socketcluster - npm Package Compare versions

Comparing version 0.9.20 to 0.9.21

index.js

@@ -72,2 +72,3 @@ var path = require('path');
     rebootOnSignal: true,
+    useSmartBalancing: false,
     clusterEngine: 'iocluster'
@@ -310,2 +311,3 @@ };
         protocolOptions: self.options.protocolOptions,
+        useSmartBalancing: self.options.useSmartBalancing,
         checkStatusTimeout: self.options.connectTimeout * 1000,
@@ -312,0 +314,0 @@ statusURL: self._paths.statusURL,

package.json

 {
   "name": "socketcluster",
   "description": "SocketCluster - A Highly parallelized WebSocket server cluster to make the most of multi-core machines/instances.",
-  "version": "0.9.20",
+  "version": "0.9.21",
   "homepage": "https://github.com/topcloud/socketcluster",
@@ -18,4 +18,4 @@ "contributors": [
     "iocluster": ">=1.0.1",
-    "loadbalancer": ">=0.9.13",
-    "socketcluster-server": ">=0.9.24"
+    "loadbalancer": ">=0.9.14",
+    "socketcluster-server": ">=0.9.25"
   },
@@ -22,0 +22,0 @@ "keywords": [

README.md

 SocketCluster
 ======
 
-**See bottom of this page for benchmark results.**
+### Latest benchmark (concurrency) v0.9.20:
 
+**See bottom of page for benchmark tests.**
+
 SocketCluster is a fast, highly scalable HTTP + WebSocket (engine.io) server which lets you build multi-process 
@@ -80,2 +82,3 @@ realtime systems/apps that make use of all CPU cores on a machine/instance.
     // storeController: 'store.js', // Optional
+    // useSmartBalancing: true, // Optional - If true, load balancing will be based on session id instead of IP address. Defaults to false.
     rebootWorkerOnError: false, // Optional, makes debugging easier - Defaults to true (should be true in production),
@@ -372,11 +375,13 @@ addressSocketLimit: 50 // Optional, prevents malicious clients from hogging up unlimited sockets (memory) on your server - Defaults to 30
 
+### Throughput
+
+The goal of this test was to see how many JavaScript (JSON) objects SocketCluster could process each second on a decent machine.
+
 #### Procedure
 
-For this CPU benchmark, we tested SocketCluster on an 8-core Amazon EC2 m3.2xlarge instance running Linux.
-The test procedure here was similar to Benchmark #1 with a few changes:
-* Instead of connecting a new client every 5 seconds, we created a new one every second.
-* The maximum number of connections created was set at 170K.
-* The messages were fully bidirectional this time - The client sent a 'ping' event containing a JavaScript object (cast to JSON) like in Benchmark #1 but instead of just counting it,
-the server responded to that ping event with a 'pong' event.
-* Fewer processes were used this time: 5 load balancers, 5 workers and 2 stores.
+For this CPU benchmark, SocketCluster was tested on an 8-core Amazon EC2 m3.2xlarge instance running Linux.
+* A new client was created every second until there were 100 concurrent clients.
+* The maximum number of messages sent was set to be 170K (1700 messages per second per client).
+* The messages were fully bidirectional - The client sent a 'ping' event containing a JavaScript object (cast to JSON) and the server responded with a 'pong' JavaScript object. That object had a 'count' property to indicate the total number of pings received so far by the current worker.
+* SocketCluster was setup to use 5 load balancers, 5 workers and 2 stores.
 
@@ -394,2 +399,30 @@ #### Observations
 
-![alt tag](https://raw.github.com/topcloud/socketcluster/master/benchmarks/socketcluster_v0.9.8.png)
+![alt tag](https://raw.github.com/topcloud/socketcluster/master/benchmarks/socketcluster_v0.9.8.png)
+
+
+### Concurrency
+
+The goal of this test was to see how many concurrent clients SocketCluster could comfortably handle.
+
+#### Procedure
+
+SocketCluster was deployed on an 8-core Amazon EC2 m3.2xlarge instance running Linux.
+The SocketCluster client was run on the largest possible 32-core Amazon EC2 c3.8xlarge instance running Linux - This was necessary in order to be able to simulate 42K concurrent users from a single machine.
+
+* Clients were created (connected) at a rate of approximately 160 per second.
+* The maximum number of concurrent clients was set to 42K - This is a limit of the client, not the server.
+* Each client sent a 'ping' message every 6 seconds on average. The payload of the 'ping' event was a JavaScript object (cast to JSON), the response was a 'pong' object containing the total number of pings received by the current worker so far.
+* SocketCluster was setup to run with 4 load balancers, 3 workers and 1 store.
+
+#### Observations
+
+* CPU (of busiest worker) peaked to around 60% near the end while new connections where still being created (at rate of 160 per second).
+* Once connections settled at 42K, the CPU use of the busiest worker dropped to around 45%
+* The store didn't do much work - In reality only 7 CPU cores were fully exploited.
+* The load average was under 2 (out of a possible 8), so there was plenty of room for more users.
+* Memory usage was negligible when compared to CPU usage.
+* The huge 32-core EC2 client machine could not get very far past 42K connections - CPU usage on the client was approaching 100% on all 32 cores. Past a certain point, the client would start lagging and the load on the server would drop.
+
+#### Screenshots
+
+![alt tag](https://raw.github.com/topcloud/socketcluster/master/benchmarks/sc_42k_clients.png)

New alerts

License Policy Violation

License

This package is not allowed per your license policy. Review the package's license to ensure compliance.

Found 1 instance in 1 package

Fixed alerts

License Policy Violation

License

This package is not allowed per your license policy. Review the package's license to ensure compliance.

Found 1 instance in 1 package

Improved metrics

Total package byte prevSize

253167

increased by29.78%

Number of package files

10

increased by11.11%

Lines of code

769

increased by0.26%

Number of lines in readme file

425

increased by8.42%

Published versions last week

1

increased byInfinity%

Dependency changes

• Updatedloadbalancer@>=0.9.14

• Updatedsocketcluster-server@>=0.9.25