Jump to content
XPEnology Community
  • 0

Anyone know if It is possible to build a fast network between a Mac and Syno-xpenology ?


Mary Andrea

Question

Anyone know if It is possible to build a fast network between a Mac and Syno-xpenology with a direct ethernet connection?

Because I’ve a Mac from a 10g thunderbolt OWC ethernet adapter to a USB 2.5 ethernet adapter into Xpeno, but I never reach that two devices connect together (without passthrough Modem) . Of course I won’t passthrough Modem because it’s a No 10g port modem, and I have not a Switch) , I am asking if its possible to mount a fast Net with direct ethernet connection and how have to setup Syno and Mac. thk for suggestion.   

Link to comment
Share on other sites

24 answers to this question

Recommended Posts

  • 0
On 11/21/2020 at 12:37 PM, Mary Andrea said:

Anyone know if It is possible to build a fast network between a Mac and Syno-xpenology with a direct ethernet connection?

So, as anyone did it , I answer myself. Maybe someone is trying to make a "direct connection" between your computer and xpenology. After two days testing, errors and trying ...  I can affirm that a direct connection can be made, and I got triple transfer speed. The most important thing is: 1) hardware speed ethernet (NIC 10G, 5G, 2,5G USB, Thunderbolt adapters or Cards) and 2) SET UP DHCP SERVER in DSM. DSM will do a new IP for your ethernet device and you can connect to Server by Samba... .I increase transferences from 108mb/s to w235-r275 mb/s. (without modem through)

Edited by Mary Andrea
  • Like 1
Link to comment
Share on other sites

  • 0

I'm setting up a connection between a Mac (Hackintosh) with 10GBe ethernet card, a Xpenology setup with Sonnet 10Gbe card and I'm waiting for a Netgear switch 10Gbe (2 ports 10Gbe), and then I will post the result.

 

In the meanwhile I was noting that Synology DSM 6.2.3u2 is based on a very old Linux kernel:

 

unmane -r
4.4.59+

 

and from version 4.15 the Linux kernel is supporting Thunderbolt devices, and networking over Thunderbolt too. I'm wondering if/whether Synology DSM 7.0 will have the right version of Linux kernel to add Thunderbolt support; if it will be the case we can connect Macs over Thunderbolt and have a super fast access! What do you think @IG-88 ?

Link to comment
Share on other sites

  • 0
8 hours ago, Mary Andrea said:

-Could you be more explicit about "where" . What do you "where" suggest?

the forums most upper section " Information"  /  "Readers News & Rumours" its not the place for a request of information, the area has a good enough names to not mistake it for a request area

Link to comment
Share on other sites

  • 0

 

On 11/21/2020 at 4:37 PM, Mary Andrea said:

if It is possible to build a fast network between a Mac and Syno-xpenology with a direct ethernet connection?

6 hours ago, Hackaro said:

I'm setting up a connection between a Mac (Hackintosh) with 10GBe ethernet card, a Xpenology setup with Sonnet 10Gbe card and I'm waiting for a Netgear switch 10Gbe (2 ports 10Gbe), and then I will post the result.

 

you dont  need a switch, you can connect both nic's direct

i have a 1G and 10G in both nas and computer, the 1G in connected to a switch and the 10G connects directly between nas and pc and have static ip addresses of a different range, so when using the 10G ip address of the nas from pc the 10G connection will be used

10G switches are still expansive and often not passive cooled (beside noise level its also a sign for a heavier power consumption)

 

6 hours ago, Hackaro said:

In the meanwhile I was noting that Synology DSM 6.2.3u2 is based on a very old Linux kernel:

 

synology mod's its kernel and uses it for all systems as base so they do not update to often to newer kernels

and if you think 4.4.59 is old, 3615/17 have 3.10.105 as base

7 hours ago, Hackaro said:

I'm wondering if/whether Synology DSM 7.0 will have the right version of Linux kernel to add Thunderbolt support; if it will be the case we can connect Macs over Thunderbolt and have a super fast access! What do you think

 

the closed preview version of 7.0 had base kernels 4.4.180 and 3.10.108 and i don't think they will switch to a new kernel in beta or final

(one of the reasons might be that dsm 7.0 is much older as synology worked unexpectedly long on that afair about 1.5 years over the planed release schedule by now ?)

