1 # SPDX-License-Identifier: GPL-2.0-only
3 # IP Virtual Server configuration
6 tristate "IP virtual server support"
7 depends on NET && INET && NETFILTER
8 depends on (NF_CONNTRACK || NF_CONNTRACK=n)
10 IP Virtual Server support will let you build a high-performance
11 virtual server based on cluster of two or more real servers. This
12 option must be enabled for at least one of the clustered computers
13 that will take care of intercepting incoming connections to a
14 single IP address and scheduling them to real servers.
16 Three request dispatching techniques are implemented, they are
17 virtual server via NAT, virtual server via tunneling and virtual
18 server via direct routing. The several scheduling algorithms can
19 be used to choose which server the connection is directed to,
20 thus load balancing can be achieved among the servers. For more
21 information and its administration program, please visit the
22 following URL: <http://www.linuxvirtualserver.org/>.
24 If you want to compile it in kernel, say Y. To compile it as a
25 module, choose M here. If unsure, say N.
30 bool "IPv6 support for IPVS"
31 depends on IPV6 = y || IP_VS = IPV6
32 select IP6_NF_IPTABLES
35 Add IPv6 support to IPVS.
40 bool "IP virtual server debugging"
42 Say Y here if you want to get additional messages useful in
43 debugging the IP virtual server code. You can change the debug
44 level in /proc/sys/net/ipv4/vs/debug_level
47 int "IPVS connection table size (the Nth power of 2)"
51 The IPVS connection hash table uses the chaining scheme to handle
52 hash collisions. Using a big IPVS connection hash table will greatly
53 reduce conflicts when there are hundreds of thousands of connections
56 Note the table size must be power of 2. The table size will be the
57 value of 2 to the your input number power. The number to choose is
58 from 8 to 20, the default number is 12, which means the table size
59 is 4096. Don't input the number too small, otherwise you will lose
60 performance on it. You can adapt the table size yourself, according
61 to your virtual server application. It is good to set the table size
62 not far less than the number of connections per second multiplying
63 average lasting time of connection in the table. For example, your
64 virtual server gets 200 connections per second, the connection lasts
65 for 200 seconds in average in the connection table, the table size
66 should be not far less than 200x200, it is good to set the table
69 Another note that each connection occupies 128 bytes effectively and
70 each hash entry uses 8 bytes, so you can estimate how much memory is
73 You can overwrite this number setting conn_tab_bits module parameter
74 or by appending ip_vs.conn_tab_bits=? to the kernel command line
75 if IP VS was compiled built-in.
77 comment "IPVS transport protocol load balancing support"
79 config IP_VS_PROTO_TCP
80 bool "TCP load balancing support"
82 This option enables support for load balancing TCP transport
83 protocol. Say Y if unsure.
85 config IP_VS_PROTO_UDP
86 bool "UDP load balancing support"
88 This option enables support for load balancing UDP transport
89 protocol. Say Y if unsure.
91 config IP_VS_PROTO_AH_ESP
92 def_bool IP_VS_PROTO_ESP || IP_VS_PROTO_AH
94 config IP_VS_PROTO_ESP
95 bool "ESP load balancing support"
97 This option enables support for load balancing ESP (Encapsulation
98 Security Payload) transport protocol. Say Y if unsure.
100 config IP_VS_PROTO_AH
101 bool "AH load balancing support"
103 This option enables support for load balancing AH (Authentication
104 Header) transport protocol. Say Y if unsure.
106 config IP_VS_PROTO_SCTP
107 bool "SCTP load balancing support"
110 This option enables support for load balancing SCTP transport
111 protocol. Say Y if unsure.
113 comment "IPVS scheduler"
116 tristate "round-robin scheduling"
118 The robin-robin scheduling algorithm simply directs network
119 connections to different real servers in a round-robin manner.
121 If you want to compile it in kernel, say Y. To compile it as a
122 module, choose M here. If unsure, say N.
125 tristate "weighted round-robin scheduling"
127 The weighted robin-robin scheduling algorithm directs network
128 connections to different real servers based on server weights
129 in a round-robin manner. Servers with higher weights receive
130 new connections first than those with less weights, and servers
131 with higher weights get more connections than those with less
132 weights and servers with equal weights get equal connections.
134 If you want to compile it in kernel, say Y. To compile it as a
135 module, choose M here. If unsure, say N.
138 tristate "least-connection scheduling"
140 The least-connection scheduling algorithm directs network
141 connections to the server with the least number of active
144 If you want to compile it in kernel, say Y. To compile it as a
145 module, choose M here. If unsure, say N.
148 tristate "weighted least-connection scheduling"
150 The weighted least-connection scheduling algorithm directs network
151 connections to the server with the least active connections
152 normalized by the server weight.
154 If you want to compile it in kernel, say Y. To compile it as a
155 module, choose M here. If unsure, say N.
158 tristate "weighted failover scheduling"
160 The weighted failover scheduling algorithm directs network
161 connections to the server with the highest weight that is
164 If you want to compile it in kernel, say Y. To compile it as a
165 module, choose M here. If unsure, say N.
168 tristate "weighted overflow scheduling"
170 The weighted overflow scheduling algorithm directs network
171 connections to the server with the highest weight that is
172 currently available and overflows to the next when active
173 connections exceed the node's weight.
175 If you want to compile it in kernel, say Y. To compile it as a
176 module, choose M here. If unsure, say N.
