Add Evaluation Results and Add SimulateRace

README.md

@@ -9,8 +9,10 @@ project
 │   TestCase1.json
 │   TestCase2.json
 │   TestCase3.json
+│   TestCase3Modified.json
 │   TestCase2EC2.json
 │   TestCase3EC2.json
+│   TestCase3EC2Modified.json
 │
 └───doc
 │   │   doc.md // design description
@@ -41,11 +43,21 @@ project
 
 ## Performance Evaluation
 
+Test cases are:
+
+![Screenshot](../TestCases.png)
+
+Test Case 3 modified is to check how performance of the network changes when a node 2 and node 3 is connected!
+
+The data are in the data folder, and I skipped the first 2 entries due to some processes start up slower than others. 
+There are around 1000 data points for each measurement criteria. I average all data points to got the results in the table. 
+Lookup Response Time is the time elapsed between when the Buyer first sends out a lookup request and when the Buyer receives a reply from Seller! 
+
 **Local Computer**
 
-The data are in the data folder, and I skipped the first few entries due to some processes start up slower than others. 
+My local computer is a 4 core 16GB RAM. 
 
-| TestCase    | Peer   | Lookup ResponseTime (ms) | lookupRPC (ms)  | replyRPC (ms)  | buyRPC (ms)  |
+| TestCase    | Peer   | Lookup Response Time (ms) | lookupRPC latency (ms)  | replyRPC latency (ms)  | buyRPC latency (ms)  |
 | :---        |    :----: | :----:  |          :---: | :---: | ---: |
 | TestCase3.json  | Node 0    |   41.36     | 6.04  |  4.6  |  3.5  |
 | TestCase3.json   | Node 1   |   35.72   |  6.3  |  3.4  |  3.3  |
@@ -62,8 +74,38 @@ The data are in the data folder, and I skipped the first few entries due to some
 | TestCase3Modified.json   | Node 5   |    _    |  7.2  |  14  |  _  |
 | TestCase3Modified.json   | Node 6   |    _    |  8.4  |  7  |  _  |
 
+As we can see from the table, the lookup response time increases drastically when a new connection between node 2 and node 3 is introduced. 
+The increase in response time for node 0 and 1 is bigger than increase in response time for node 2. This is due to node 3 being a connecting node for node 0 and node 1 to the Seller, but now node 3 also have to serve node 2.
+The increase in response time can be due to (1) our single machine has to create extra threads to handle service call and context switching takes more time; (2) node 3 becoming a bottleneck since it has so many clients to serve!
+There are some increases in latency in lookupRPC and replyRPC, but this is much smaller compare to the increase in lookup response time!
+
 **EC2 Cluster**
 
+I deploy each peer on a t2-micro EC2 instance. I skipped the first 2 entries due to some processes start up slower than others!
+
+| TestCase    | Peer   | Lookup Response Time (ms) | lookupRPC latency (ms)  | replyRPC latency (ms)  | buyRPC latency (ms)  |
+| :---        |    :----: | :----:  |          :---: | :---: | ---: |
+| TestCase3.json  | Node 0          |   34.5  | 5.77  |  5.4  |  5.9  |
+| TestCase3.json   | Node 1         |   29.7  |  6.6  |  4.8  |  3.9  |
+| TestCase3.json   | Node 2         |   26.1  |  5.7  | 4 |  3.9  |
+| TestCase3.json   | Node 3         |    _    |  6.5  |  5  |  _  |
+| TestCase3.json   | Node 4         |    _    |   5.3 |  4.6  |  _  |
+| TestCase3.json   | Node 5         |    _    |   4.7 |  6.04  |  _  |
+| TestCase3.json   | Node 6         |    _    |   4.9 |  5.3  |  _  |
+| TestCase3Modified.json  | Node 0    |   41.1  | 7.3  |  5.0  |  5.3  |
+| TestCase3Modified.json   | Node 1   |   35.7  |  6.3  | 4.8  |  3.6  |
+| TestCase3Modified.json   | Node 2   |   28.8  |  6.3  | 3.8 |  3.1  |
+| TestCase3Modified.json   | Node 3   |    _    |  7.4  |   5  |  _  |
+| TestCase3Modified.json   | Node 4   |    _    |  5.7  |  4.7  |  _  |
+| TestCase3Modified.json   | Node 5   |    _    |  4.03  |  6  |  _  |
+| TestCase3Modified.json   | Node 6   |    _    |  4.64  |  4.6  |  _  |
+
+Similarly, on a distributed environment, the response time for lookup request increase when we introduce a new connection between 2 and 3.
+The increase in response time for node 0 and 1 is bigger than increase in response time for node 2. This is due to node 3 being a connecting node for node 0 and node 1 to the Seller, but now node 3 also have to serve node 2.
+However, the increase is much less compare to single server! This can be due the cost of new connection is distributed among involved peers on different machine, 
+while on a single server, the machine incurs the whole cost. 
+
+From the experimental results, we can conclude that when a 
 
