AI, Machine Learning, Deep Learning 등 관련 컴퓨터 과학의 역사가 반세기가 넘다보니까,
공부를 시작하는 사람 입장에서는 어디서부터 시작해야 할지 막막하다.
원래 시작점이 수학(미분, 선형대수학, 통계학 등)이니까 대학원 수준의 수학 공부부터 해야 할까?
이렇게 하면 좋기는 하지만, 최근 몇년 사이에 TensorFlow 같은 Library가 잘 만들어져 있고 이것을 활용하는 수준으로 서비스를 개발하는 입장이라면 그냥 얇은 수학 지식만 가지고 TensorFlow를 활용하는 것이 훨씬 시간과 개발자의 에너지를 아낄 수 있다.
게다가 YouTube 영상 중에 TensorFlow 예제를 직접 보여주는 것이 있어서, 영상을 먼저 보고 공부를 시작하면 길잡이를 만난 것처럼 공부할 수 있다. (아래 영상을 보는 것을 추천 ^^)
NVIDIA TensorRT Model Optimizer
A library to quantize and compress deep learning models for optimized inference on GPUs
South Carolina 대학의 'Open Virtual Switch Lab Series' 문서를 바탕으로 내가 실습한 내용을 이곳에 정리함. ( Network namespace 개념부터 차곡차곡 쌓아 올리면서 Open vSwitch Use Case를 설명하기 때문에 공부하는 사람에게 많은 도움이 된다 )
위 그림에 묘사된 것과 같이 Network를 구성하기 위해 아래 명령을 작성했다. (따라해보면 위 그림과 똑같은 Network 만들어진다)
## root namespace에 존재하는 모든 network interface를 조회
$ ip link
## 네임스페이스 my-ns-a, my-ns-b 를 생성
$ ip netns add my-ns-a
$ ip netns add my-ns-b
## Linux kernel에 존재하는 모든 namespace 조회
$ ip netns
my-ns-b
my-ns-a
## 'my-ns-a' 네임스페이스에 존재하는 network interface 조회
$ ip netns exec my-ns-a ip link
1: lo: <LOOPBACK> mtu 65536 qdisc noop state DOWN mode DEFAULT group default qlen 1000
link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
## 가상 스위치 'sw1'를 생성
$ ovs-vsctl add-br sw1
## root namespace에 존재하는 network interface를 조회
$ ip link
... 중간 생략 ...
47: ovs-system: <BROADCAST,MULTICAST> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
link/ether 86:3d:02:69:23:4f brd ff:ff:ff:ff:ff:ff
48: sw1: <BROADCAST,MULTICAST> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
link/ether 16:68:07:5d:c0:40 brd ff:ff:ff:ff:ff:ff
## Open vSwitch에 namespace를 연결하기
## 1) veth peer 생성하기
$ ip link add my-ns-a-eth0 type veth peer name sw1-eth1
$ ip link add my-ns-b-eth0 type veth peer name sw1-eth2
$ ip link
... 중간 생략 ...
47: ovs-system: <BROADCAST,MULTICAST> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
link/ether 86:3d:02:69:23:4f brd ff:ff:ff:ff:ff:ff
48: sw1: <BROADCAST,MULTICAST> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
link/ether 16:68:07:5d:c0:40 brd ff:ff:ff:ff:ff:ff
51: sw1-eth1@my-ns-a-eth0: <BROADCAST,MULTICAST,M-DOWN> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
link/ether be:01:52:6f:4b:58 brd ff:ff:ff:ff:ff:ff
52: my-ns-a-eth0@sw1-eth1: <BROADCAST,MULTICAST,M-DOWN> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
link/ether 96:24:a4:bf:78:f3 brd ff:ff:ff:ff:ff:ff
53: sw1-eth2@my-ns-b-eth0: <BROADCAST,MULTICAST,M-DOWN> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
link/ether 46:d4:ad:57:18:20 brd ff:ff:ff:ff:ff:ff
54: my-ns-b-eth0@sw1-eth2: <BROADCAST,MULTICAST,M-DOWN> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
link/ether 2a:78:4d:57:db:37 brd ff:ff:ff:ff:ff:ff
## 2) veth peer를 각각의 namepace에 연결하기 (Attaching to namespaces)
$ ip link set my-ns-a-eth0 netns my-ns-a
$ ip link set my-ns-b-eth0 netns my-ns-b
$ ip netns exec my-ns-a ip link
1: lo: <LOOPBACK> mtu 65536 qdisc noop state DOWN mode DEFAULT group default qlen 1000
link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
52: my-ns-a-eth0@if51: <BROADCAST,MULTICAST> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
link/ether 96:24:a4:bf:78:f3 brd ff:ff:ff:ff:ff:ff link-netnsid 0
$ ip netns exec my-ns-b ip link
1: lo: <LOOPBACK> mtu 65536 qdisc noop state DOWN mode DEFAULT group default qlen 1000
link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
54: my-ns-b-eth0@if53: <BROADCAST,MULTICAST> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
link/ether 2a:78:4d:57:db:37 brd ff:ff:ff:ff:ff:ff link-netnsid 0
## 3) 가상 스위치 sw1에 veth peer를 연결하기 (Attaching veth peer to switch sw1)
$ ovs-vsctl add-port sw1 sw1-eth1
$ ovs-vsctl show
...
