Concurrent page access statistics are simple, and concurrent page visits

Concurrent page access statistics are simple, and concurrent page visits

Page access statistics may show you how to implement them easily when you go to school. For example, you can save a page access count in servletContext and Add 1 for each access; or save the operation records to the database every time you access the database. This processing method will be used for experiments. In actual applications, this will have a significant impact on the system performance.

The first method is to add a synchronization lock when modifying the public variables because the number of page visits is a public variable. The synchronization lock will significantly slow the access speed; the second method is the same, and frequent database access is not a reasonable method.

Not long ago, a friend of mine asked me to help them write a simple page statistics code. 1. The requirement is to save the page access IP address, time, and other available information, the access information to be saved in the future can be expanded. 2. The current access speed cannot be affected. 3. A certain amount of concurrent access is supported.
After receiving this request from a friend, I thought of the following points: 1. How to filter the pages we need for statistics; 2. Separating access from statistics, the access information is not stored in the access thread. The other thread saves the access information to the database. 3. A public queue can be used to save the access information; 4. You can save a certain amount of access information in batches.
　　

Solution:

1. For the first question, I have provided two methods. 1. Use a set to save all the pages for statistics, and then judge whether the current request is in the statistics column in the Filter. 2. Introduce a public code in the JSP page, use CounterUtils like this in the code. addCounter (request); this method has the advantage that it is easier to maintain pages that require statistics, and it feels more efficient, and does not require Filter interception. However, it is impossible for a friend to use the first method.

2. Every access, we save the information to be saved into an object, and then put it into the queue. Then another thread regularly saves the information.

So I wrote a simple demo to my friends. It wasn't long before I was returned. After testing, the database is not saved until the concurrency reaches 200, and the access slows down. In the end, the heap memory overflows.
There is no way to test with loadRunner in the current machine, and detect memory changes through the jconsole java tool. The test is the same as my friend said. It can run normally at the beginning. When the concurrency reaches a certain amount, the storage starts to slow down. At last, I don't know how to keep the entire storage thread from running, in this way, the queue is getting bigger and bigger, and the memory size of the heap naturally overflows.

　　
From the above situation, we can also think that a queue may not support such a large number of concurrent accesses, so we want to use multiple queues for storage, using methods similar to table sharding and database sharding, different requests are allocated to different queues, so the following method is used:

　　
Some code is as follows:
1. initialize and generate the sequence list set list

/*** Generate a queue Array Based on urls * @ return */private static queue list <RequestStc> [] initUris () {Digester digester = new Digester (); String path = null; try {path = CounterUtils. class. getClassLoader (). getResource ("urls. xml "). toURI (). getPath ();} catch (URISyntaxException e1) {e1.printStackTrace ();} UriRuleSet ruleSet = new UriRuleSet (); ruleSet. addRuleInstances (digester); try {// uri set uris = digester. pars E (new File (path); // hashCode base BaseHash = uris! = Null? Uris. size ()/3:1; parameter list <RequestStc> [] listArr = new parameter list [BaseHash]; for (int I = 0; I <listArr. length; I ++) listArr [I] = new javaslist <RequestStc> (); return listArr;} catch (Exception e) {e. printStackTrace (); return null ;}}

 

2. encapsulate the request as a required statistical object

/*** Add a request statistics * @ param request */public static void addCounter (HttpServletRequest request) {// encapsulate the user statistics request, in addition, the hash algorithm is used to distribute data to different queues. RequestStc stc = new RequestStc (); stc. setIp (request. getRemoteAddr (); stc. setUri (request. getRequestURI (); stc. setNow (Calendar. getInstance (). getTime (); buffers [request. hashCode () % BaseHash]. push (stc );}

 

3. Round-Robin into the list queue set

/*** Execution statistics */private void processCount () {try {// round-robin queue while (true) {Thread. sleep (Sleep_MS); if (buffers = null) {break;} Thread th = null; for (int I = 0, len = buffers. length; I <len; I ++) {buffers list <RequestStc> stcList = null; if (buffers [I]. size ()> = Execute_Base) {// copy the buffers array element and clear it. Start a thread to save synchronized (buffers [I]) to the queue. {stcList = (queue list <RequestStc>) buffers [I]. clone (); buffers [I]. clear () ;}th = new Thread (new ExecuteThread (stcList); th. start () ;}}} catch (Exception e) {e. printStackTrace ();}}

 

4. The ExecuteThread thread is used to save access logs in batches.

// Save the database in batches. This is omitted.

5. How to intercept access requests that require statistics
Method 1: Determine whether the uri is in the column needing statistics
Method 2: add the JAVA code such as CounterUtils. addCounter (request) to the jsp to be counted );

After modification, in the loadRunner and tomcat tests, it can basically achieve the maximum concurrency of tomcat and above users, and occupy a small amount of resources.
PS: the above method is purely a personal solution. If there is any problem, please do not hesitate to advise!