Router Features with Synology NAS

[sfx2000]

That connection buys redundancy and failover in case one link fails for whatever reason - and a LAG'ed connection does offer more capacity in a heavily loaded situation, which in most home networks is next to impossible actually...

Take it from me - been there - building out networks to support 20M users...

 

[sfx2000]

I think you're wrong?

It's a check-box feature, similar to tri-radio/dual-band AC5300 class - and most folks misunderstand what that means - and the marketeer's leverage that to sell more boxes...

 

[L&LD]

It's a check-box feature, similar to tri-radio/dual-band AC5300 class - and most folks misunderstand what that means - and the marketeer's leverage that to sell more boxes...

I still think you're wrong, in this specific case.

 

[sfx2000]

I still think you're wrong, in this specific case.

Do the testing, show the numbers...

 

[L&LD]

Do the testing, show the numbers...

I'm not the one with the hardware.

 

[sfx2000]

Bandwidth is a funny thing with how it's measured - in some cases it's speed for a single client, and for servers, it's more about the number of concurrent connections.

LAG does improve bandwidth - which goes to my statement about capacity - which is bandwidth in my book - as I qualify "speed" as another parameter... and like I mentioned above, LAG does offer benefits like redundant connections and fail-over - but at the same time, a bit cannot exceed the limits of the PHY - anybody familiar with networking understands that...

Something to read... https://docs.oracle.com/cd/E36784_01/html/E37516/gmsab.html

Might put things into context..

 

[L&LD]

Bandwidth is a funny thing with how it's measured - in some cases it's speed for a single client, and for servers, it's more about the number of concurrent connections.

LAG does improve bandwidth - which goes to my statement about capacity - which is bandwidth in my book - as I qualify "speed" as another parameter... and like I mentioned above, LAG does offer benefits like redundant connections and fail-over - but at the same time, a bit cannot exceed the limits of the PHY - anybody familiar with networking understands that...

Something to read... https://docs.oracle.com/cd/E36784_01/html/E37516/gmsab.html

Might put things into context..

I understand the difference between bandwidth and speed as you're describing it.

I don't understand why it won't improve this situation? (I feel that you may have skimmed this thread too quickly).

 

[RMerlin]

From what I know, LAG only helps if you have multiple clients simultaneously accessing the device. So if you have a NAS that uses LAG, it means that two desktops would be able to push the NAS's throughput up to a theoretical 2 Gbps.

If you have one PC that uses LAG and one NAS that uses LAG, I would assume this is once again a single client talking to a single device, therefore only 1 Gbps would be used.

I have never tested the theory however - I see no need here to LAG my NAS.

 

[sfx2000]

So if I had to use a metaphor - sometimes this helps...

Imaging yourself at Costco, Walmat, the local Grocery Store - the lines are long... the time taken to complete the checking/transaction - this is speed... and that's pretty much a fixed thing.

The number of aisle/tills open/checkers available - this is bandwidth - and this is what LAG does - each transaction still happens at 1Gb, but there's more aisles open as a function of how many links are aggregated...

(yes, I had do this once for a business sponsor of a project, explaining things like he was 5 years old - smart guy, but not a networking/technical guy)

 

[sfx2000]

In telecom, we call this erlangs - it's a comms theory item - but basically it's the same concept as what I discussed above - or the number of elevators in a building vs. the number of people that need to access a given floor...

One of my favorite interview questions actually - outside of "how many barbers live in China", which is basically the same issue rephrased as this also takes into the issue of demand vs. capacity...

(and this goes back to a statement I made some time back - latency matters - latency in the "tell it to me like I'm 5" is how much time it takes for that transaction - which in a normal one, takes a fixed amount of time - but if the customer has to fiddle with the checkbook/credit card, or asks questions, this extends that - and not all the customers are the same - which leads to jitter on the link).

I've got a couple of smart guys now on this thread - and I'm a bit surprised that I have to explain it like this...

 

[L&LD]

I understood what you mean (as I said so too) before your analogy (which I use too, btw).

What I don't understand is why two devices connected to each other in LAG cannot transfer at almost 2x the speed? Both 'know' the connections are there, both have the 'connection'. What is preventing them from using it fully then (unless it is an O/S limitation)?

 

[sfx2000]

What is preventing them from using it fully then (unless it is an O/S limitation)?

Because they can and they can't...

LAG will improve availability, but it cannot increase the bit rate...

 

[kirb112]

I think you're wrong?

