TL-SG2008P LACP issue with Synology DS220+
This thread has been locked for further replies. You can start a new thread to share your ideas or ask questions.TL-SG2008P LACP issue with Synology DS220+
This thread has been locked for further replies. You can start a new thread to share your ideas or ask questions.
Hello,
New Omada user here, having an issue with TL-SG2008P and my Synology NAS with LACP.
My switch is managed by an OC200, LAG1 LACP enabled:
From Synology's end, LACP is enabled and both ports connected, link speed appears as 2000mbit (same results with static lag and balance xor)
however when doing file transfer to/from two other pcs which are connected directly to the switch, the total bandwidth appears to be 1gbit.
if i remove the lag/lacp configuration and use "adaptive load balancing" on the synology alone, the correct speed is achieved, so it looks like something on the switch is not being applied correctly with regard to LACP.
would appreciate any hints on this, seems like i'm missing something here
thanks!
- Copy Link
- Subscribe
- Bookmark
- Report Inappropriate Content
Hi @NinjaMonkey
Thanks for posting in our business forum.
This is normal and not a bug and @KJK has well-explained everything to you.
Note that we do not have layer 3+4 LAG for you. We only have layer 2 or 3 LACP. Not 4. There are switches on the market support layer 4 LACP which can fix the problem you experience. Quote below:
NAS to one wired and two wireless clients achieves 2gbit,
NAS to two different wired clients tops out at 1gbit
What you have done is verifying the LACP is effective. 2Gbps speed has been reached by multi-devices.
For more knowledge about layer 4 LACP, you can search it online. We don't have related materials as we do not have such a product to support it.
- Copy Link
- Report Inappropriate Content
Thanks @Clive_A ,
indeed @KJK 's replies were very informative but it only makes me think that there is some bug in the LACP layer 3 algorithm, (SRC IP + DST IP/ DST IP / SRC IP)
in both outlined scenarios it should be able to achieve maximum speed, since all clients are using different IPs,
even by layer 2 lacp i should be able to max out, since the wired clients are directly connected to the switch and have different mac addresses.
or is my understanding of this completely wrong?
- Copy Link
- Report Inappropriate Content
Hi @NinjaMonkey
Each computer will get a 1Gbps maximum regardless of your NIC link speed. That'll be the picture on the right which is correct. The entire bandwidth is gonna be 2Gbps.
Layer 3 LAG is not gonna break the limit of your NIC even if your NIC is a 10Gbps one. No.
If you wanna have 5Gbps from a group of 5-port LAG, then you need layer 4 LAG which can help you aggregate the speed to 5Gbps on your 10Gb NIC.
If you mean the picture on the left is on the hash algorithm MAC address(layer 2), that is possible. When calculating the hash value, sometimes some devices can be assigned to the same link. Which means they share one port bandwidth. Only when the hash values are different, do you get load balanced between the links.
- Copy Link
- Report Inappropriate Content
Thank you @Clive_A
I think we are still having a bit of trouble understanding eachother,
NAS is uplinked to SG2008P via 2x1gbit in L2/L3 LAG,
all 5 clients are connected to the same SG2008P,
3 wired clients each with its own 1gbit link, and 2 wifi clients via 2 different EAP225.
when transferring data from the NAS and to the 3 wired clients simultaneously, the total bandwidth coming out of the NAS is 1gbit. so about 333mbit/s per each client, so LAG is only half utilized.
when transferring data from the NAS and to 1x wired clients + 2 wi-fi clients simultaneously, i'm getting total bandwidth of 2gbit from the NAS, or 667mbit/s per each client, so full utilization of the LAG.
in both scenarios, Layer2/3 LACP should be able to split the load equally and allow total bandwidth of 2Gbit from the NAS, is this not correct?
- Copy Link
- Report Inappropriate Content
“in both scenarios, Layer2/3 LACP should be able to split the load equally and allow total bandwidth of 2Gbit from the NAS, is this not correct?”
Well, it is not correct. The process of selecting a port for any data flow uses a pure mathematical algorithm. The volume of the traffic does not play any role there. The key part of that algorithm is the calculation of those hash values. You have a choice of what that calculation is based on, usually MAC and/or IP addresses, but you will not know how exactly that calculation is done. To my knowledge, each company has its own proprietary hash calculation formula and keeps it secret. If there are just a couple traffic sources, there are just a couple of hash values and the chances that they match one and the some port are very high. If there are plenty of traffic sources, there are many hash values so the chances that they match different ports increase substantially. That usually results in a better load distribution, but not always. For example, if a certain source produces the majority of traffic, most of the traffic, of course, will be flowing by just one port. Only the other hand, even if there are only two traffic sources, you may get “lucky” and two different ports get matched. Like it or not, that’s how it works in those kind of LAGs and it is not just TP-Link specific.
- Copy Link
- Report Inappropriate Content
- Copy Link
- Report Inappropriate Content
Information
Helpful: 0
Views: 2577
Replies: 16
Voters 0
No one has voted for it yet.