SYSTEM WARNING: 'date_default_timezone_get(): It is not safe to rely on the system's timezone settings. You are *required* to use the date.timezone setting or the date_default_timezone_set() function. In case you used any of those methods and you are still getting this warning, you most likely misspelled the timezone identifier. We selected the timezone 'UTC' for now, but please set date.timezone to select your timezone.' in '/usr/share/mantis/www/core.php' line 264

0002042: GUI for interface bonding - MantisBT
MantisBT - Endian Firewall
View Issue Details
0002042Endian FirewallNetwork related (VPN, uplinks)public2009-07-28 01:412010-09-24 09:20
tomakos 
peter-endian 
normalfeatureN/A
acknowledgedopen 
 
 
0002042: GUI for interface bonding
I have just set up a linux server with 4 Gigabit-NICs which are "bonded".
This means, that the 4 NICs are being joined together and have only one IP. When clients talk with this server, the load is balanced on the 4 NICs.

It would be great, if endian would support bonding, too. In this case, I could have e.g. 2 green and 2 orange interfaces and the traffic green<>orange would be with 2 Gigabit, instead of 1.

Bonding is supported by modern linux kernels by default. It is just a setting that someone has to do!

There are 7 different modes of bonding, differeng in the way and level of OSI-layer that the traffic is balanced on the bonded NICs..
These modes are explained here: http://www.mjmwired.net/kernel/Documentation/networking/bonding.txt#434 [^]

No tags attached.
Issue History
2009-07-28 01:41tomakosNew Issue
2009-07-28 01:41tomakosAssigned To => peter-endian
2009-07-28 01:48tomakosNote Added: 0002794
2009-07-28 01:53tomakosNote Added: 0002797
2009-07-28 01:56tomakosNote Edited: 0002794
2009-07-28 02:00tomakosNote Edited: 0002794
2009-07-28 02:23tomakosNote Edited: 0002794
2009-07-28 07:37luca-endianNote Added: 0002801
2009-07-28 09:56tomakosNote Added: 0002802
2009-07-28 11:02tomakosNote Edited: 0002802
2009-07-28 17:14peter-endianNote Added: 0002810
2009-07-28 21:21tomakosNote Added: 0002812
2010-09-24 09:20peter-endianStatusnew => acknowledged
2010-09-24 09:20peter-endianSummaryLink aggregation AKA "bonding" (=combining multiple NICs to the switch) => GUI for interface bonding

Notes
(0002794)
tomakos   
2009-07-28 01:48   
(edited on: 2009-07-28 02:23)
Keywords:
link balancing, load balancing, multiple uplinks, link aggregation, transparent, combine DSL, balance-rr, round-robin, round robin, active-backup, balance-xor, 802.3ad, balance-tlb, balance-alb, layer2, layer2+3, layer3+4, Ethernet trunk, NIC teaming, port channel, port teaming, port trunking, link bundling, EtherChannel, Multi-Link Trunking (MLT), NIC bonding, Network Fault Tolerance NFT, WAN, WAN uplink

(0002797)
tomakos   
2009-07-28 01:53   
Please see also the closely related:
http://bugs.endian.com/view.php?id=802 [^]
http://bugs.endian.com/view.php?id=2041 [^]
(0002801)
luca-endian   
2009-07-28 07:37   
you are talking about the Peter's super-secret-feature.. shhhh ;)
http://blog.endian.com [^]
(0002802)
tomakos   
2009-07-28 09:56   
(edited on: 2009-07-28 11:02)
I am not a master of this subject, but I have set up a server who's NICs are bonded and have read a little about bonding of internet links. So let me explain with the little of what I know about the subject:

There is a difference when speaking about combining some LAN-NICs and WAN-NICs (=uplinks to the internet).

When bonding LAN-NICs,
**********************
e.g. with one of the existing modes called 802.3ad (AKA "Link Aggregation"), those bonded NICs recieve 1 IP. The switch to which those NICs are connected needs be a manageable switch and to support this protocol. You go into the settings of the switch and tell him: "Those NICs are bonded, so please treat them as one and don't be confused about that 1 IP and scrambled traffic." The switch then handles the traffic appropriately.

There are 7 different modes and 3 different hash-policies.. The ultimate mode is balanced-rr, which truely spreads all traffic on the NICs, down to the packet level. It's comparable with a RAID0 on those NICs.. Even a single file that you send will be spread over the multiple NICs.. I have not managed this, since my switch doesn't seem to support it. And secondly there is a lot of loss, since the packets are scrambled so much, that a lot of packages get dropped because of timeouts.. The next best choice is 802.3ad with hash-policy "layer3+4". This means, that the kernel will analyse traffic based on the IP and the protocol used. With this way at least, the kernel will try to span traffic to the same IP over different NICs according to the protocol used. E.g. FTP traffic towards IP1 via NIC1 and SIP-Telephony-Traffic towards IP1 via NIC2.


When bonding Internet Uplinks,
******************************
AKA "Link balancing", there is one problem that makes it impossible to use the same bonding techniques as with the LAN NICs: You can't influence the other side, that what would be the switch in you LAN: The internet providers!

So the balance-rr and 802.3ad modes are out of the game for sure.
But there are other modes of bonding, that may be of interest for further researching:

511 balance-tlb or 5
512
513 Adaptive transmit load balancing: channel bonding that
514 does not require any special switch support. The
515 outgoing traffic is distributed according to the
516 current load (computed relative to the speed) on each
517 slave. Incoming traffic is received by the current
518 slave. If the receiving slave fails, another slave
519 takes over the MAC address of the failed receiving
520 slave.

and:

527 balance-alb or 6
528
529 Adaptive load balancing: includes balance-tlb plus
530 receive load balancing (rlb) for IPV4 traffic, and
531 does not require any special switch support. The
532 receive load balancing is achieved by ARP negotiation.
533 The bonding driver intercepts the ARP Replies sent by
534 the local system on their way out and overwrites the
535 source hardware address with the unique hardware
536 address of one of the slaves in the bond such that
537 different peers use different hardware addresses for
538 the server.
539
540 Receive traffic from connections created by the server
541 is also balanced.


I don't know if those modes of bonding are really usable for internet-uplink-bonding. I have found some other approaches in the internet:
- Changing the routing tables
- Making DNS-Round-Robins

I hope those information is somewhat helpful!

Best regards,
Tom

P.S. Those lines are taken of the "bonding.txt" THE source of information for setting up a NIC bond: http://www.mjmwired.net/kernel/Documentation/networking/bonding.txt [^] [^]
You will find all needed information about the modes and hash policies there!

(0002810)
peter-endian   
2009-07-28 17:14   
actually our bonding is implemented using XOR, which is only for high availability and not for link aggregation.
Could be changed manually easily however by editing /etc/rc.d/rc.bondconfig and using the respective options to the modprobe bonding line
(0002812)
tomakos   
2009-07-28 21:21   
Bonding is already integrated?! Or do you mean for the upcoming 2.3?

Well in any case: That is nice! Is there a way how I can help making the different modes and hash-modes availiable via the frontend of Endian?!