Thank you for the input guys. Is someone else able to confirm one of these positions?

 

[L&LD]

Thank you for the input guys. Is someone else able to confirm one of these positions?

Unless someone else can confirm or deny one side or the other, I will concede to sfx2000 here.

(Hopefully, I learned 'right').

 

[kirb112]

So if I had to use a metaphor - sometimes this helps...

Imaging yourself at Costco, Walmat, the local Grocery Store - the lines are long... the time taken to complete the checking/transaction - this is speed... and that's pretty much a fixed thing.

The number of aisle/tills open/checkers available - this is bandwidth - and this is what LAG does - each transaction still happens at 1Gb, but there's more aisles open as a function of how many links are aggregated...

(yes, I had do this once for a business sponsor of a project, explaining things like he was 5 years old - smart guy, but not a networking/technical guy)

Does this mean that multi-file transfers will be faster? Let's say I am transferring a folder of 1000 songs or 1000 pictures. Does LAG enable these to be spread out over the interfaces? Say, 500 per cable in a 2x LAG setup? Ultimately resulting in the data being transferred more quickly?

 

[L&LD]

Does this mean that multi-file transfers will be faster? Let's say I am transferring a folder of 1000 songs or 1000 pictures. Does LAG enable these to be spread out over the interfaces? Say, 500 per cable in a 2x LAG setup? Ultimately resulting in the data being transferred more quickly?

That is a great example. (Hope the answer is 'yes', but it still may need to be two different clients needing files from the NAS).

@sfx2000 what do you say?

 

[sfx2000]

That is a great example. (Hope the answer is 'yes', but it still may need to be two different clients needing files from the NAS).

@sfx2000 what do you say?

In the typical home network - probably not - depends on the capabilities of the NAS, and the clients - even my Intel Based QNAP (which supports SMB3) - two clients cannot saturate it and the link... and that's two Win10 Core i7 machines over gigabit ethernet... there's certain filesystem efficiencies and tuning on the NAS side, but that typically would be primary concern, not the links from the NAS to the Switch, and from the switch to the clients.

If there were 100 clients, then definitely LAG would help...but then one would be probably running with a managed switch where one has a bit more control, and even an 8 port managed switch (like Netgear GS108T or similar) usually will have higher aggregate bandwidth across the switch than what one would find with a consumer Router/AP...

Can always try it - but in my honest opinion, it's introducing a level of complexity that isn't really needed, and perhaps not much benefit in a small network.

 

[sfx2000]

That is a great example. (Hope the answer is 'yes', but it still may need to be two different clients needing files from the NAS).

@sfx2000 what do you say?

Want to add one more comment here about Link Aggregation - again, a lot of folks think that using LAG will double speed - and again, I assert it doesn't - it will increate capacity, but a single client, on a single session, even if one have four links in the LAG group, all IP packets have to travel over the same wire - and this goes into how IP is encapsulated by ethernet - and that's how it is... This is how LAG works, and yes, it's counter intuitive to folks that look at RAID or MIMO (in the case of Wireless).

Looking at the diagram above, switch is on the left, and the NAS is on the right - you see two links - but again, these are datalink/phy later links at the MAC level - the IP session sees both as one link - as per LAG - which is what this discussion is about.

At the session level however, it's a single connection from client to server, as such, it's carried thru the transport as a single session, driving down into the stack onto on MAC layer session, which lives on a single wire - the other wire is redundant in a single user/session connection.

So - in usage - Alice's client can have a session, and Bob's client can as well - the nice thing with LAG, is Alice gets her own, while Bob has his own...

So yes, overall thruput (Alice's 1GB session, and Bob's 1GB session) is doubled, Alice will only get 1GB whether Bob is on the NAS or not..

This make things any clearer?

 

[baddog]

Yes, thank you!

 

[Dr Tone]

LAG on a desktop is handy only for redundancy or if you are doing multiple things to another device on your switch that also has LAG. Example might be VMs on the desktop or multiple file copies. LAG is also nice if saturating 1Gb for a file copy but still have snappy network/internet because of the the other half of the LAG.

I find it most useful on my server that has iSCSI luns mounted on the NAS. The server has LAG so I copy at a full 1Gb to server which then uses the other half of the LAG to get a full 1Gb to the Lun on the NAS. If I didn't have LAG on that server I'd only be able to get 50% of 1Gb when copying files to and from the server because half would have to be allocated to iSCSI.