HP 1920-48G : LAG hash algorithm/policy

[JO4]

Hi,

The network of my (small) office is based on several HPE 1920-48G (JG927A). The network is currently organized like this :

Computers ---RJ45--- Switch #2 ---2x fiber--- Switch #1 ---RJ45--- Servers & firewall

The 2 switchs are separate of 3 floors.

Four VLANs exist :
- VLAN 10 (10.25.10.0/24) : workstations (company computers which can access the servers)
- VLAN 20 (10.25.20.0/24) : guest (guest computers which can only go to Internet, no servers)
- VLAN 30 (10.25.30.0/24) : servers (webservers and NAS)
- VLAN 40 (10.25.40.0/24) : admin (management network)

The firewall (a pfSense box) is connected to each VLAN (the host .1 on every subnet). The 2 switchs are connected through a LAG with LACP enabled.

Employees regularly transfer large files (> 100 GB) to one of the two NAS. The problem is when multiple employees transfer simultaneously : the speed is divided.

In this case, the file must go through the firewall so I have this :

* First employee :

IP src : 10.25.10.6 (a workstation)
MAC src : 11:11:11:11:11:11 (the workstation MAC)
IP dst : 10.25.30.5 (NAS #1)
DST mac : 99:99:99:99:99:99 (the pfSense MAC, the next hop to go the 10.25.30.0/24)

* Second employee :

IP src : 10.25.10.7 (a workstation)
MAC src : 22:22:22:22:22:22 (the workstation MAC)
IP dst : 10.25.30.10 (NAS #2)
DST mac : 99:99:99:99:99:99 (the pfSense MAC, the next hop to go the 10.25.30.0/24)

I have no option to change the LAG hash algorithm / policy used for the LAG.

First question : which LAG hash algorithm / policy is used for the HPE 1920-48G ? (I found nothing in the documentation or on Internet).

I suspect that the "VLAN and MAC dst" are used. If this is the case, my two transfers will go through the same fiber and it will not go faster. Is that correct ?

Second question : is it possible to change the mode through the CLI ? (see. https://community.hpe.com/t5/Web-and-Unmanaged/HP-1920-Link-Aggregation-trunk-hash-policy/m-p/7016440)

---

I borrowed from a friend 2 switches from a other constructor. The network diagram will look like this :

Computers ---RJ45--- Switch #2 ---4xRJ45--- NEW switch #3 ---fiber @ 10Gbps--- NEW switch ---RJ45--- Servers & firewall (connected with a 10Gpbs interface)

The goal is to interconnect the office and the "datacenter" with a 10Gbps link (the firewall will also be connected with a 10Gbps interface) and to have 4 employees transferring at the same time to 1 Gbps.

On the new switch #3, I can create a LAG with LACP and choose the hash algorithm "IP src/IP dst".

In this configuration, the LAG LACP enabled between the HPE 1920-48G and the new switch are not going to work in the "same mode".

Second question : how is it going to happen?

Thank you very much for you help.

Best regards,

[JO4 · Accepted solution]

I found the solution.

1. The default mode of LAG hashing policy is "destination-mac, source-map".

2. You can change the mode to "destination-ip" and/or "source-ip". You must connect to the hidden CLI and use the command "link-aggregation load-sharing mode destination-ip source-ip" (to show the current configuration value : "display link-aggregation load-sharing mode ").

Thread closed.