 

 

Link to comment
Share on other sites

  • 0
On 11/25/2020 at 8:52 PM, IG-88 said:

you dont  need a switch, you can connect both nic's direct

i have a 1G and 10G in both nas and computer, the 1G in connected to a switch and the 10G connects directly between nas and pc and have static ip addresses of a different range, so when using the 10G ip address of the nas from pc the 10G connection will be used

10G switches are still expansive and often not passive cooled (beside noise level its also a sign for a heavier power consumption)

 

 

 

that's a good idea, thanks! I will experiment, do I need a cross cable? 

Link to comment
Share on other sites

  • 0
25 minutes ago, Hackaro said:

 

that's a good idea, thanks! I will experiment, do I need a cross cable? 

I don't know exactly if my cable its "cross", I've bought a Cat6 and works (Cat5 too, but if you will buy a Cable, better its Cat7 because prices between Cat6 and Cat7 itsnt make differences, but Cat7 performance its notoriously better. 

Link to comment
Share on other sites

  • 0
On 11/30/2020 at 3:45 PM, Hackaro said:

that's a good idea, thanks! I will experiment, do I need a cross cable? 

no, since 1G network the spec covers this so its automatic detection for direct connection

 

On 11/30/2020 at 4:14 PM, Mary Andrea said:

I don't know exactly if my cable its "cross", I've bought a Cat6 and works (Cat5 too, but if you will buy a Cable, better its Cat7 because prices between Cat6 and Cat7 itsnt make differences,

 

the detection for direct connection is not about the cable

offical its cat 6 for 10G but even a good 5e cable can do the job on shorter connections (5-10m?)

better "wire" does not always mean the whole cable is good, the plugs and sloppy work can also ruin the quality

 

Quote

but Cat7 performance its notoriously better. 

 

if you already have a 10G connection with a 6a cable (like 10m, the spec for 6a covers 100m) its not going to get faster with a cat 7 (or 7a/8) wire build into the cable

that will be more important if you use longer cable connections like 50-100m

 

the difference in wire quality between 6a and 7 is not that huge just 500MHz vs 600MHz bandwidth

7a (1000MHz) and 8 (2000MHz)  will have substantial better quality wires

so if you go beyond cat 6a it should at least be 7a to get anything worth the money

 

also if its about latency (like iSCSi, SAN) then it should not be 10G Base-T, SFP+ with DAC or optical fiber might be better in that regard and there are "cheap" 4 or 5 port SFP+ switsches since about a year now so on long term it might be better to start now with SFP+

i'm now stuck with 3 nic's with 10G Base-T and there is no cheap passive switch with at least 3 10G ports

10G Base-T is interesting in a changing/variable environment as you can also connect the cable to a 1G switch, thats not possible with SFP+

 

 

 

Edited by IG-88
  • Like 2
Link to comment
Share on other sites

  • 0
On 11/30/2020 at 9:00 PM, IG-88 said:

no, since 1G network the spec covers this so its automatic detection for direct connection

 

@IG-88 Sorry for the late reply but I needed a second 10Gbe card to connect things properly. At the moment I'm still using an Asus switch XG-U2008, it was very cheap at half the price so I got it, and now I have my 10Gb ring between Mac (Hackintosh actually) and Nas Xpenology . Here are some speed tests: 

 

https://www.dropbox.com/s/db6h5xp9vpl24lj/Screenshot 2020-12-10 at 19.32.11.png?dl=0

 

considering that: 

  • I'm already using jumbo frames, SMB2/3 without signing and 7 cables' type (one 5meters, the other one 1,5 meters), 
  • I'm copying a single big file (around 25GB ISO file), so I guess it's an ideal condition, 
  • btw AFP speed are more or less the same,
  • R/W cache on NAS is made by a pair of NVMe 1TB Crucial disks,
  • the source (in Hackintosh) is on an Crucial NVMe 1TB as well,

I'm pretty disappointed of the performances showed, I was quite confident to be able to get better speeds, especially in Write mode. Am I missing something? 

