Help with Math/Tech

[bdub76]

I currently have a DS920+ with WD Red Plus drives with a SHR raid of sizes 4TB and 6TB. I saw online that folks are using usb 3 to 2.5 ethernet dongles to increase their write speed. This seems like a cheap upgrade if you have a 2.5 switch. However, the internal write speeds of those drives would be much lower than the 2.5 ethernet. The data sheet says about 180M/s, which would be lower than the Ethernet speed in MB/s.

So would the upgrade only give me a theoretical increase around 55MB/s? Or is there something else I’m missing?

https://documents.westerndigital.com/content/dam/doc-library/en_us/assets/public/western-digital/product/internal-drives/wd-red-plus-hdd/product-brief-western-digital-wd-red-plus-hdd.pdf

GitHub - bb-qq/r8152: Synology DSM driver for Realtek RTL8152/RTL8153/RTL8156 based adapters

Synology DSM driver for Realtek RTL8152/RTL8153/RTL8156 based adapters - bb-qq/r8152

github.com

 

[Tech9]

I wouldn't recommend any USB-to-Ethernet dongles use with high-reliability expected devices.

 

[Tech Junky]

So would the upgrade only give me a theoretical increase around 55MB/s? Or is there something else I’m missing?

If you want speed build a PC, throw the drives in or upgrade them, and put a 5ge nic into it for faster speeds with raid 10. R10 will push you to 400MB/s and sell the 920 to make up for most of the cost.

 

[bdub76]

If you want speed build a PC, throw the drives in or upgrade them, and put a 5ge nic into it for faster speeds with raid 10. R10 will push you to 400MB/s and sell the 920 to make up for most of the cost.

The drives have to be able to write the throughput. None of the HDs in the doc above come close to the throughput you're talking about. I'd have to go with SSDs, which are currently about 3ish times more expensive. Gigabit Ethernet with overhead is still below the max read/write speeds of the HDs I own. But I don't have a 2.5 nic to max them out. I doubt they'll actually do the 180 MB/s as advertised. Likewise, my WIFI 6 network doesn't come close to the 180 MB/s either.

The fastest solution I can think of today is a M2 NVME SSD connected to a Thunderbolt port such as this device:

https://www.amazon.com/gp/product/B07N48N5GR/?tag=snbforums-20

I struggle with the use case for 2.5ge or even 5ge if you can't read and write at those speeds without much more expensive media. Both of those network speeds exceed the read and write speeds of HD media. And I really want faster writes. I don't need faster reads since I'm streaming the media. And all streaming today can me done with a 100 Mbps and slower.

Or maybe I'm missing something. How can you read and write faster than the drive maximum?

 

[L&LD]

How can you read and write faster than the drive maximum?

RAID0, RAID10, etc. 2 or more drives in a form of RAID0 will multiply the single drive's rates.

As for the maximums, only expect that if the drives are less than ~10% full. As your array/drives fill up, they will slow down as the inner tracks are used on the drive platters.

Free RAID Calculator - Caclulate RAID Array Capacity and Fault Tolerance. (raid-calculator.com)

 

[bdub76]

RAID0, RAID10, etc. 2 or more drives in a form of RAID0 will multiply the single drive's rates.

As for the maximums, only expect that if the drives are less than ~10% full. As your array/drives fill up, they will slow down as the inner tracks are used on the drive platters.

Free RAID Calculator - Caclulate RAID Array Capacity and Fault Tolerance. (raid-calculator.com)

But I'm still limited by the 1Gb/s Ethernet, which is is around the low 100-125MB/s after packet overhead. And that's slower than the drive max.

So how do I benefit from the faster RAID speed when the network is so much slower?

Am I thinking about this correctly? If I were to go to SSDs, I'd gain the benefit of more reliability, but I'm limited by network speeds. 1Gb/s Ethernet isn't fast enough for 400-500 MB/s reads and writes. So I can't realize those faster read and write speeds of those drives. In order for this to work, I'd really need 10Gb/s Ethernet. But I can't do anything over WIFI that's close to those speeds.

 

[Tech Junky]

Or maybe I'm missing something. How can you read and write faster than the drive maximum?

