Multi-gigabit bandwidth without LACP

Danic

Member
Feb 6, 2015
73
27
18
jrdm.us
Is there a way to have incoming load balancing without LACP? I have a linux file server (samba 4.1.6) that gets bombarded with daily backups at 5pm every day. I know the disk array holding backups can push greater than gigabit throughput. I have extra gigabit port on this server and can easily add an additional NICs. But I only have unmanaged switch at my disposal. I've looked into some linux methods for network load balancing, but the round robin method only effects outgoing traffic. SMB Multichannel looks promising but samba doesn't support it and my clients are Windows 7. What else could I do? I rather not buy any additional hardware, otherwise I would just order up a LCAP capable switch.

As a side note: I'm the only one bothered by the fact the backups take hours instead of 10's of minutes because the network pipe is full.

Thanks for any suggestions.
 

Chuckleb

Moderator
Mar 5, 2013
1,017
330
83
Minnesota
Would a quick hack be to run multiple nics with different addresses and have backups go to designated IPs? Basically a manual load balancing. Not ideal but quick, cheap, and dirty.
 
  • Like
Reactions: Continuum

Danic

Member
Feb 6, 2015
73
27
18
jrdm.us
Good idea. Its only 18 workstations.
Just came to me: I am running pfsense router with unbound DNS. Is there a round-robin DNS lookup? Give the entry 'backup-server.local' IP's of .1 .2 and .3 ? I'll do some googling and report back.
 

EffrafaxOfWug

Radioactive Member
Feb 12, 2015
1,232
419
83
Have you tried setting up a bond using balance-alb? As long as you have supported NICs it's meant to load balance streams across different NICs.
 

Stanza

Active Member
Jan 11, 2014
205
41
28
Stagger the backups for the clients?

Eg have them fire off 10minutes apart?
 

canta

Well-Known Member
Nov 26, 2014
1,029
198
63
39
Good idea. Its only 18 workstations.
Just came to me: I am running pfsense router with unbound DNS. Is there a round-robin DNS lookup? Give the entry 'backup-server.local' IP's of .1 .2 and .3 ? I'll do some googling and report back.
if you are having lacp switch, it is easy to do LAG

as already mention, if you do not have LACP switch capabl, NIC bonding is the quick way

somehow redhat introduces nic teaming, I tried and managed to make it work with hassle.

I propose:
1) LACP if possible
2) or NIC bonding, this is very easy, and do not forget to pick "load balancing_alb".

by playing DNS, and two IP addresses.: this is a headache for troubleshooting in the future when something goes wacky. my rule: KIS ( Keep It Simple)
 

maze

Active Member
Apr 27, 2013
556
84
28
multiple nic's and do a round-rubin of the dns name. Should be a quickfix.

Alternatively, get a cheap managed switch - like a hp 1810g (i think its called) it'll do LACP and wont set you back a lot.
 

Danic

Member
Feb 6, 2015
73
27
18
jrdm.us
Thanks for all the suggestions
Stagger the backups for the clients?
I could add a delay. The backup script is ran at shutdown, and everyone leaves at 5pm.

Have you tried setting up a bond using balance-alb? As long as you have supported NICs it's meant to load balance streams across different NICs.
I have in the past and it didn't work. Only outgoing traffic was balanced. Limitation of crappy marvel yukon NICs perhaps. Tomorrow I'll go into the office, install some (personal) spare Intel NICs and try again.

I also looked into unbound and so far everything I put in the pfsense's advanced options area will causes unbound to crash. I was hoping it would be simple as
Code:
local-data: "server1 A 10.10.10.1"
local-data: "server1 A 10.10.10.2"
local-data: "server1 A 10.10.10.3"
I don't understand the config file well enough to make the proper entry.

If balance-alb doesn't pan out I'm left with getting LACP switch. Great having zero budget and only buying new equipment when something catastrophically breaks. The world of small businesses.
 

canta

Well-Known Member
Nov 26, 2014
1,029
198
63
39
Thanks for all the suggestions

I could add a delay. The backup script is ran at shutdown, and everyone leaves at 5pm.


I have in the past and it didn't work. Only outgoing traffic was balanced. Limitation of crappy marvel yukon NICs perhaps. Tomorrow I'll go into the office, install some (personal) spare Intel NICs and try again.

I also looked into unbound and so far everything I put in the pfsense's advanced options area will causes unbound to crash. I was hoping it would be simple as
Code:
local-data: "server1 A 10.10.10.1"
local-data: "server1 A 10.10.10.2"
local-data: "server1 A 10.10.10.3"
I don't understand the config file well enough to make the proper entry.

If balance-alb doesn't pan out I'm left with getting LACP switch. Great having zero budget and only buying new equipment when something catastrophically breaks. The world of small businesses.
newegg has running sale for 16-Port ZyXEL Fanless L2 Web Managed Rackmountable Switch (GS1900-16) for $129.99 - $60 with promo codeEMCAXKK59 = $69.99

I am resisting not to buy due already has two GS7224T with LACP, and one GS7216T without LACP.....
must resist....

want proven works? LACP is the answer :D

I tried bonding with Intel NIC, this worked

get Intel NIC!! and trash out marvel one .....
 

Danic

Member
Feb 6, 2015
73
27
18
jrdm.us
Thanks for the suggestions everyone. For now I'm running multiple NICs and multiple IPs and assigning specific workstation to specific IPs