performance is dreadful

[tessus]

I've setup a rsync backup job from my TS-419P II (RAID5 with 4x 3TB disks) to my TS-651 (RAID5 with 6x 6TB disks).
They are connected via a 1 GBit router and CAT6 cables (also tried 5E). There's no other traffic on this switch. Yet, the transfer speed is 21MB/s. This means I have to wait 3 days for transferring 5TB of data. This must be a joke.

I have also checked the network info on both QNAPs and both are connected with 1 Gbps.

Have you had any experience with those 2 models and can you elaborate on this dreadful performance? Copying files from my MacBook to a USB disk is 4 times faster. I'm sorry, but RAID5 should be reasonably fast for reading sequential data (most files are between 400MB and 6GB) and a transfer rate of 100MB/s on a 1GBit LAN is definitely not uncommon (I've seen it even up to almost the max theoretical bandwidth), so what's going on here?

I have quite some experience with storage subsystems and this is definitely odd. Yeah, ok, I'm not using fibre channel and SSDs, but come on... 21MB/s when it rather should be about 5 times that fast?

Any insights would be highly appreciated. Cheers.

[Moogle Stiltzkin]

performance on the ts-419p II is not that great to begin with (outdated nas)
https://www.techpowerup.com/review/qnap ... ii/12.html


maybe there are other things

- are cables ok?
- are you using latest qts?
- if both are qnap, why use rsync? use rtrr instead (lan right?)
- what is your HBS settings? i'm assuming you are using HBS? some settings in hbs can impact old models like SSL (which i disable for my ts-509 pro)
- did you run test on hdds? are they in good condition?

[tessus]

Thanks for the reply, but as you can see even from your link the read speed is still way over 21 MB/s for RAID5 with 4 disks.

> are cables ok?

Yes. As mentioned in my first post I tried different cables. And the cables are ok.

> are you using latest qts?

Yes.

> if both are qnap, why use rsync? use rtrr instead (lan right?)

Ha, that's the same speed or even a bit slower. Btw, rrtr is also using rsync in the background.

> what is your HBS settings? i'm assuming you are using HBS? some settings in hbs can impact old models like SSL (which i disable for my ts-509 pro)

I'm not using HBS. HBS makes no sense in my env. I'm using static volumes, and the files I'm currently rsyncing are unique. No dedup possible.

[Moogle Stiltzkin]

yeh static vols should be the best performance. hm....

[dolbyman]

the 419pII has probably 100% cpu load ... it's a 10 year old cellphone type processor ..it's just that slow

[spikemixture]

Thanks for the reply, but as you can see even from your link the read speed is still way over 21 MB/s for RAID5 with 4 disks.

> are cables ok?

Yes. As mentioned in my first post I tried different cables. And the cables are ok.

> are you using latest qts?

Yes.

> if both are qnap, why use rsync? use rtrr instead (lan right?)

Ha, that's the same speed or even a bit slower. Btw, rrtr is also using rsync in the background.

> what is your HBS settings? i'm assuming you are using HBS? some settings in hbs can impact old models like SSL (which i disable for my ts-509 pro)

I'm not using HBS. HBS makes no sense in my env. I'm using static volumes, and the files I'm currently rsyncing are unique. No dedup possible.

Hopefully someone here will fix your problem.
What software are you using to get from box to box?
I have basically given up trying to get faster than 40MB/s and I have 10GBE !
U might want to try HBSv3 as allegedly that is faster (not for me). I get upto 180MB/s copying the same data to a USB3 8bay Probox!

[tessus]

the 419pII has probably 100% cpu load ... it's a 10 year old cellphone type processor ..it's just that slow

Yep, it's close to a hundred percent. Which is really weird for a simple rsync process. But if the processor is really that slow, it might also impact IO. Unfortunately there's no `vmstat` on that system, but I believe the values in the r column would be rather high. But the fact that the rsync process uses a lot of system CPU pretty much confirms this.

[tessus]

What software are you using to get from box to box?

Backup station -> Remote Replication -> Rsync (File-Level Backup)

I have basically given up trying to get faster than 40MB/s and I have 10GBE !

If this is the case I won't buy a QNAP system anymore and build my own NAS system.
The whole reason for buying QNAP was that I didn't have to do it myself. How can a company sell a NAS that doesn't do its primary job right? I'm really puzzled by this.
All the virtualization and Media rubbish is nice to have, but what are they good for, if I can't access the data properly (in a timely manner)?

