With a large number of applications pouring into the market and joining the “battle of APP roll in”, end users are increasingly picky about the requirements of application quality. The end-to-end R & D students only pay attention to the solution of crash bugs, which can no longer meet the demands of users on the app experience. Many users will also feed back performance related problems, such as too long app startup time, page blocking, flash back, etc. With the improvement of online users’ demands at any time, many development students have evolved from offline test performance to paying more attention to online performance, so as to ensure the user experience.
Youmeng + application performance monitoring platform u-apm has been accessed and supported by many developers in the industry since it was launched free of charge at the end of last year. Youmeng + also attaches great importance to various problems that developers should be able to monitor. In recent months, based on the optimization of the original stability function, u-apm has added three performance modules: startup analysis, memory analysis and Caton analysis to comprehensively help developers improve the user experience.
The startup scenario is the first barrier for users to use the app. Slow startup and startup flash back will directly block customers’ use from the source, and even cause unsuccessful startup and new customers’ unloading. Many technical teams focus on the start-up time as a key performance monitoring indicator. The startup analysis in u-apm includes three functional modules: startup trend, slow startup analysis and startup crash analysis.
Startup analysis supports defining startup phases through preset collection and personalized customization. You can query the effects of first startup, cold startup and hot startup respectively, and cross filter queries with dimensions such as equipment, system, version and region.
You can also set the business definition of slow start for first start, cold start and hot start respectively. Generally, the time of hot start is much lower than that of cold start and first start. Slow start analysis can monitor the number of slow start devices and equipment system distribution in three cases, and support a single device to query the start timing at a fixed point to accurately locate the problem.
The crash in the startup phase should be the primary problem to be solved in daily development to avoid that users can’t continue to use it in a short time. The startup crash analysis filters out the crash list under the user-defined startup time limit, which is more convenient to locate the startup problem.
The memory analysis of u-apm provides monitoring and analysis of online oom exceptions to help developers find and locate online oom problems in time. At the same time, it provides the memory usage during the app running phase and provides key index data for application memory optimization.
In Android exceptions, not all oom problems can be judged by simply checking whether the error summary contains out of memory.
In the module of memory analysis oom exception, the internal precipitation has been exposed to the outside with the intelligent diagnosis technology that has been used for many years. You can intelligently read the error stack for matching, mine and cluster the abnormal problems that are not oom but are actually caused by oom, and the judgment rate of oom exceptions can be directly increased by 20% ~ 30%. IOS applications also added the capture of oom exceptions this time.
The memory occupation module displays the information of key memory indicators, and the distribution of equipment conditions is used as a statistical reference. It also provides the distribution of channels, versions, systems and equipment models.
The u-apm reports the device information and the Caton log with the Caton experience through the response time of the main thread. All IOS collection and Android dual terminal Caton collection are provided for free, and such a large amount of log data does not need to be paid by traffic.
In addition, if developers find it very troublesome to look at the wrong stacks one by one, they can use u-apm to provide the function of the Caton module by using the aggregation algorithm, which is equivalent to seeing the contents of 200 stacks in about one minute, effectively saving developers a lot of time for mining problems. The Caton module supports two aggregation forms: positive order and reverse order:
• positive order aggregation: select 200 stacks that affect a large number of users and aggregate them from the top of the stack to the bottom of the stack to help customers mine the core problems causing the stuck problem
• reverse order aggregation: filter 200 stacks that affect a large number of users and aggregate them from the bottom of the stack to the top of the stack to help customers mine the core problems causing the stuck problem
The depth of the first two sub modules is up to 50, and the first two sub modules can display the detailed information
Using OpenAPI to call error data
The error data in the background of u-apm has many applications in the daily business of developers. For example, regularly pour the application quality data into the weekly report of the technical team and send it to all departments of the company, or display today’s real-time error rate and other performance indicators in the company’s own background. The new OpenAPI can solve the demand for flexible access to the wrong data collected by the application, support the flexible call of data within 90 days of real-time / offline, and open the permission for free.
Monitoring alarm upgrade
The monitoring alarm function in u-apm has been upgraded recently to support more flexible and personalized monitoring alarm settings. In addition to calling OpenAPI to process data and set alarms, it is more convenient for developers to use the monitoring alarm function in the u-apm background:
a. Flexible alarm effective time:
Developers can add the effective time period of the alarm, such as 9:00 to 19:00 from Monday to Friday and 12:00 to 20:00 on weekends. They can flexibly set the working hours without being disturbed by invalid information.
b. Key error type / single error alarm: developers can select the error type that needs your attention
Or direct continuous attention alarm for an error in repair
c. Combined alarm trigger conditions
Developers can flexibly set the desired alarm trigger conditions through a variety of indicators and threshold or contrast rules in the combination of intersection / Union:
d. Multiple alarm access channels
If the developer also has requirements for the touch channel of monitoring alarm, it can consider using the company’s office software for group touch, and pay attention to and repair the application problems together with other colleagues in the same group.
Error capture upgrade
In the recent upgrade of u-apm, the types of error capture have increased significantly:
Android supports Java and native crash collection; ANR； And error types added c#, Lua for unity SDK;
IOS crash support
In addition to the above functions, u-apm has also updated the UI design of Yunzhen machine, the overall loading speed of API upload symbol table page, rendering and other functions. For details and free use, please go to:https://www.umeng.com/apm?&ut…