 
 ## Build and Test
@@ -72,14 +114,14 @@ The data are in the data folder, and I skipped the first few entries due to some
 
 The project is build with Java 1.8, Gradle 6.7, and gRPC 1.36.0
 
-To run:
+To compile the code:
 ``` 
 $ ./gradlew clean build
 ```
 
-**hopcount** is hardcoded to be 1 for milestone 1. Professor confirmed that it's okay to have the two peers: one buyer, one seller for milestone 1.
+### To deploy on EC2 instance 
 
-Image ID for the customized EC2 instance:
+View instruction in doc/testVerify.md. Customized image with Java 8, git installed and repo cloned:
 
 {
 "ImageId": "ami-07916b33d72291f85"
@@ -91,7 +133,7 @@ Image ID for the customized EC2 instance:
 - [X] Test case 1: 1 buyer and 1 seller 
 - [X] Test case 2: Race condition 2 buyers 1 seller
 - [X] Test case 3: 7 peers 3 Buyers 2 Sellers 2 peers
-= [ ] Simulate a race condition 
+- [X] Simulate a race condition 
 - [X] Running test case on a single server EC2
 - [X] Deploy test case on EC2 cluster
 - [X] Measure latency

TestCase3EC2Modified.json

@@ -0,0 +1,73 @@
+{
+  "N": "7",
+  "K": "4",
+  "network": [
+    {
+      "id": "0",
+      "IPAddress": "172.31.45.247",
+      "port": "8080",
+      "neighbors": [
+        "3",
+        "1"
+      ]
+    },
+    {
+      "id": "1",
+      "IPAddress": "172.31.91.178",
+      "port": "8080",
+      "neighbors": [
+        "0",
+        "2",
+        "3"
+      ]
+    },
+    {
+      "id": "2",
+      "IPAddress": "172.31.88.28",
+      "port": "8080",
+      "neighbors": [
+        "1",
+        "3",
+        "4"
+      ]
+    },
+    {
+      "id": "3",
+      "IPAddress": "172.31.55.0",
+      "port": "8080",
+      "neighbors": [
+        "0",
+        "1",
+        "2",
+        "5"
+      ]
+    },
+    {
+      "id": "4",
+      "IPAddress": "172.31.18.128",
+      "port": "8080",
+      "neighbors": [
+        "2",
+        "6"
+      ]
+    },
+    {
+      "id": "5",
+      "IPAddress": "172.31.85.180",
+      "port": "8080",
+      "neighbors": [
+        "3",
+        "6"
+      ]
+    },
+    {
+      "id": "6",
+      "IPAddress": "172.31.95.38",
+      "port": "8080",
+      "neighbors": [
+        "4",
+        "5"
+      ]
+    }
+  ]
+}

build.gradle

@@ -75,22 +75,11 @@ protobuf {
     }
 }
 
-task p2pSeller(type: JavaExec) {
+task simulateRace(type: JavaExec) {
     classpath sourceSets.main.runtimeClasspath
-    main = "com.p2p.grpc.Seller"
+    main = "com.p2p.SimulateRace"
 }
 
-task p2pBuyer(type: JavaExec) {
-    classpath sourceSets.main.runtimeClasspath
-    main = "com.p2p.grpc.Buyer"
-}
-
-task p2pPeer(type: JavaExec) {
-    classpath sourceSets.main.runtimeClasspath
-    main = "com.p2p.grpc.PeerImpl"
-}
-
-
 test {
     useJUnitPlatform()
 }