Header Only - DO NOT REMOVE - Extreme Networks
Question

EAPS and Load Sharing

  • 29 November 2014
  • 26 replies
  • 889 views

Userlevel 2
Hi,

I want to double my ring bandwidth between one BD8800 and two X460. Which one is the best, using EAPS over load sharing ports (ex. LACP) or use multiple domain rings with different ways (clockwise and conter clockwise)?

Best regards,

26 replies

Userlevel 6
Hello Lazuardi I would create a LAG connection between the two switches over that link. Thanks P
Userlevel 6
Hello Lazuardi I would create a LAG connection between the two switches over that link. Thanks PHello Linn the LAG is what we call sharing it is created using the command enable sharing Grouping algorithm address LACP You can either set the group as static which means it will not use LACP or add LACP where the two sides communicate to establish the LAG group. The effect is that you share the load over all of the links in the LAG. It will not be balanced because the traffic is sent over a particular port based on the hash and is flow based. So for example if I have two flows one can be a ping and the other can be a TFTP connection. They can be assigned to two different ports but the data I'll not be the same. Hope that helps
Userlevel 6
Hello Lazuardi I would create a LAG connection between the two switches over that link. Thanks PHey Lazuardi

You can change the CRC to customize the hash. The command is

"configure sharing address-based custom"
"configure sharing address-based custom [ipv4 [L3-and-L4 | source-only |"
"destination-only | source-and-destination] | hash-algorithm [xor | crc-16 |"
"crc-32 [lower | upper]]]"

I would recommend calling GTAC to help set this up to make sure that everything is set up properly.

P
Userlevel 4
IF there are more number of switches then can go for EAPS and other solutions
Userlevel 3
Hello Lazuardi I would create a LAG connection between the two switches over that link. Thanks PHi Paul ,

Could you explain more about LAG connection please? basic setup, configuration and possible outcome? Thanks.
Userlevel 2
Now i'm trying EAPS over LACP. Any drawbacks I should know of this method? Why this method is better than using multiple domain rings with different ways?
Userlevel 6
Hello Lazuardi. You can do Another EAPS ring but that won't add bandwidth per SE because you are still using the same ports tx/rx just sending the traffic in a different direction. It is a solution if you want to use all of the links of the ring since you can change the blocked port. Using a LAG actually adds more bandwidth to the existing ring allowing you to keep the VLANs in tact and you can add them with usually no impact to the traffic, however, I do recommend adding them in off hours. If you do two separate rings with two different domains you will have to make sure they are separate and connect them with just one link or route between them on one switch so you don't have a loop. I hope that helps. P
Userlevel 2
Hi Paul,

Do you mean that I cannot have the same protected VLANs on all ring domains of the same devices?

Best regards,
Userlevel 6
Hey Lazuadri Yes a VLAN can only be one one domain if they share the same physical ring. if you are using spatial-reuse on EAPS, where one domain goes one direction and the other domain goes in the other direction you have to divide the VLANs in half where half are on one domain and the other half on the other domain. P
Userlevel 3
Hello Lazuardi I would create a LAG connection between the two switches over that link. Thanks PUtilization is high on Link 1 only. I want to use load balance on both links.
Please kindly see the below output and advice thanks.

* I try L2 static, L3_L4 and LACP. all the same. Ports are not load balance. *

enable sharing 1 grouping 1,4

enable sharing 1 grouping 1,4 L3_L4

enable sharing 1 grouping 1,4 algorithm address-based L3_L4 lacp

CoreSW # sh port 1,4 utilization bandwithPort Link Link Rx Peak Rx Tx Peak Tx
State Speed % bandwidth % bandwidth % bandwidth % bandwidth
================================================================================
Link_1> A 1000 82.06 94.94 48.23 55.72
Link_2> A 1000 0.25 0.30 1.34 1.57
================================================================================
> indicates Port Display Name truncated past 8 characters
Link State: A-Active, R-Ready, NP-Port Not Present, L-Loopback

Please help me how to load balance these two links together. Thanks sir.
Userlevel 6
Hello Lazuardi I would create a LAG connection between the two switches over that link. Thanks PHey Lazuardi

you will not get a perfect load balance. that is why it is called sharing or link aggregation versus load balanced. It all depends on how the addresses are hashed. I see you are using L3 and L4 which should hash based on src and dst IP and IP port number. How much traffic are you providing across the link? is this production or in a lab?

Thanks
P
Userlevel 3
Hello Lazuardi I would create a LAG connection between the two switches over that link. Thanks PHi Paul Russo,

It is in Production live network. Please advice how to load share on two links.
Really appreciate your help and thanks for your time.

We do need your advice.

Thanks sir.
Userlevel 2
Hi Paul,

Any specific reason of selecting L3_L4 or CRC algorithms? What is default default hashing algorithm for LACP?