As L&LD pointed out R0/R10 multiply the R/W speed per additional disk or pair in the case of R10. This isn't something you see with other modes though.

WIFI I can get 1.5gbps but, through a wired connection I'm able to hit those speeds of 400MB/s+ using a USBC 5GE adapter on my laptop to the 5GE port on the server that has the disks in it.

But I'm still limited by the 1Gb/s Ethernet

This is why you need to build your own "NAS" and put a proper NIC inside if you have clients that need that kind of bandwidth. Anything is possible when you release the disks from a "NAS" and put them in a PC and expand beyond the cheap HW NAS OEM's use. ~$700 for a 920 is outrageous for the CPU/RAM/NIC when you could take that $700 and have so much better performance and not be locked into a 4 bay enclosure.

 

[bdub76]

WIFI I can get 1.5gbps but, through a wired connection I'm able to hit those speeds of 400MB/s+ using a USBC 5GE adapter on my laptop to the 5GE port on the server that has the disks in it.

What adapter are you using?

 

[Tech Junky]

What adapter are you using?

Intel AX411 which requires a 12/13th gen CPU due to it being cnvi2. They added ~300mbps to the speed by combining 2.4/5 bands into a single pipe. They're dirt cheap at $18.

 

[L&LD]

But I'm still limited by the 1Gb/s Ethernet, which is is around the low 100-125MB/s after packet overhead. And that's slower than the drive max.

So how do I benefit from the faster RAID speed when the network is so much slower?

Am I thinking about this correctly? If I were to go to SSDs, I'd gain the benefit of more reliability, but I'm limited by network speeds. 1Gb/s Ethernet isn't fast enough for 400-500 MB/s reads and writes. So I can't realize those faster read and write speeds of those drives. In order for this to work, I'd really need 10Gb/s Ethernet. But I can't do anything over WIFI that's close to those speeds.

Yes, you have to design a whole system to get the speeds you want. A network is a chain, the weakest link rules those top speeds.

SSDs are not more reliable than HDDs (depending on which models are being used).

You can get 10Gbps speeds, wirelessly. But again, you need to pony up the cash to get the hardware to do so (see the latest announcements of WiFi 7 class routers).

If you're looking for a single client to hit ~500MB/s, you're limited to 5GbE and higher wired solutions.

If you're looking for multiple clients to achieve those speeds in aggregate, then look for LAG solutions, to get closer to those speeds.

I don't think you've stated why you need these speeds?

In any case, prepare to exponentially increase your budget for each doubling of the current speeds you're seeing today.

Speed isn't cheap. And there are no shortcuts.

I would not recommend a DIY NAS if this will be your first experience with one. And, you value the data you want it to store.

 

[bdub76]

Yes, you have to design a whole system to get the speeds you want. A network is a chain, the weakest link rules those top speeds.

SSDs are not more reliable than HDDs (depending on which models are being used).

You can get 10Gbps speeds, wirelessly. But again, you need to pony up the cash to get the hardware to do so (see the latest announcements of WiFi 7 class routers).

If you're looking for a single client to hit ~500MB/s, you're limited to 5GbE and higher wired solutions.

If you're looking for multiple clients to achieve those speeds in aggregate, then look for LAG solutions, to get closer to those speeds.

I don't think you've stated why you need these speeds?

In any case, prepare to exponentially increase your budget for each doubling of the current speeds you're seeing today.

Speed isn't cheap. And there are no shortcuts.

I would not recommend a DIY NAS if this will be your first experience with one. And, you value the data you want it to store.

My backbone network gear is almost five years old, so I'm reviewing it for potential upgrades. But the math doesn't make sense to me. I upgraded the NAS last year from a 7 year old two bay Synology. I didn't want to risk drive failure, and since the hardware was old, I purchased the DS920+, which I have been mostly happy with. I picked up a AX86U as an AP to replace my old r7000p running DD-WRT. Kong wasn't releasing any new updates, and I wanted to refresh it. I also wanted a router capable unit as a fail over in case my switch and router go bad. I'm still running a 6 year old HP 1g managed switch, and a 5 year old PC Engine APU2 as a gateway.