To the moderators of this forum: "c r a p" is not necessarily a swear word. The first definition in the dictionary is: something of extremely poor quality (• nonsense. • unwanted articles; rubbish.)

U might want to try HBSv3 as allegedly that is faster (not for me). I get upto 180MB/s copying the same data to a USB3 8bay Probox!

HBS is not an option for me and after that transfer I will never touch it again.

180MB/s sounds more like it. What the heck are these QNAP guys thinking?

[QuantumBiker]

If the systems are on different subnet and you are not just using router as a switch - could it be the router setup? Just because a router has a 1G interface does not mean it can route traffic at that rate. Several things can slow it down.

[QuantumBiker]

Oh - and do you have compression turned on for transfer? Sometimes that can really load down the box.

[tessus]

Thanks for the ideas.

could it be the router setup?
Oh - and do you have compression turned on for transfer? Sometimes that can really load down the box.

No, the router works just fine and I can transfer with 115MB/s on that switch. Just not with the 2 QNAPs. As mentioned before, the only traffic currently on that router is the file transfer.
They are on the same subnet, btw. They stand next to each other physically and even use ports next to each other on the router. hahaha.
Too bad quantum entanglement hasn't been leveraged for backups yet.

Compression is not turned on, neither is rsync via ssh. I have turned off everything that could degrade performance.

I even checked the queue depth of the disks. All seems fine. I begin to think that performance on QNAP systems is generally bad. In light of this debacle, I've done some research and QNAP NAS systems have the worst performance of all. As I said, no more QNAP for me.

[Trexx]

Speed of devices at both ends make a difference. Backup option settings make a difference.

Rsync where depending on setting may have to compare checksums running on a very old box 419 will not be fast. If you want fast use RTRR.

Don’t blame Qnap. Choose more optimal settings or get faster hw.


Sent from my iPhone using Tapatalk

[tessus]

Rsync where depending on setting may have to compare checksums running on a very old box 419 will not be fast. If you want fast use RTRR.

Hmm, I tried using RTRR and it wasn't faster at all. But I cancelled the job after 2 minutes. Maybe I have to wait.

I still have 2 more TB to copy after the current job. It should be done in another day or so. I will then use RTRR, but I doubt it will be any faster. Especially since rtrr also uses rsync in the background.

Also, I think I mentioned I did not use any rsync options (in the UI) that would slow down the process. Anyway, I'll let you know the result after the RTRR session. And it is QNAPs fault, when I get faster speed with any other device. So please stop your condescending tone and try to help with more constructive feedback. In previous posts I have mentioned that RTRR was not faster. Maybe reading before blurting out something helps. If you read the previous posts, you will see that I have already established that "Speed of devices at both ends make a difference. Backup option settings make a difference."

[P3R]

Yep, it's close to a hundred percent. Which is really weird for a simple rsync process.

The thread was really solved here.

With Rsync both systems need to compute checksums of all data. That have always been a brick wall with slow CPUs and always will.

Hmm, I tried using RTRR and it wasn't faster at all. But I cancelled the job after 2 minutes. Maybe I have to wait.

Ehhh, yes.

Especially since rtrr also uses rsync in the background.

I know that several people claim that but I have never seen it stated by Qnap or any other reliable source. So what source(s) do you have for that claim?

In my experience RTRR have the best performance on LAN connections, especially for large bulk transfers. One need to be careful with the options though.

[tessus]

Especially since rtrr also uses rsync in the background.

I know that several people claim that but I have never seen it stated by Qnap or any other reliable source. So what source(s) do you have for that claim?

In my experience RTRR have the best performance on LAN connections, especially for large bulk transfers. One need to be careful with the options though.

Hmm, I think I've read it somewhere, but upon testing RRTR my QNAPs no rsync processes were spawned. And yes, the performance was better. It was still slow (around 39MB/s) but it would have reduced my 3 days copy process by almost 50%. The only problem is that it uses its own directory structure on the target and thus it's only usable with manual intervention.
I am still disappointed over the performance. IMO, it should be at least twice as fast. But I will remenber this for the future that for a one-go data transfer, RRTR is the way to go.