Re: NIO+Multithreaded TCPSocket listener, very low cpu utilisation
So, what is next? Can we enable some sort of profiling to see what is going on?
Re: NIO+Multithreaded TCPSocket listener, very low cpu utilisation
Mediocre result... let me create a java equivalent program so we have a direct comparison.. These are the tests for a similar program in java. bombardier -c 200 -n 1 http://localhost:8081 Bombarding http://localhost:8081/ with 1 requests using 200 connections 1 / 1 [==] 100.00% 0s Done! StatisticsAvg StdevMax Reqs/sec 24527.29 14797.09 46439 Latency8.25ms15.52ms 235.79ms HTTP codes: 1xx - 0, 2xx - 1, 3xx - 0, 4xx - 0, 5xx - 0 others - 0 Throughput: 3.87MB/s
Re: NIO+Multithreaded TCPSocket listener, very low cpu utilisation
Result: bombardier -c 200 -n 1 http://localhost: Bombarding http://localhost: with 1 requests using 200 connections 1 / 1 [===] 100.00% 1m24s Done! StatisticsAvg StdevMax Reqs/sec 122.27 909.33 20363 Latency 49.98ms 192.16ms 1.07s HTTP codes: 1xx - 0, 2xx - , 3xx - 0, 4xx - 0, 5xx - 0 others - 1 Errors: the server closed connection before returning the first response byte. Make sure the server returns 'Connection: close' response header before closing the connection - 1 Throughput:36.89KB/s Mediocre result... let me create a java equivalent program so we have a direct comparison..
Re: NIO+Multithreaded TCPSocket listener, very low cpu utilisation
On Thursday, 16 November 2017 at 18:44:11 UTC, Daniel Kozak wrote: It works for me because I have multiple threads, but when I use only one thread per pool (defaultPoolThreads(1)), it obviosly blocks, which is correct behavior Ok, let me force the: "defaultPoolThreads(8)" and let me re-test it. By the way, if the socket is blocking you can remove all the socketSet and the socket.select as they are not needed anymore. like this.. https://dpaste.dzfl.pl/ca09e4c54789
Re: NIO+Multithreaded TCPSocket listener, very low cpu utilisation
On Thursday, 16 November 2017 at 18:20:36 UTC, Daniel Kozak wrote: Hmm works ok for me. What OS? Dne 16. 11. 2017 12:05 dop. napsal uživatel "kdevel via Digitalmars-d-learn": [...] I'm running MacOS..
Re: NIO+Multithreaded TCPSocket listener, very low cpu utilisation
On Wednesday, 15 November 2017 at 23:04:46 UTC, kdevel wrote: On Wednesday, 15 November 2017 at 13:31:46 UTC, Daniel Kozak wrote: This one works ok for me, but I am on linux: https://dpaste.dzfl.pl/f54decee45bc It works, but it does not handle two connects in parallel. STR: 1. start the binary in console 1 2. telnet localhost in console 2 3. telnet localhost in console 3 4. enter a [RETURN] in console (3) observed: nothing (the "thread" handling the first connect blocks) expected: response On my machine defaultPoolThreads() returns 1. This can easily be increased: ``` defaultPoolThreads(8); ``` There is also another problem with Socket.select: It may return -1. This frequently happens when the ACCEPT socket is non-blocking and the select is interrupted (by the GC?) then errno == EINTR. Also not having a timeout in the Socket.select of handle_socket allows for DOS attacks like the one above. In case of a timeout select also returns -1. Hi Guys, so i have tried the latest changes with destroy(socket) rather than socket.close(). Still the program is extremely slow. I see that the httpclinet is making successfully 100 request at a time and then it freezes for 5/10 seconds and then it process another 100 request. As Kdevel pointed out, this could be related to the Socket.select blocking on the interupt or it would be GC kicking in. I also have seen the value computed for the defaultThread is not correct, hence i manually started the TaskPool(int thread) manually. I'm going to share my benchmark program so you can guys test with the same tool. Unfortunately it seems that there is some underline issue which is blocking the program and preventing to process all the incoming requests in a performant fashion, or at-least utilising the full CPU cycles. reagrds ade90036
Re: NIO+Multithreaded TCPSocket listener, very low cpu utilisation
