlikely that customers will ever experience an increase of their mainframe costs due to Virtel because:
- Most systems are "MSU-capped" so that they will not automatically resize into a higher MSU band in case of high CPU consumption to avoid unwillingly increasing the
systems' costs
- Most systems are sized for average use at 60-70% of capacity: with overheads ranging between one and five 1000th of capacity, Virtel has very little impact on overall
system use, and very little chance of impacting the MSU cap
- Virtel doesn't need to run in the CICS, IMS, or TSO partition it serves: if that partition is running close to capacity, Virtel can be configured to run in a partition
with lower use (for example a batch partition) and to serve the CICS, IMS or TSO applications in cross-domain mode to avoid impacting the MSU cap
- The comparison above is based on a standard Virtel configuration where exchanges between Virtel and the application it serves are through VTAM... But Virtel can be
configured to bypass VTAM and use cross-memory (EXCI) exchanges with the application it serves to further reduce the CPU footprint
Estimate Virtel overhead
To estimate for your own application the overhead of a Virtel-based screen-based web access or user-friendly GUI send us average and peak 3270 emulation traffic statistics
(connection/transaction counts per period, bytes per transaction/connection, daily transaction count) and MSU base cost.
Read more