179 tristate "locality-based least-connection scheduling"
181 The locality-based least-connection scheduling algorithm is for
182 destination IP load balancing. It is usually used in cache cluster.
183 This algorithm usually directs packet destined for an IP address to
184 its server if the server is alive and under load. If the server is
185 overloaded (its active connection numbers is larger than its weight)
186 and there is a server in its half load, then allocate the weighted
187 least-connection server to this IP address.
189 If you want to compile it in kernel, say Y. To compile it as a
190 module, choose M here. If unsure, say N.
193 tristate "locality-based least-connection with replication scheduling"
195 The locality-based least-connection with replication scheduling
196 algorithm is also for destination IP load balancing. It is
197 usually used in cache cluster. It differs from the LBLC scheduling
198 as follows: the load balancer maintains mappings from a target
199 to a set of server nodes that can serve the target. Requests for
200 a target are assigned to the least-connection node in the target's
201 server set. If all the node in the server set are over loaded,
202 it picks up a least-connection node in the cluster and adds it
203 in the sever set for the target. If the server set has not been
204 modified for the specified time, the most loaded node is removed
205 from the server set, in order to avoid high degree of replication.
207 If you want to compile it in kernel, say Y. To compile it as a
208 module, choose M here. If unsure, say N.
211 tristate "destination hashing scheduling"
213 The destination hashing scheduling algorithm assigns network
214 connections to the servers through looking up a statically assigned
215 hash table by their destination IP addresses.
217 If you want to compile it in kernel, say Y. To compile it as a
218 module, choose M here. If unsure, say N.
221 tristate "source hashing scheduling"
223 The source hashing scheduling algorithm assigns network
224 connections to the servers through looking up a statically assigned
225 hash table by their source IP addresses.
227 If you want to compile it in kernel, say Y. To compile it as a
228 module, choose M here. If unsure, say N.
231 tristate "maglev hashing scheduling"
233 The maglev consistent hashing scheduling algorithm provides the
234 Google's Maglev hashing algorithm as a IPVS scheduler. It assigns
235 network connections to the servers through looking up a statically
236 assigned special hash table called the lookup table. Maglev hashing
237 is to assign a preference list of all the lookup table positions
240 Through this operation, The maglev hashing gives an almost equal
241 share of the lookup table to each of the destinations and provides
242 minimal disruption by using the lookup table. When the set of
243 destinations changes, a connection will likely be sent to the same
244 destination as it was before.
246 If you want to compile it in kernel, say Y. To compile it as a
247 module, choose M here. If unsure, say N.
250 tristate "shortest expected delay scheduling"
252 The shortest expected delay scheduling algorithm assigns network
253 connections to the server with the shortest expected delay. The
254 expected delay that the job will experience is (Ci + 1) / Ui if
255 sent to the ith server, in which Ci is the number of connections
256 on the ith server and Ui is the fixed service rate (weight)
259 If you want to compile it in kernel, say Y. To compile it as a
260 module, choose M here. If unsure, say N.
263 tristate "never queue scheduling"
265 The never queue scheduling algorithm adopts a two-speed model.
266 When there is an idle server available, the job will be sent to
267 the idle server, instead of waiting for a fast one. When there
268 is no idle server available, the job will be sent to the server
269 that minimize its expected delay (The Shortest Expected Delay
270 scheduling algorithm).
272 If you want to compile it in kernel, say Y. To compile it as a
273 module, choose M here. If unsure, say N.
275 comment 'IPVS SH scheduler'
277 config IP_VS_SH_TAB_BITS
278 int "IPVS source hashing table size (the Nth power of 2)"
282 The source hashing scheduler maps source IPs to destinations
283 stored in a hash table. This table is tiled by each destination
284 until all slots in the table are filled. When using weights to
285 allow destinations to receive more connections, the table is
286 tiled an amount proportional to the weights specified. The table
287 needs to be large enough to effectively fit all the destinations
288 multiplied by their respective weights.
290 comment 'IPVS MH scheduler'
292 config IP_VS_MH_TAB_INDEX
293 int "IPVS maglev hashing table index of size (the prime numbers)"
297 The maglev hashing scheduler maps source IPs to destinations
298 stored in a hash table. This table is assigned by a preference
299 list of the positions to each destination until all slots in
300 the table are filled. The index determines the prime for size of
301 the table as 251, 509, 1021, 2039, 4093, 8191, 16381, 32749,
302 65521 or 131071. When using weights to allow destinations to
303 receive more connections, the table is assigned an amount
304 proportional to the weights specified. The table needs to be large
305 enough to effectively fit all the destinations multiplied by their
308 comment 'IPVS application helper'
311 tristate "FTP protocol helper"
312 depends on IP_VS_PROTO_TCP && NF_CONNTRACK && NF_NAT && \
316 FTP is a protocol that transfers IP address and/or port number in
317 the payload. In the virtual server via Network Address Translation,
318 the IP address and port number of real servers cannot be sent to
319 clients in ftp connections directly, so FTP protocol helper is
320 required for tracking the connection and mangling it back to that of
323 If you want to compile it in kernel, say Y. To compile it as a
324 module, choose M here. If unsure, say N.
327 bool "Netfilter connection tracking"
328 depends on NF_CONNTRACK
330 The Netfilter connection tracking support allows the IPVS
331 connection state to be exported to the Netfilter framework
332 for filtering purposes.
335 tristate "SIP persistence engine"
336 depends on IP_VS_PROTO_UDP
337 depends on NF_CONNTRACK_SIP
339 Allow persistence based on the SIP Call-ID