So thanks for the suggestions, i have fixed HTTP response not postman cal also parse the headers correctly!! happy days. I have removed the duration from the Socket.select but the application seems to process a bunch or requests and then it stalls for several seconds (3/5) and then it resumes. The httpclinet which i'm using to test the application is reporting: "connection timeout". Could this be caused by the GC? ```updated code import std.algorithm : remove; import std.conv : to; import core.thread: Thread; import std.socket : InternetAddress, Socket, SocketException, SocketSet, TcpSocket, SocketShutdown; import core.time : Duration, dur; import std.stdio : writeln, writefln; import std.parallelism : task, TaskPool; string to_retlf (string s) { import std.algorithm; import std.string; return s .lineSplitter .map!(a => chomp (a)) .join ("\r\n"); } void main(string[] args) { ushort port; if (args.length >= 2) port = to!ushort(args[1]); else port = ; auto listener = new TcpSocket(); assert(listener.isAlive); listener.blocking = false; listener.bind(new InternetAddress(port)); listener.listen(100); writefln("Listening on port %d.", port); auto taskPool = new TaskPool(8); string response = "HTTP/1.1 200 OK Server: dland:v2.076.1 Date: Tue, 11 Nov 2017 15:56:02 GMT Content-Type: text/html; charset=UTF-8 Content-Length: 51 Hello World!".to_retlf; new Thread({ auto listeningSet = new SocketSet(); while(true) { listeningSet.add(listener); if (Socket.select(listeningSet, null, null)) { if (listeningSet.isSet(listener))// connection request { Socket socket = listener.accept(); assert(socket.isAlive); //writefln("Connection from %s established.", socket.remoteAddress().toString()); auto task = task!handle_socket(socket, response); taskPool.put(task); } } listeningSet.reset(); } }).start(); } void handle_socket(Socket socket, string response) { auto socketSet = new SocketSet(); while(true) { socketSet.add(socket); if (Socket.select(socketSet, null, null)) { char[1024] buf; auto datLength = socket.receive(buf[]); if (datLength == Socket.ERROR) writeln("Connection error."); else if (datLength != 0) { //writefln("Received %d bytes from %s: \"%s\"", datLength, socket.remoteAddress().toString(), buf[0..datLength]); //writefln("Writing response"); socket.send(response); } // release socket resources now socket.close(); break; } socketSet.reset(); } ``` Regards
Re: NIO+Multithreaded TCPSocket listener, very low cpu utilisation
On Tuesday, 14 November 2017 at 21:09:40 UTC, kdevel wrote: On Tuesday, 14 November 2017 at 19:57:54 UTC, ade90036 wrote: while(true) { listeningSet.add(listener); if (Socket.select(listeningSet, null, null, dur!"nsecs"(150)) > 0) { Why do you ever timeout? This loop consumes 100 % (a single core) when idle on my machine. Thanks for you feedback, i'll make those cosmetic changes to the HTTP response Why do i specify a timeout in the Socket.select? I have noticed that if i don't specify a timeout the Socket.select seems to block every so often for several seconds (2/5sec) and i have read in another post to specify a timeout value to the call. I'll retest now.
NIO+Multithreaded TCPSocket listener, very low cpu utilisation
Hi Forum, Let's cut the chase, i'm a newby in Dlang. I have 15+ years experience in java and 7+ years experience in C++. I found D very fascinating and the sugar coated syntax very appealing to my style of coding. (groovy like) I've been trying to learn Dland and bring it thought the motions by creating a very simple and basic TCPSocket listerner that when you send a request it responds with an HTTP response over a specific port. (localhost:4445) I have the code working on a worked thread and when the socket accept() it defers the processing of the request (socket) in a different thread backed by TaskPool(8). i have 8 logical core, which is a macBook pro retina 16gb, i7. What i'm expecting to see is the CPU of my 8 core I7 go through the roof and nearly melt (hope not) but at-least have the fan on at sustainable level and obtain full CPU utilisation. What i'm benchmarking it against is a JAVA NIO2 implementation. This implementation achieves very high CPU utilisation and high throughput. The process utilisation averages 400% at times and the fan is really searching for cold air. (Nic) However, when i run the Dlang program i see i very poor CPU utilisation. The fan is always in silent mode. Not sure if you are familiar with MacOS cpu metrics, but they are based per core. So dland program reports 100% under the process monitor (which equates to one core) and the overall system CPU utilisation is 13%. I would have expected to see a much higher cpu utilisation but it is not happening. I have been trying different variation of the same implementation but not luck. I'm starting to suspect that this is an BUG related to the macOS but i would like to confirm or atleast have a second pair of eyes having a look. ```Code import std.algorithm : remove; import std.conv : to; import core.thread: Thread; import