Release 4.8 net/netfilter/ipvs/ip_vs_sed.c
/*
* IPVS: Shortest Expected Delay scheduling module
*
* Authors: Wensong Zhang <wensong@linuxvirtualserver.org>
*
* This program is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License
* as published by the Free Software Foundation; either version
* 2 of the License, or (at your option) any later version.
*
* Changes:
*
*/
/*
* The SED algorithm attempts to minimize each job's expected delay until
* completion. The expected delay that the job will experience is
* (Ci + 1) / Ui if sent to the ith server, in which Ci is the number of
* jobs on the ith server and Ui is the fixed service rate (weight) of
* the ith server. The SED algorithm adopts a greedy policy that each does
* what is in its own best interest, i.e. to join the queue which would
* minimize its expected delay of completion.
*
* See the following paper for more information:
* A. Weinrib and S. Shenker, Greed is not enough: Adaptive load sharing
* in large heterogeneous systems. In Proceedings IEEE INFOCOM'88,
* pages 986-994, 1988.
*
* Thanks must go to Marko Buuri <marko@buuri.name> for talking SED to me.
*
* The difference between SED and WLC is that SED includes the incoming
* job in the cost function (the increment of 1). SED may outperform
* WLC, while scheduling big jobs under larger heterogeneous systems
* (the server weight varies a lot).
*
*/
#define KMSG_COMPONENT "IPVS"
#define pr_fmt(fmt) KMSG_COMPONENT ": " fmt
#include <linux/module.h>
#include <linux/kernel.h>
#include <net/ip_vs.h>
static inline int
ip_vs_sed_dest_overhead(struct ip_vs_dest *dest)
{
/*
* We only use the active connection number in the cost
* calculation here.
*/
return atomic_read(&dest->activeconns) + 1;
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
wensong zhang | wensong zhang | 24 | 100.00% | 1 | 100.00% |
| Total | 24 | 100.00% | 1 | 100.00% |
/*
* Weighted Least Connection scheduling
*/
static struct ip_vs_dest *
ip_vs_sed_schedule(struct ip_vs_service *svc, const struct sk_buff *skb,
struct ip_vs_iphdr *iph)
{
struct ip_vs_dest *dest, *least;
int loh, doh;
IP_VS_DBG(6, "%s(): Scheduling...\n", __func__);
/*
* We calculate the load of each dest server as follows:
* (server expected overhead) / dest->weight
*
* Remember -- no floats in kernel mode!!!
* The comparison of h1*w2 > h2*w1 is equivalent to that of
* h1/w1 > h2/w2
* if every weight is larger than zero.
*
* The server with weight=0 is quiesced and will not receive any
* new connections.
*/
list_for_each_entry_rcu(dest, &svc->destinations, n_list) {
if (!(dest->flags & IP_VS_DEST_F_OVERLOAD) &&
atomic_read(&dest->weight) > 0) {
least = dest;
loh = ip_vs_sed_dest_overhead(least);
goto nextstage;
}
}
ip_vs_scheduler_err(svc, "no destination available");
return NULL;
/*
* Find the destination with the least load.
*/
nextstage:
list_for_each_entry_continue_rcu(dest, &svc->destinations, n_list) {
if (dest->flags & IP_VS_DEST_F_OVERLOAD)
continue;
doh = ip_vs_sed_dest_overhead(dest);
if ((__s64)loh * atomic_read(&dest->weight) >
(__s64)doh * atomic_read(&least->weight)) {
least = dest;
loh = doh;
}
}
IP_VS_DBG_BUF(6, "SED: server %s:%u "
"activeconns %d refcnt %d weight %d overhead %d\n",
IP_VS_DBG_ADDR(least->af, &least->addr),
ntohs(least->port),
atomic_read(&least->activeconns),
atomic_read(&least->refcnt),
atomic_read(&least->weight), loh);
return least;
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
wensong zhang | wensong zhang | 179 | 78.85% | 1 | 8.33% |
julian anastasov | julian anastasov | 18 | 7.93% | 5 | 41.67% |
julius volz | julius volz | 7 | 3.08% | 1 | 8.33% |
stephen hemminger | stephen hemminger | 7 | 3.08% | 1 | 8.33% |
simon kirby | simon kirby | 6 | 2.64% | 1 | 8.33% |
patrick schaaf | patrick schaaf | 4 | 1.76% | 1 | 8.33% |
hannes eder | hannes eder | 3 | 1.32% | 1 | 8.33% |
simon horman | simon horman | 3 | 1.32% | 1 | 8.33% |
| Total | 227 | 100.00% | 12 | 100.00% |
static struct ip_vs_scheduler ip_vs_sed_scheduler =
{
.name = "sed",
.refcnt = ATOMIC_INIT(0),
.module = THIS_MODULE,
.n_list = LIST_HEAD_INIT(ip_vs_sed_scheduler.n_list),
.schedule = ip_vs_sed_schedule,
};
static int __init ip_vs_sed_init(void)
{
return register_ip_vs_scheduler(&ip_vs_sed_scheduler);
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
wensong zhang | wensong zhang | 16 | 100.00% | 1 | 100.00% |
| Total | 16 | 100.00% | 1 | 100.00% |
static void __exit ip_vs_sed_cleanup(void)
{
unregister_ip_vs_scheduler(&ip_vs_sed_scheduler);
synchronize_rcu();
}
Contributors
| Person | Tokens | Prop | Commits | CommitProp |
wensong zhang | wensong zhang | 15 | 83.33% | 1 | 50.00% |
julian anastasov | julian anastasov | 3 | 16.67% | 1 | 50.00% |
| Total | 18 | 100.00% | 2 | 100.00% |
module_init(ip_vs_sed_init);
module_exit(ip_vs_sed_cleanup);
MODULE_LICENSE("GPL");
Overall Contributors
| Person | Tokens | Prop | Commits | CommitProp |
wensong zhang | wensong zhang | 289 | 79.61% | 1 | 5.88% |
julian anastasov | julian anastasov | 21 | 5.79% | 6 | 35.29% |
hannes eder | hannes eder | 14 | 3.86% | 2 | 11.76% |
sven wegener | sven wegener | 10 | 2.75% | 1 | 5.88% |
julius volz | julius volz | 7 | 1.93% | 1 | 5.88% |
stephen hemminger | stephen hemminger | 7 | 1.93% | 1 | 5.88% |
simon kirby | simon kirby | 6 | 1.65% | 1 | 5.88% |
patrick schaaf | patrick schaaf | 4 | 1.10% | 1 | 5.88% |
simon horman | simon horman | 3 | 0.83% | 1 | 5.88% |
adrian bunk | adrian bunk | 1 | 0.28% | 1 | 5.88% |
michael opdenacker | michael opdenacker | 1 | 0.28% | 1 | 5.88% |
| Total | 363 | 100.00% | 17 | 100.00% |
Information contained on this website is for historical information purposes only and does not indicate or represent copyright ownership.