Best regards,
Userlevel 6
Hey Lazar do the default is crc16 I believe. Changing it to crc32 may help but I would recommend testing it if you can. P
Userlevel 6
Hey Lazuardi

I if you just use LACP the algorithm is L2. The reason to use L3-L4 is to get more info to hash off of which should get a better spread of traffic. If you use CRC32 verses CRC16 the that should help as well.

Thanks
P
Userlevel 2
Hi Paul,

In your opinion, which one is better for spreading, L3_L4, CRC16 or CRC32? Any license required for custom CRC16 and CRC32 sharing hash, especially for X460?

Best regards,
Userlevel 4
License is not required Try both and save the best one
Userlevel 2
Hi,

Will CRC16 and CRC32 give more ethernet farmes reordering overhead than L3_L4?

http://en.wikipedia.org/wiki/Link_aggregation

Best regards,
Userlevel 3
Hi,

I try this command and not supported in XOS 15.3.3.5 model X450a-24x.

"configure sharing address-based custom"

Is there any way to change the CRC to customize the hash for X450a-24x? Thanks.

BRgds,
Userlevel 6
Hey Paul

The command configure sharing address-based custom should be in 15.3 as it has been in XOS since 12.3. The crc-32 however is not there we only supported crc-16

"This command specifies the part of the packet header that the switch examines to select the egress"
"port for address-based load-sharing trunks. The address-based load-sharing setting is global and applies to all load-sharing trunks, or LAGs, that are address-based and configured with a custom algorithm. You change this setting by issuing the command again with a different option."
"The addressing information examined is based on the packet protocol as follows:"
"• IPv4 packets—Uses the source and destination IPv4 addresses and Layer4 port numbers as specified with this command."
"• IPv6 packets—Uses the source and destination IPv6 addresses and Layer4 port numbers."
"• MPLS packets—Uses the top, second, and reserved labels and the source and destination IP"
"addresses."
"• Non-IP Layer 2—Uses the VLAN ID, the source and destination MAC addresses, and the ethertype."
"The xor hash algorithm guarantees that the same egress port is selected for traffic distribution based"
"on a pair of IP addresses, Layer4 ports, or both, regardless of which is the source and which is the"
"destination."

Hope this helps

P
Userlevel 3
Hey Paul

The command configure sharing address-based custom should be in 15.3 as it has been in XOS since 12.3. The crc-32 however is not there we only supported crc-16

"This command specifies the part of the packet header that the switch examines to select the egress"
"port for address-based load-sharing trunks. The address-based load-sharing setting is global and applies to all load-sharing trunks, or LAGs, that are address-based and configured with a custom algorithm. You change this setting by issuing the command again with a different option."
"The addressing information examined is based on the packet protocol as follows:"
"• IPv4 packets—Uses the source and destination IPv4 addresses and Layer4 port numbers as specified with this command."
"• IPv6 packets—Uses the source and destination IPv6 addresses and Layer4 port numbers."
"• MPLS packets—Uses the top, second, and reserved labels and the source and destination IP"
"addresses."
"• Non-IP Layer 2—Uses the VLAN ID, the source and destination MAC addresses, and the ethertype."
"The xor hash algorithm guarantees that the same egress port is selected for traffic distribution based"
"on a pair of IP addresses, Layer4 ports, or both, regardless of which is the source and which is the"
"destination."

Hope this helps

P

xos 15.3 model X450a. Cannot use custom

"enable sharing 10 grouping 10-11 algorithm address-based custom lacp"




X450a-24t.2 # enable sharing 10 grouping 10-11 algorithm address-based ?
Execute the command
health-check link aggregation group handled by health-check application
L2 Based on layer 2 MAC (default)
L3_L4 Based on layer 3 IP and layer 4 port
lacp link aggregation group handled by LACP
X450a-24t.2 # enable sharing 10 grouping 10-11 algorithm address-based
Userlevel 6
Hey Paul

Thanks apologize the command reference says that it is in there as of 12.3 code.

I do not believe there is anyway to change the crc in the 450a

P
Userlevel 2
Hi Paul,

There is no CRC32 custom algorithm options on our BD8800 and X460. Our BD8800 cannot use custom algorithm with error "No slots currently in the system support the custom algorithm". When I try to use L3-L4, our BD8800 cannot connect to all devices connected with load sharing. It seem no problem with CRC16 and L3_L4 on our X460. Is it normal?

Best regards,
Userlevel 6
Hey Lazuardi that is possible if your 8800 modules are older. What do you mean "BD8800 cannot connect to all devices connected with load sharing"

Is the traffic being more distributed on the links?

P
Userlevel 2
Hi Paul,

How could new buying BD8800 modules become older? I'm using 15.6.1.4 EXOS.

I mean of "cannot connect" when using L3_L4 LACP algorithm is cannot ping the load sharing connected device (X460) IP which previously could if I using L2 LACP algorithm on BD8800.

Best regards,

Reply