std.socket : InternetAddress, Socket, SocketException, SocketSet, TcpSocket, SocketShutdown; import core.time : Duration, dur; import std.stdio : writeln, writefln; import std.parallelism : task, TaskPool; void main(string[] args) { ushort port; if (args.length >= 2) port = to!ushort(args[1]); else port = 4447; auto listener = new TcpSocket(); assert(listener.isAlive); listener.blocking = false; listener.bind(new InternetAddress(port)); listener.listen(100); writefln("Listening on port %d.", port); auto taskPool = new TaskPool(8); new Thread({ auto listeningSet = new SocketSet(); while(true) { listeningSet.add(listener); if (Socket.select(listeningSet, null, null, dur!"nsecs"(150)) > 0) { if (listeningSet.isSet(listener))// connection request { Socket socket = null; scope (failure) { writefln("Error accepting"); if (socket) socket.close(); } socket = listener.accept(); assert(socket.isAlive); assert(listener.isAlive); //writefln("Connection from %s established.", socket.remoteAddress().toString()); auto task = task!handle_socket(socket); taskPool.put(task); } } listeningSet.reset(); } }).start(); } void handle_socket(Socket socket) { auto socketSet = new SocketSet(); while(true) { socketSet.add(socket); if (Socket.select(socketSet, null, null, dur!"nsecs"(150)) > 0) { char[1024] buf; auto datLength = socket.receive(buf[]); if (datLength == Socket.ERROR) writeln("Connection error."); else if (datLength != 0) { //writefln("Received %d bytes from %s: \"%s\"", datLength, socket.remoteAddress().toString(), buf[0..datLength]); //writefln("Writing response"); socket.send("HTTP/1.1 200 OK Server: dland:v2.076.1 Date: Tue, 11 Nov 2017 15:56:02 GMT Content-Type: text/plain; charset=UTF-8 Content-Length: 32 Hello World!"); } // release socket resources now socket.shutdown(SocketShutdown.BOTH); socket.close(); break; } socketSet.reset(); } } ``` You help in understanding this matter is extremelly helpfull. Regards
Re: rawRead using a struct with variable leght
On Monday, 5 June 2017 at 16:30:53 UTC, Era Scarecrow wrote: On Monday, 5 June 2017 at 16:04:28 UTC, ade90036 wrote: Unfortunately the struct doesn't know at compile time what the size of the constant_pool array, or at-least was not able to specify it dynamically. It also won't know ahead of time how many fields, methods or attributes you have either. First I'd say all the arrays will have to be redefined to use [], rather than a fixed size. Glancing at the chapter information, you're probably not going to have an easy time, and will have to simply have to fill in the fields individually in order followed by allocating the arrays and probably filling/loading those immediately (although it's possible the array contents are done at the end, though it seems doubtful). Thanks for the reply. I guess i didnt pick such an easy task afterall. I shall parse each fields individually. Thanks again..
rawRead using a struct with variable leght
Hi everyone, I'm trying out Dland, always been and have been a big fan. So to give it a good run i wanted to create is a java class parser, based on the spec released here. ( https://docs.oracle.com/javase/specs/jvms/se7/html/jvms-4.html.) The class file can be represented in the following "struct" like: ClassFile { u4 magic; u2 minor_version; u2 major_version; u2 constant_pool_count; cp_infoconstant_pool[constant_pool_count-1]; u2 access_flags; u2 this_class; u2 super_class; u2 interfaces_count; u2 interfaces[interfaces_count]; u2 fields_count; field_info fields[fields_count]; u2 methods_count; method_infomethods[methods_count]; u2 attributes_count; attribute_info attributes[attributes_count]; } where: u4 == ubyte[4] --> integer u2 == ubyte[2] --> short u1 == ubyte[1] --> byte I have the first 4 fields parsing, however i stumble upon an example where you can use a rawRead() with a struct, therefore representing the entire structure and then reading from rawRead() like d ClassFile[1] classFileStruct; f.rawRead(claddFileStruct); Unfortunately the struct doesn't know at compile time what the size of the constant_pool array, or at-least was not able to specify it dynamically. What is the best approach to go about parsing such structure? Should i have structs to represent the blocks of fixed fields from the structure and then parsing each variable (length) structure manually? Something like this? ClassHeader[1] classHeader; f.rawRead(classHeader); CpInfo[] cpInfo = new CpInfo[classHeader.constant_pool_count]; f.rawRead(cpInfo); ` Really appreciate any knowledgeable suggestions. Regards Ade