I'm thinking about going to 2.5g as the Protectli units that I'm looking at come now with 2.5g nics. But given that I gain no benefit from the faster speed since the only part of my network, where I use any speed at all is my NAS. That's why I brought up that dongle as a possibility. And my Internet is a only 250/10. The 250 for my use cases is overkill. I mostly stream. The only benefit I get from 250 is updating PS5 games. And the 10 is terrible. It makes backing up any substantial data to a cloud service painful.

The reality is that speeds are mismatched mess for consumer gear right now. It doesn't make sense to me as I look at upgrades to replace old equipment.

 

[Tech Junky]

speeds are mismatched mess for consumer gear

They always will be and that's when you get creative.

For the ISP you might consider switching to T-Mobile or VZW. In TM get 70-100UL and DL average of 200+ for $50/mo.

If you run newer HDD you can see per disk of up to 235MB/s depending on the disk. There's also the new Seagate mach2 drives that can hit 500+ with dual actuators coming soon.

If you're going to overhaul you need to figure out what's a priority and then work your way from there. The NAS is your bottleneck for data speeds at this point. There are some cheap 2.5 switches for under $200 and closer to $100 with fewer ports.

So, if you only have a single device that needs higher speeds for backups or pulling data direct wiring that to the disks would be cheap and simple. If that's the case then just using an enclosure would make more sense than a NAS. Maybe even a DAS that's more robust and capable of raid would be a better idea.

List out what you have currently and see where things lead as to planning. Not everything needs to be at max speed form the sound of it so far. But yes if there's a slow link somewhere in the path then that's the most you can expect to get if the data has to pass through it.

Kind of like these goofy routers that can push 3gbps on wifi and then they put a 1gbps port on them.

 

[L&LD]

The reality is what we make it. Manufacturers don't build for our specific needs, we choose what is the closest match to those needs.

You need to upgrade your NAS to 2.5GbE speeds or greater.

You need to upgrade your router to at least a GT-AX6000 or better.

You may possibly need to upgrade your cable runs, and/or switches to match the speed of the hardware you decide on above.

None of the above needs to be done all at once. But the design of the network needs to be solid, and firm, from the start.

Two years ago, I was enjoying 1Gbps symmetrical ISP speeds. With a single RT-AX86U connected to my 2.5GbE NAS. Every client that was used to connect to the NAS was noticeably more responsive. Not faster though (they were still limited to 1GbE speeds, of course).

When I got two RT-AX86Us, and used them in wired (2.5GbE) backhaul mode, the difference when accessing the NAS was noticed (via the 1GbE connections everything else, wired, on the network had access to). Surprisingly, my Intel AX210 equipped laptop was the best connection to the NAS, over WiFi.

I bought two 2.5GbE switches since then and have a vastly upgraded network experience now, wired or wirelessly. With that same 1Gbps ISP connection too.

When/if, I have the chance to upgrade the network again, all the equipment I have today, will be replaced. The new network will be expensive, but it will also be built to purpose too.

If you're thinking today of ~500MB/s speeds, don't consider 2.5GbE equipment at all. Build the network you want, once. Even if it takes a little longer to get there. The end result will be worth the trouble.

 

[bdub76]

The reality is what we make it. Manufacturers don't build for our specific needs, we choose what is the closest match to those needs.

You need to upgrade your NAS to 2.5GbE speeds or greater.

You need to upgrade your router to at least a GT-AX6000 or better.

You may possibly need to upgrade your cable runs, and/or switches to match the speed of the hardware you decide on above.

None of the above needs to be done all at once. But the design of the network needs to be solid, and firm, from the start.

Two years ago, I was enjoying 1Gbps symmetrical ISP speeds. With a single RT-AX86U connected to my 2.5GbE NAS. Every client that was used to connect to the NAS was noticeably more responsive. Not faster though (they were still limited to 1GbE speeds, of course).

When I got two RT-AX86Us, and used them in wired (2.5GbE) backhaul mode, the difference when accessing the NAS was noticed (via the 1GbE connections everything else, wired, on the network had access to). Surprisingly, my Intel AX210 equipped laptop was the best connection to the NAS, over WiFi.

