system:benches10gbps:loadbalancing
Differences
This shows you the differences between two versions of the page.
| Next revision | Previous revision | ||
|
system:benches10gbps:loadbalancing [2012/10/16 12:33] ze created |
system:benches10gbps:loadbalancing [2012/10/17 17:35] (current) ze |
||
|---|---|---|---|
| Line 2: | Line 2: | ||
| allowing us to get 800k connections/s (2 clients, 3 servers) without | allowing us to get 800k connections/s (2 clients, 3 servers) without | ||
| loadbalancing them. | loadbalancing them. | ||
| - | |||
| - | FIXME: graph not available yet. Will wait until the bench are over. | ||
| Monitoring graphs for the different benches can be found | Monitoring graphs for the different benches can be found | ||
| Line 129: | Line 127: | ||
| 40-45k conn/s. | 40-45k conn/s. | ||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit12/rempart-firewall/interfaces_eth1_bps.png|bps]] | ||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit12/rempart-firewall/interfaces_eth1_pkt.png|pkt]] | ||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit12/rempart-firewall/ipvs_conn.png|conn per sec]] | ||
| ===== ipvs_tab_bits ===== | ===== ipvs_tab_bits ===== | ||
| Line 143: | Line 144: | ||
| That give us some better overall performances, with like 350k conn/s, | That give us some better overall performances, with like 350k conn/s, | ||
| but there is some weird downfall some time in the beginning. | but there is some weird downfall some time in the beginning. | ||
| + | |||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit20/rempart-firewall/interfaces_eth1_bps.png|bps]] | ||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit20/rempart-firewall/interfaces_eth1_pkt.png|pkt]] | ||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit20/rempart-firewall/ipvs_conn.png|conn per sec]] | ||
| + | |||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit20b/rempart-firewall/interfaces_eth1_bps.png|bps]] | ||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit20b/rempart-firewall/interfaces_eth1_pkt.png|pkt]] | ||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit20b/rempart-firewall/ipvs_conn.png|conn per sec]] | ||
| + | |||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit20c/rempart-firewall/interfaces_eth1_bps.png|bps]] | ||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit20c/rempart-firewall/interfaces_eth1_pkt.png|pkt]] | ||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit20c/rempart-firewall/ipvs_conn.png|conn per sec]] | ||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit20c/rempart-firewall/ipvs_cur.png|ipvs connection tracking]] | ||
| + | |||
| ===== smp affinity ===== | ===== smp affinity ===== | ||
| Line 154: | Line 169: | ||
| get on the same 8 threads, but the performances is more stable. and gets | get on the same 8 threads, but the performances is more stable. and gets | ||
| stable around 280k conn/s. | stable around 280k conn/s. | ||
| + | |||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit20-af | ||
| + | finity-a/rempart-firewall/interfaces_eth1_bps.png|bps]] | ||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit20-af | ||
| + | finity-a/rempart-firewall/interfaces_eth1_pkt.png|pkt]] | ||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit20-af | ||
| + | finity-a/rempart-firewall/ipvs_conn.png|conn per sec]] | ||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit20-af | ||
| + | finity-a/rempart-firewall/ipvs_cur.png|ipvs connection tracking]] | ||
| Line 159: | Line 183: | ||
| 6-9, 18-21), we use more threads, and get better performances, stable | 6-9, 18-21), we use more threads, and get better performances, stable | ||
| around 370k/s. | around 370k/s. | ||
| + | |||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit20-af | ||
| + | finity-b/rempart-firewall/interfaces_eth1_bps.png|bps]] | ||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit20-af | ||
| + | finity-b/rempart-firewall/interfaces_eth1_pkt.png|pkt]] | ||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit20-af | ||
| + | finity-b/rempart-firewall/ipvs_conn.png|conn per sec]] | ||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit20-af | ||
| + | finity-b/rempart-firewall/ipvs_cur.png|ipvs connection tracking]] | ||
| If we change cpu before changing thread (0-3, 6-9, 12-15, 18-21), the | If we change cpu before changing thread (0-3, 6-9, 12-15, 18-21), the | ||
| performances are lightly better, around 380k/s. | performances are lightly better, around 380k/s. | ||
| + | |||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit20-af | ||
| + | finity-c/rempart-firewall/interfaces_eth1_bps.png|bps]] | ||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit20-affinity-c/rempart-firewall/interfaces_eth1_pkt.png|pkt]] | ||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit20-affinity-c/rempart-firewall/ipvs_conn.png|conn per sec]] | ||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit20-affinity-c/rempart-firewall/ipvs_cur.png|ipvs connection tracking]] | ||
| ===== higher hash size ===== | ===== higher hash size ===== | ||
| Line 169: | Line 208: | ||
| + | with 24 bits: | ||
| + | |||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit24-affinity-c/rempart-firewall/interfaces_eth1_bps.png|bps]] | ||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit24-affinity-c/rempart-firewall/interfaces_eth1_pkt.png|pkt]] | ||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit24-affinity-c/rempart-firewall/ipvs_conn.png|conn per sec]] | ||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit24-affinity-c/rempart-firewall/ipvs_cur.png|ipvs connection tracking]] | ||
| + | |||
| + | with 28 bits (same test, twice) : | ||
| + | |||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit28-affinity-c-bis/rempart-firewall/interfaces_eth1_bps.png|bps]] | ||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit28-affinity-c-bis/rempart-firewall/interfaces_eth1_pkt.png|pkt]] | ||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit28-affinity-c-bis/rempart-firewall/ipvs_conn.png|conn per sec]] | ||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit28-affinity-c-bis/rempart-firewall/ipvs_cur.png|ipvs connection tracking]] | ||
| + | |||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit28-affinity-c/rempart-firewall/interfaces_eth1_bps.png|bps]] | ||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit28-affinity-c/rempart-firewall/interfaces_eth1_pkt.png|pkt]] | ||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit28-affinity-c/rempart-firewall/ipvs_conn.png|conn per sec]] | ||
| + | [[http://www.hagtheil.net/files/system/benches10gbps/loadbalancing/ipvs-bit28-affinity-c/rempart-firewall/ipvs_cur.png|ipvs connection tracking]] | ||
| + | |||
| + | ====== Conclusion ====== | ||
| + | |||
| + | * As always, affinity helps. In this case, not only with performance, but with stability. | ||
| + | * if you have alot of connection to balance, increase the bits to at least 20. It doesn't take much memory, and it helps alot. | ||
system/benches10gbps/loadbalancing.1350390794.txt.gz ยท Last modified: 2012/10/16 12:33 by ze
Page Tools
Except where otherwise noted, content on this wiki is licensed under the following license: CC Attribution-Noncommercial-Share Alike 4.0 International
