47 | | [[BR]] |
48 | | == 5.Kerrighed modules - Source Code Structure Analysis == |
49 | | === 5.1 Kerrighed 0.70 Totally 7 Modules === |
50 | | [[Image(krg0.7arch.png)]] |
51 | | * Kerrighed extension of Linux Kernel |
52 | | * Process & load balancing - '''Aragorn''' |
53 | | * Containers - '''Gandalf''' |
54 | | * Synchronization - '''Elrond''' |
55 | | * Ghosts - '''Nazgul''' |
56 | | * Communication - '''Gimli''', '''Gloin''' |
57 | | * Tools - '''lluvatar''' |
58 | | |
59 | | ==== 5.1.1 Probably the Key modules for this project ==== |
60 | | |
61 | | * '''Gandalf module''': Global memory management -> ('''For Container''') |
62 | | * /kerrighed-2.3.0/libs/include/kerrighed_tools.h |
63 | | * /kerrighed-2.3.0/modules/tools/krg_services.h |
64 | | * /kerrighed-2.3.0/modules/mm/mmap_handler.h[[BR]] |
65 | | |
66 | | '''/* Memory Part -> Distributed Shared Memory (DSM) , MPI&OpenMP */''' |
67 | | |
68 | | * '''Aragon module''': Global process management -> ('''For Migration''') |
69 | | * /kerrighed-2.3.0/modules/proc/distant_syscalls.c |
70 | | * /kerrighed-2.3.0/modules/scheduler/mosix_probe.c |
71 | | * /kerrighed-2.3.0/modules/tools/krg_arch.h |
72 | | * /kerrighed-2.3.0/modules/rpc/rpcid.h |
73 | | * /kerrighed-2.3.0/modules/epm/fork_delay.c |
74 | | * /kerrighed-2.3.0/modules/epm/migration.c |
75 | | * /kerrighed-2.3.0/modules/epm/procfs.c[[BR]] |
76 | | |
77 | | '''/* Process & Load balancing Part -> MPI , OpenMP */''' [[BR]] |
78 | | |
79 | | '''Process migration involves two key phases:''' [[BR]] |
80 | | (1) migration of the process execution context [[BR]] |
81 | | (2) maintenance of links between the migrated process and the resources [[BR]] |
82 | | |
83 | | First task: '''Nazgul module''' ---> second task: '''Gandalf module''' -> container & linker |
84 | | |
85 | | * '''Elrond module''': Synchronization tools for parallel applications -> ('''For Synchronization''') |
86 | | * /kerrighed-2.3.0/modules/epm/ghost_process_api.h[[BR]] |
87 | | |
88 | | '''/* Synchronization Part -> MPI , OpenMP */''' [[BR]] |
89 | | |
90 | | {{{ |
91 | | Elrond provides distributed locks, barriers, semaphores, and wait condition to processes in a implementation that |
92 | | supports process checkpointing as well as migration between nodes. |
93 | | }}} |
94 | | |
95 | | * '''Gimli''': |
96 | | * /kerrighed-2.3.0/libs/libkerrighed/libcomm.c |
97 | | * /kerrighed-2.3.0/libs/include/comm.h |
98 | | * /kerrighed-2.3.0/libs/include/kerrighed_tools.h |
99 | | * /kerrighed-2.3.0/modules/arch/um/krg_um.c |
100 | | * /kerrighed-2.3.0/tools/krg_services.h |
101 | | * /kerrighed-2.3.0/tools/krg_arch.h |
102 | | * /kerrighed-2.3.0/modules/fs/kerfs/dir_io_linker.c |
103 | | * /kerrighed-2.3.0/modules/fs/kerfs/inode_io_linker.c |
104 | | * /kerrighed-2.3.0/modules/fs/kerfs/file_io_linker.c [[BR]] |
105 | | |
106 | | |
107 | | |
108 | | {{{ |
109 | | All these distributed services are based on Gimli for their communications. Gimli is a service providing high |
110 | | performance communication inside the cluster and a complete abstraction of the communication system to kerrighed that |
111 | | makes it portable on various interconnection networks. |
112 | | }}} |
113 | | |
114 | | |
115 | | |
116 | | [[BR]] |