Bridge sw1
Port sw1
Interface sw1
type: internal
Port sw1-eth1
Interface sw1-eth1
...
$ ovs-vsctl add-port sw1 sw1-eth2
$ ovs-vsctl show
...
Bridge sw1
Port sw1
Interface sw1
type: internal
Port sw1-eth2
Interface sw1-eth2
Port sw1-eth1
Interface sw1-eth1
...
## 가상 스위치의 network port를 activate 하기. (Turning up the network port)
$ ip link set sw1-eth1 up
$ ip link set sw1-eth2 up
$ ip link
...
51: sw1-eth1@if52: <NO-CARRIER,BROADCAST,MULTICAST,UP> mtu 1500 qdisc noqueue master ovs-system state LOWERLAYERDOWN mode DEFAULT group default qlen 1000
link/ether be:01:52:6f:4b:58 brd ff:ff:ff:ff:ff:ff link-netns my-ns-a
53: sw1-eth2@if54: <NO-CARRIER,BROADCAST,MULTICAST,UP> mtu 1500 qdisc noqueue master ovs-system state LOWERLAYERDOWN mode DEFAULT group default qlen 1000
link/ether 46:d4:ad:57:18:20 brd ff:ff:ff:ff:ff:ff link-netns my-ns-b
...
## 각각의 namespace에 IP address를 할당하기
$ ip netns exec my-ns-a ip link set dev my-ns-a-eth0 up
$ ip netns exec my-ns-b ip link set dev my-ns-b-eth0 up
$ ip netns exec my-ns-a ip address add 192.168.1.10/24 dev my-ns-a-eth0
$ ip netns exec my-ns-b ip address add 192.168.1.20/24 dev my-ns-b-eth0
## 설정 정보 확인하기
$ ip netns exec my-ns-a ip addr
1: lo: <LOOPBACK> mtu 65536 qdisc noop state DOWN group default qlen 1000
link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
52: my-ns-a-eth0@if51: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000
link/ether 96:24:a4:bf:78:f3 brd ff:ff:ff:ff:ff:ff link-netnsid 0
inet 192.168.1.10/24 scope global my-ns-a-eth0
valid_lft forever preferred_lft forever
inet6 fe80::9424:a4ff:febf:78f3/64 scope link
valid_lft forever preferred_lft forever
$ ip netns exec my-ns-b ip addr
1: lo: <LOOPBACK> mtu 65536 qdisc noop state DOWN group default qlen 1000
link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
54: my-ns-b-eth0@if53: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UP group default qlen 1000
link/ether 2a:78:4d:57:db:37 brd ff:ff:ff:ff:ff:ff link-netnsid 0
inet 192.168.1.20/24 scope global my-ns-b-eth0
valid_lft forever preferred_lft forever
inet6 fe80::2878:4dff:fe57:db37/64 scope link
valid_lft forever preferred_lft forever
## namespace 'my-ns-a'의 routing table 확인하기
$ ip netns exec my-ns-a ip route
192.168.1.0/24 dev my-ns-a-eth0 proto kernel scope link src 192.168.1.10
## namespace 'my-ns-b'의 routing table 확인하기
$ ip netns exec my-ns-b ip route
192.168.1.0/24 dev my-ns-b-eth0 proto kernel scope link src 192.168.1.20
## namespace 'my-ns-a'에서 bash shell 시작하기
$ ip netns exec my-ns-a bash
$ ifconfig
my-ns-a-eth0: flags=4163<UP,BROADCAST,RUNNING,MULTICAST> mtu 1500
inet 192.168.1.10 netmask 255.255.255.0 broadcast 0.0.0.0
inet6 fe80::9424:a4ff:febf:78f3 prefixlen 64 scopeid 0x20<link>
ether 96:24:a4:bf:78:f3 txqueuelen 1000 (Ethernet)
RX packets 86 bytes 21517 (21.5 KB)
RX errors 0 dropped 0 overruns 0 frame 0
TX packets 13 bytes 1006 (1.0 KB)
TX errors 0 dropped 0 overruns 0 carrier 0 collisions 0
$ ping 192.168.1.20 -c 2
PING 192.168.1.20 (192.168.1.20) 56(84) bytes of data.
64 bytes from 192.168.1.20: icmp_seq=1 ttl=64 time=0.088 ms
64 bytes from 192.168.1.20: icmp_seq=2 ttl=64 time=0.079 ms
--- 192.168.1.20 ping statistics ---
2 packets transmitted, 2 received, 0% packet loss, time 1007ms
rtt min/avg/max/mdev = 0.079/0.083/0.088/0.004 ms
$ traceroute 192.168.1.20
traceroute to 192.168.1.20 (192.168.1.20), 64 hops max
1 192.168.1.20 0.452ms 0.003ms 0.002ms