Abstract：
BACKGROUND: The Fast Health Interoperability Resources (FHIR) standard was proposed and released to solve the interoperability problems of the electronic health records. The FHIR Subscription resources are used to establish real-time event notifications from the FHIR server to another system. There are several communication channels such as rest-hook and websocket. The objective of our work is to compare the performance of the FHIR subscription using the rest-hook and websocket channels.
METHODS: HAPI FHIR server, python websocket clients and HTTP endpoints were used to measure the processor and memory usage of the two subscription channels. Tests were performed with 5, 10, 15, 20, 30, 40, 50, 60, 70 and 80 clients. The performance was logged using windows performance monitor.
RESULTS: The rest-hook subscription showed near six-fold increase in resource utilization when increasing the clients from 5 to 80. On the contrary, the websocket subscription channel did not reach a two-fold increase.
CONCLUSIONS: The type of the subscription channel should be carefully selected and load distribution should be considered when the number of clients grows.
METHODS: HAPI FHIR server, python websocket clients and HTTP endpoints were used to measure the processor and memory usage of the two subscription channels. Tests were performed with 5, 10, 15, 20, 30, 40, 50, 60, 70 and 80 clients. The performance was logged using windows performance monitor.
RESULTS: The rest-hook subscription showed near six-fold increase in resource utilization when increasing the clients from 5 to 80. On the contrary, the websocket subscription channel did not reach a two-fold increase.
CONCLUSIONS: The type of the subscription channel should be carefully selected and load distribution should be considered when the number of clients grows.
摘要：
背景：提出并发布了快速健康互操作性资源（FHIR）标准，以解决电子健康记录的互操作性问题。FHIR订阅资源用于建立从FHIR服务器到另一个系统的实时事件通知。有几个通信通道，如rest-hook和websocket。我们工作的目的是使用rest-hook和websocket通道比较FHIR订阅的性能。
方法：HAPIFHIR服务器，pythonwebsocket客户端和HTTP端点用于测量两个订阅通道的处理器和内存使用情况。对5、10、15、20、30、40、50、60、70和80个客户进行测试。使用windows性能监视器记录性能。
结果：当将客户端从5个增加到80个时，rest-hook订阅显示资源利用率增加了近六倍。相反,websocket订阅频道没有达到两倍的增长。
结论：应仔细选择订阅通道的类型，并在客户端数量增加时考虑负载分配。
方法：HAPIFHIR服务器，pythonwebsocket客户端和HTTP端点用于测量两个订阅通道的处理器和内存使用情况。对5、10、15、20、30、40、50、60、70和80个客户进行测试。使用windows性能监视器记录性能。
结果：当将客户端从5个增加到80个时，rest-hook订阅显示资源利用率增加了近六倍。相反,websocket订阅频道没有达到两倍的增长。
结论：应仔细选择订阅通道的类型，并在客户端数量增加时考虑负载分配。
