Analysis

---

Figure 6-6 demonstrates that serialization and deserialization have a negative impact on transaction rates. We can also conclude that the cost of serialization is discernible when the serialized objects are nonprimitive and more complex. Yet another conclusion, not surprising, is that the serialization cost increases with the size of the object. Computing cost and response times are significantly higher for transactions involving the biggest object, Bank.

Figure 6-8 reiterates the fact that the transaction rate costs increase steadily with higher volumes. We infer this from the fact that the difference in transaction rate between the local and remote versions of an object is smaller with one user than with four users. This is an important consideration for heavily used transactions. The results of Figure 6-9 are quite surprising considering that it implies that object complexity due to inheritance does not affect the serialization costs. At least the single level inheritance did not affect the costs, as the results indicate. The cost of serialization did decrease when the application developer took responsibility for the format of serialized data, but the improvement is about (3-2.6)/2.6*100 = 15.4% of the transaction rate for two virtual users.

---