2 # IP Virtual Server configuration
4 menu "IP: Virtual Server Configuration"
5 depends on INET && NETFILTER
8 tristate "IP virtual server support (EXPERIMENTAL)"
9 depends on INET && NETFILTER
11 IP Virtual Server support will let you build a high-performance
12 virtual server based on cluster of two or more real servers. This
13 option must be enabled for at least one of the clustered computers
14 that will take care of intercepting incoming connections to a
15 single IP address and scheduling them to real servers.
17 Three request dispatching techniques are implemented, they are
18 virtual server via NAT, virtual server via tunneling and virtual
19 server via direct routing. The several scheduling algorithms can
20 be used to choose which server the connection is directed to,
21 thus load balancing can be achieved among the servers. For more
22 information and its administration program, please visit the
23 following URL: <http://www.linuxvirtualserver.org/>.
25 If you want to compile it in kernel, say Y. To compile it as a
26 module, choose M here. If unsure, say N.
29 bool "IP virtual server debugging"
32 Say Y here if you want to get additional messages useful in
33 debugging the IP virtual server code. You can change the debug
34 level in /proc/sys/net/ipv4/vs/debug_level
37 int "IPVS connection table size (the Nth power of 2)"
41 The IPVS connection hash table uses the chaining scheme to handle
42 hash collisions. Using a big IPVS connection hash table will greatly
43 reduce conflicts when there are hundreds of thousands of connections
46 Note the table size must be power of 2. The table size will be the
47 value of 2 to the your input number power. The number to choose is
48 from 8 to 20, the default number is 12, which means the table size
49 is 4096. Don't input the number too small, otherwise you will lose
50 performance on it. You can adapt the table size yourself, according
51 to your virtual server application. It is good to set the table size
52 not far less than the number of connections per second multiplying
53 average lasting time of connection in the table. For example, your
54 virtual server gets 200 connections per second, the connection lasts
55 for 200 seconds in average in the connection table, the table size
56 should be not far less than 200x200, it is good to set the table
59 Another note that each connection occupies 128 bytes effectively and
60 each hash entry uses 8 bytes, so you can estimate how much memory is
63 comment "IPVS transport protocol load balancing support"
66 config IP_VS_PROTO_TCP
67 bool "TCP load balancing support"
70 This option enables support for load balancing TCP transport
71 protocol. Say Y if unsure.
73 config IP_VS_PROTO_UDP
74 bool "UDP load balancing support"
77 This option enables support for load balancing UDP transport
78 protocol. Say Y if unsure.
80 config IP_VS_PROTO_ESP
81 bool "ESP load balancing support"
84 This option enables support for load balancing ESP (Encapsultion
85 Security Payload) transport protocol. Say Y if unsure.
88 bool "AH load balancing support"
91 This option enables support for load balancing AH (Authentication
92 Header) transport protocol. Say Y if unsure.
94 comment "IPVS scheduler"
98 tristate "round-robin scheduling"
101 The robin-robin scheduling algorithm simply directs network
102 connections to different real servers in a round-robin manner.
104 If you want to compile it in kernel, say Y. To compile it as a
105 module, choose M here. If unsure, say N.
108 tristate "weighted round-robin scheduling"
111 The weighted robin-robin scheduling algorithm directs network
112 connections to different real servers based on server weights
113 in a round-robin manner. Servers with higher weights receive
114 new connections first than those with less weights, and servers
115 with higher weights get more connections than those with less
116 weights and servers with equal weights get equal connections.
118 If you want to compile it in kernel, say Y. To compile it as a
119 module, choose M here. If unsure, say N.
122 tristate "least-connection scheduling"
125 The least-connection scheduling algorithm directs network
126 connections to the server with the least number of active
129 If you want to compile it in kernel, say Y. To compile it as a
130 module, choose M here. If unsure, say N.
133 tristate "weighted least-connection scheduling"
136 The weighted least-connection scheduling algorithm directs network
137 connections to the server with the least active connections
138 normalized by the server weight.
140 If you want to compile it in kernel, say Y. To compile it as a
141 module, choose M here. If unsure, say N.
144 tristate "locality-based least-connection scheduling"
147 The locality-based least-connection scheduling algorithm is for
148 destination IP load balancing. It is usually used in cache cluster.
149 This algorithm usually directs packet destined for an IP address to
150 its server if the server is alive and under load. If the server is
151 overloaded (its active connection numbers is larger than its weight)
152 and there is a server in its half load, then allocate the weighted
153 least-connection server to this IP address.
155 If you want to compile it in kernel, say Y. To compile it as a
156 module, choose M here. If unsure, say N.
159 tristate "locality-based least-connection with replication scheduling"
162 The locality-based least-connection with replication scheduling
163 algorithm is also for destination IP load balancing. It is
164 usually used in cache cluster. It differs from the LBLC scheduling
165 as follows: the load balancer maintains mappings from a target
166 to a set of server nodes that can serve the target. Requests for
167 a target are assigned to the least-connection node in the target's
168 server set. If all the node in the server set are over loaded,
169 it picks up a least-connection node in the cluster and adds it
170 in the sever set for the target. If the server set has not been
171 modified for the specified time, the most loaded node is removed
172 from the server set, in order to avoid high degree of replication.
174 If you want to compile it in kernel, say Y. To compile it as a
175 module, choose M here. If unsure, say N.
178 tristate "destination hashing scheduling"
181 The destination hashing scheduling algorithm assigns network
182 connections to the servers through looking up a statically assigned
183 hash table by their destination IP addresses.
185 If you want to compile it in kernel, say Y. To compile it as a
186 module, choose M here. If unsure, say N.
189 tristate "source hashing scheduling"
192 The source hashing scheduling algorithm assigns network
193 connections to the servers through looking up a statically assigned
194 hash table by their source IP addresses.
196 If you want to compile it in kernel, say Y. To compile it as a
197 module, choose M here. If unsure, say N.
200 tristate "shortest expected delay scheduling"
203 The shortest expected delay scheduling algorithm assigns network
204 connections to the server with the shortest expected delay. The
205 expected delay that the job will experience is (Ci + 1) / Ui if
206 sent to the ith server, in which Ci is the number of connections
207 on the the ith server and Ui is the fixed service rate (weight)
210 If you want to compile it in kernel, say Y. To compile it as a
211 module, choose M here. If unsure, say N.
214 tristate "never queue scheduling"
217 The never queue scheduling algorithm adopts a two-speed model.
218 When there is an idle server available, the job will be sent to
219 the idle server, instead of waiting for a fast one. When there
220 is no idle server available, the job will be sent to the server
221 that minimize its expected delay (The Shortest Expected Delay
222 scheduling algorithm).
224 If you want to compile it in kernel, say Y. To compile it as a
225 module, choose M here. If unsure, say N.
227 comment 'IPVS application helper'
231 tristate "FTP protocol helper"
232 depends on IP_VS && IP_VS_PROTO_TCP
234 FTP is a protocol that transfers IP address and/or port number in
235 the payload. In the virtual server via Network Address Translation,
236 the IP address and port number of real servers cannot be sent to
237 clients in ftp connections directly, so FTP protocol helper is
238 required for tracking the connection and mangling it back to that of
241 If you want to compile it in kernel, say Y. To compile it as a
242 module, choose M here. If unsure, say N.