Edited by Hackaro
Link to comment
Share on other sites

  • 0
12 hours ago, Hackaro said:

I'm pretty disappointed of the performances showed, I was quite confident to be able to get better speeds,

whats the performace you see?

there are test programs like iperf  where you can test just the network performance (independent  from storage influence)

if that works out you can start looking for other source for low performance

when  writing big files with win10 and my nas i see ~1000MB/s for as long as the ram can be used as cache, after this it lower

 

m.2 nvme does not necessarily deliver great performance, the new QLC based devices can only hold up with the expected speed for the time until the TLC cache is filled, after that its measly because QLC is terrible at write speeds

Link to comment
Share on other sites

  • 0
On 12/18/2020 at 11:53 PM, IG-88 said:

whats the performace you see?

there are test programs like iperf  where you can test just the network performance (independent  from storage influence)

if that works out you can start looking for other source for low performance

when  writing big files with win10 and my nas i see ~1000MB/s for as long as the ram can be used as cache, after this it lower

 

m.2 nvme does not necessarily deliver great performance, the new QLC based devices can only hold up with the expected speed for the time until the TLC cache is filled, after that its measly because QLC is terrible at write speeds

 

I guess I have to study a bit more! ... that's the initial test' results:

https://www.dropbox.com/s/d7hlhkae5lc9p7a/Screenshot 2020-12-20 at 12.59.06.png?dl=0

 

so the network is the bottleneck here... ny NAS has only 8GB RAM so I really don't know about the ram as cache ... maybe I should consider another type of NVMe ... mine are those one: https://www.anandtech.com/show/13512/the-crucial-p1-1tb-ssd-review and I read the following: 

 

Quote

When the SLC cache fills up, writes to the P1 get very slow. The P1 doesn't bypass the cache when it is full, so everything written to the drive is written to SLC first before being folded into QLC blocks. (This helps the P1 offer similar partial power loss protection to the Crucial MX series of SATA SSDs.) The P1 also tends to keep data in SLC so it can serve as a read cache instead of aggressively folding data into QLC blocks during idle time.

 

too sad... :-(

I have to dig further on the NVMe technology, I guess. Thanks for now! I'll report back.

Edited by Hackaro
add info
Link to comment
Share on other sites

  • 0
4 hours ago, Hackaro said:

NVMe technology, I guess. Thanks for now! I'll report back.

its not about nvme, its about storage technology and cache strategy's

your link

https://www.anandtech.com/show/13512/the-crucial-p1-1tb-ssd-review/7

shows very clear how terrible such qlc base dirves can be if the cache is the only thing it relies on

its pretty much like with cmr disks and thencintruducing smr, parts of the smr disk (5%?) user cmr and are used as extended cache, when the cache is full and there are still data hammering the drive then it fallls back to its native speed and for smr thats pretty low, and the same goes for QLC cells, the write speed is verry low, that works ok for a typical desktop use but especially as cache drive its really not a good choice

 

4 hours ago, Hackaro said:

o the network is the bottleneck here...

NOT the bottleneck? if you have 9.9GBit of 10GBit then its maxed out already

 

4 hours ago, Hackaro said:

I really don't know about the ram as cache

 

you dont have to, afaik dsm does use ram as cache as long as its not used  for other things or is reserved

that get important if you measure performance just by copying files, you need to know the pipeline to interpret the results

 

 

  • Thanks 1
Link to comment
Share on other sites

  • 0
On 12/20/2020 at 5:32 PM, IG-88 said:

you dont have to, afaik dsm does use ram as cache as long as its not used  for other things or is reserved

that get important if you measure performance just by copying files, you need to know the pipeline to interpret the results

 

Sorry for the late reply, and Happy New year! :-)

 

I had a lot of troubles with my 10Gbe ring (CRC errors) and after many tests I ended up discovering that my Asus 10Gbe switch was the culprit. Connecting NAS and Mac peer-to-peer made all the errors disappears.

 

Now I sent back the switch and after thinking a bit about what we've discussed I've opted out for a 2,5Gbe ring instead of a 10Gbe, it's much more cheaper and it'll probably have performances that correctly match the Crucial NVMe disks I've bought. Maybe in the future I will consider selling my Crucial NVMe's and change them with much better (and expensive!) NVMe disks. But, for example, the Synology NVMe disks are enterprise's level and they are super expensive! 

 

My BigMac (Hack actually!) has already a 2,5Gbe Intel NIC onboard so this is also an advantage. Can you suggest me a Realtek 8125 card that can work well with Jun's loader for DS918+? Thanks! 

 

PS: Can I safely update to DSM 6.2.3-25426 Update 3 as well?

Edited by Hackaro
Link to comment
Share on other sites

  • 0
3 hours ago, Hackaro said:

Can you suggest me a Realtek 8125 card that can work well with Jun's loader for DS918+? Thanks! 

