Charts 3 and 4 show the same results as charts 1 and 2 in the Test Setup topic but use the concept of gain. Gain is the ratio between the time to perform an action using a direct Oracle connection and performing the same action using P6 Compression Server (every other parameter viz. number of users etc. is held constant). As is obvious with charts 1 and 2, charts 3 and 4 show that the benefits of P6 Compression Server are magnified with larger databases. For example, on a Medium WAN P6 Compression Server gets data 4.8 times faster than a two-tier setup, but 2.3 times faster on a small Database. This is a result of P6 Compression Server being able to compress more data with larger loads. The greater the latency, the greater the benefit of P6 Compression Server.
Chart 3: Gain using P6 Compression Server on a Large Database
Chart 4: Login Gain on a Small Database
The next set of charts explains how P6 Compression Server scales with larger numbers of concurrent users. Chart 5 shows the results with up to 15 concurrent users on the large database. Chart 6 shows login performance on the small database. As expected, the performance of P6 Compression Server degrades beyond five concurrent users except for the slow WAN. The bottlenecks in this case are two-fold:
Both sets of data were generated using the SQLPlayer application which emulates P6 Professional. (Note: All references to "PM" in graphics or graphics titles refers to P6 Professional.)
Chart 5: Scalability on a Large Database with Number of Concurrent Users
Chart 6: Scalability on a Small Database with Number of Concurrent Users
The next set of charts shows the impact of using a single NIC on P6 Compression Server rather than two NICs as recommended in the P6 Professional Compression Server Administration Guide. Chart 7 shows the performance on a large database. It is immediately obvious that performance degrades over 50% when using a single NIC from one to five users. Chart 8 shows the same behavior for a small database. The degradation in performance is not as dramatic in this case but is still evident.
Chart 7: Performance Difference between One and Two NICs on P6 Compression Server for Large Data
Chart 8: Performance Difference between One and Two NICs on P6 Compression Server for Small Data
The next set of charts shows open project performance with the small database. Chart 9 shows the raw numbers comparing P6 Compression Server with a direct connection to the database. Chart 10 shows how this translates into gain. As seen here, P6 Compression Server has a greater gain for opening a project when compared to login. This indicates that on high-latency networks, using P6 Compression Server provides performance gains even on small databases.
Chart 9: Open Project Performance with a Small Database
Chart 10: Login and Open Project Gains for a Small Database
The next chart shows the impact of CPU on P6 Compression Server performance. It can be seen that the use of a better CPU translates to better performance even for a single user. Furthermore, the difference between a single user and four concurrent users is more pronounced with the low-end box.
Chart 11: P6 Compression Server performance on high end and low end box
Legal Notices
Copyright © 2005, 2016,
Oracle and/or its affiliates. All rights reserved.
Last Published Thursday, September 22, 2016