117 | | === 5.2 Kerrighed 2.3.0 Totally 14 Modules === |
118 | | [[Image(KerrighedArch.png)]] |
119 | | || RPC || || |
120 | | || Hotplug || || |
121 | | || KDDM || || |
122 | | || Capability || || |
123 | | || ProcFS || || |
124 | | || Sync || || |
125 | | || IPC || || |
126 | | || FS || || |
127 | | || MM || || |
128 | | || Dynamic Streams || || |
129 | | || Ghost || || |
130 | | || Proc || || |
131 | | || EPM || || |
132 | | || Global Scheduler || || |
133 | | * Load Kerrighed Moudles Messages |
134 | | {{{ |
135 | | $ dmesg |
136 | | Start loading Kerrighed... |
137 | | Init Kerrighed worker(s)...Init Kerrighed low-level framework... |
138 | | Init kerrighed syscall mechanism |
139 | | Kerrighed tools - init module |
140 | | TIPC: Started in network mode |
141 | | TIPC: Own node address <1.1.2>, network identity 1 |
142 | | Try to enable bearer on lo:<5>TIPC: Enabled bearer <eth:lo>, discovery domain <1.1.0>, priority 10 |
143 | | ok |
144 | | Try to enable bearer on eth0:<5>TIPC: Enabled bearer <eth:eth0>, discovery domain <1.1.0>, priority 10 |
145 | | ok |
146 | | Try to enable bearer on eth1:<4>TIPC: Bearer <eth:eth1> priority adjustment required 10->9 |
147 | | TIPC: Enabled bearer <eth:eth1>, discovery domain <1.1.0>, priority 9 |
148 | | ok |
149 | | RPC initialisation done |
150 | | Init Kerrighed low-level framework (nodeid 1) : done |
151 | | Init Kerrighed distributed services... |
152 | | KDDM initialisation : start |
153 | | KDDM set init |
154 | | KDDM set init : done |
155 | | KDDM initialisation done |
156 | | KerMM initialisation : start |
157 | | KerMM initialisation done |
158 | | DVFS initialisation : start |
159 | | FAF: initialisation : start |
160 | | FAF: initialisation : done |
161 | | DVFS initialisation done |
162 | | KerIPC initialisation : start |
163 | | KerIPC initialisation done |
164 | | Proc initialisation: start |
165 | | Proc initialisation: done |
166 | | EPM initialisation: start |
167 | | EPM initialisation: done |
168 | | Init Kerrighed distributed services: done |
169 | | Kerrighed... loaded! |
170 | | TIPC: Established link <1.1.2:eth0-1.1.3:eth0> on network plane B |
171 | | Kerrighed is running on 2 nodes |
172 | | |
173 | | }}} |
174 | | [[BR]] |
175 | | === 5.3 Private Variables === |
176 | | * '''mmap''' (map pages of memory - by adding a MAP_LOCAL flag) |
177 | | * /kerrighed-2.3.0/libs/include/kerrighed_tools.h |
178 | | * /kerrighed-2.3.0/modules/tools/krg_services.h |
179 | | * /kerrighed-2.3.0/modules/mm/mman.h |
180 | | |
181 | | {{{ |
182 | | It is possible to ensure that memory will be allocated in the virtual address space of each thread of the process |
183 | | without linking it to a container, thus ensuring that no coherence will be maintained between the different copies of |
184 | | that memory segment across the different threads. |
185 | | }}} |
186 | | |
187 | | [[BR]] |
188 | | === 5.4 OpenMP on Kerrighed === |
189 | | * Standard '''Pthread''' library replaced by '''krgthread''' library proving the Posix interface on Kerrighed Clusters. |
190 | | * Containers to support shared variables in the cluster. |
191 | | * A thread using kerrighed synchronization primitives can migrate and be checkpointed at any time, and except when |
192 | | blocked in the OS. |