 

i dont have a card but with a 918+ the latest 8125 driver is compiling so no problems if its a newer version

 

3 hours ago, Hackaro said:

PS: Can I safely update to DSM 6.2.3-25426 Update 3 as well?

yes, afaik there was not special problem with that

 

Link to comment
Share on other sites

  • 0
On 1/17/2021 at 8:02 AM, Hackaro said:

 

I've opted out for a 2,5Gbe ring instead of a 10Gbe

I've got good performances with 2.5G, and didn't fully satisfied when I've had replaced it with Qnap10G Card. 

(Can you see comparative performance connecting a Mac (by thunderbolt3) to Server Xpenology alternatively with usb2.5gb (Realtek) and QXG-10G1T (Aquantia 107). I found Qnap Card transferences so unstable speed (start fast and after seconds down), and more stable on USB 2.5g.

comparative.png

Edited by Mary Andrea
Link to comment
Share on other sites

  • 0

 

On 1/18/2021 at 12:48 PM, Mary Andrea said:

I've got good performances with 2.5G, and didn't fully satisfied when I've had replaced it with Qnap10G Card. 

(Can you see comparative performance connecting a Mac (by thunderbolt3) to Server Xpenology alternatively with usb2.5gb (Realtek) and QXG-10G1T (Aquantia 107). I found Qnap Card transferences so unstable speed (start fast and after seconds down), and more stable on USB 2.5g.

 

 I didn't opt for that convoluted and expensive solution: I just bought a QNAP switch 2,5Gbe e connected all my devices (two hackintoshes, and NAS) directly to it. All with 2,5Gb NIC on board. Write and read are consistent at full max speed if the storage source is an SSD/Nvme disk. 

Link to comment
Share on other sites

  • 0

 

3 hours ago, Hackaro said:

 I didn't opt for that convoluted and expensive solution:

Certainly, 10g home networks is now something new, and newbies necessarily fall in an expensive solutions and configuration errors. It is for that reason that I have shared my experience, so others newbies can understand that spending three times more does not imply three times more speed. Sometimes a simple solution like yours is better. I'm glad you publish both your successes and your failures.

  • Like 1
  • Thanks 1
Link to comment
Share on other sites

  • 0
On 1/17/2021 at 3:29 PM, IG-88 said:

 

i dont have a card but with a 918+ the latest 8125 driver is compiling so no problems if its a newer version

 

yes, afaik there was not special problem with that

 

 

@IG-88
I can confirm that RTL8125 works like a charm ... I have also update DSM to 6.2.3U3 without problems. I've change both the two NVMes from Crucial to Samsung 970 EVO PLUS to have better cache's performance as I've have said, but I've gotten in troubles 🤬 as you can read here ... if you have any hint ... it's welcome! :-)

Link to comment
Share on other sites

  • 0

Hi, I added an Asus XG-C100C to my Hackintosh, which worked OOB and a IBM Emulex 49Y7952 2x10GbE card to my converted px12-450 running DSM 6.2.3-25426 Update 3.

As the Emulex is FTP+ and the Asus is RJ45 I've put a 6COMGIGA 10GBase-T SFP+ RJ45 copper transceiver Module in the Emulex, turned on jumbo frames and now I've got transfer rates from around 915 MB/s write and 990 MB/s read.

Edited by jollmo
Link to comment
Share on other sites

  • 0
4 hours ago, jollmo said:

Hi, I added an Asus XG-C100C to my Hackintosh, which worked OOB and a IBM Emulex 49Y7952 2x10GbE card to my converted px12-450 running DSM 6.2.3-25426 Update 3.

As the Emulex is FTP+ and the Asus is RJ45 I've put a 6COMGIGA 10GBase-T SFP+ RJ45 copper transceiver Module in the Emulex, turned on jumbo frames and now I've got transfer rates from around 915 MB/s write and 990 MB/s read.

That was with 6m Cat5e cable, with 15m Cat6A I get 950 write / 995 read...

  • Thanks 1
Link to comment
Share on other sites

  • 0

Because of the Rj45 / SFP+ mismatch, in-between I tried a Asus XG-100F, which I couldn't make recognised under OSX at all.

My second try was a Sonnet Solo 10G SFP+, which worked oob, but was very unreliable in connection and had read rates of max 30 MB/s.

That was with a 10m Rapidcon Twinax SFP+ cable...

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Answer this question...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...