I bought two 2.5GbE switches since then and have a vastly upgraded network experience now, wired or wirelessly. With that same 1Gbps ISP connection too.

When/if, I have the chance to upgrade the network again, all the equipment I have today, will be replaced. The new network will be expensive, but it will also be built to purpose too.

If you're thinking today of ~500MB/s speeds, don't consider 2.5GbE equipment at all. Build the network you want, once. Even if it takes a little longer to get there. The end result will be worth the trouble.

But there has to be a practical use case. That’s the math problem.

The reality is that I don’t have one. Exceeding my NAS speed doesn’t provide me any benefit. Half my Internet usage is streaming, which is well below my 250 service today and well below 1g Ethernet. I can’t even saturate my current Internet speeds with torrents. I think, this is due to a lot of people having crappy upload speeds. My biggest bottleneck is my upstream for Internet. Increasing that amount would give me the opportunity to better use cloud backup services for media.

It’s a waste of money to upgrade while my current gear still works. My current upgrades are due to age and lack of future support. Any WIFI device in my house with WIFI 6 can hit my max Internet speed, so I’m good.

The current problem with WIFI is lots of devices due to the rise of IOT. The problem is that the incremental devices aren’t manufactured with the latest radios, so they don’t benefit from tech that’s supposed to help with congestion. I have 20 connected devices. And a lot of those are still on WIFI 4.

What’s your use case? Why do you have gigabit Internet?

 

[L&LD]

The lowest latency possible is my use case.

I don't have time to wait an extra second for the next 'page' in an article to appear. Nor for the updates my computers need daily, weekly, and monthly. Not to mention the cloud backup I use for certain files too.

Yes, speed has a cost. But unlimited time, I do not have (so that cost seems less than negligible, to me).



The network I've been building since I started using RMerlin firmware has been all about lower latency (whether I knew what that was or not, back then).

Airport Extreme (= molasses network), Netgear (huge step up from the airport, but laughable when the RT-N66U was set up).

All the way to the RT-AX86U and the GT-AX6000 I use today (with full 2.5GbE LAN connectivity) is because I'm chasing the lowest latency possible for my network.

If/when I have a faster ISP option, I'm sure I will do that too, for the same reasons.

The math problem you're describing is merely the unwillingness to spend a few dollars more to see if you see a benefit for your use cases. Almost any decision you make can be reversed. But theoretical/number analysis won't give you the full picture either.

The cost of doing this has been low(er) for me because I can sell the previous equipment quickly, and at a fair price (to me and the buyer). The benefits are immense though. And noticed even when I'm not on my network (while working elsewhere). But, anyone else can do the same.

A network, like the internet, is not a stagnant part of our world (it never has been). It is in our best interests to keep them updated and current. If we want to truly enjoy, rather than just endure them.

 

[bdub76]

They always will be and that's when you get creative.

For the ISP you might consider switching to T-Mobile or VZW. In TM get 70-100UL and DL average of 200+ for $50/mo.

If you run newer HDD you can see per disk of up to 235MB/s depending on the disk. There's also the new Seagate mach2 drives that can hit 500+ with dual actuators coming soon.

If you're going to overhaul you need to figure out what's a priority and then work your way from there. The NAS is your bottleneck for data speeds at this point. There are some cheap 2.5 switches for under $200 and closer to $100 with fewer ports.

Kind of like these goofy routers that can push 3gbps on wifi and then they put a 1gbps port on them.

You mean like the AX86S? In theory you should be able to get 1.2g with 2x2 at 80 with WIFI 6. But that’s practically impossible, so the vendor is probably like why bother with a faster nic.

But it also doesn’t make sense that the other routers only have one 2.5g either. It should be all of them.

 

[bdub76]

The lowest latency possible is my use case.

I don't have time to wait an extra second for the next 'page' in an article to appear. Nor for the updates my computers need daily, weekly, and monthly. Not to mention the cloud backup I use for certain files too.

Yes, speed has a cost. But unlimited time, I do not have (so that cost seems less than negligible, to me).



The network I've been building since I started using RMerlin firmware has been all about lower latency (whether I knew what that was or not, back then).

Airport Extreme (= molasses network), Netgear (huge step up from the airport, but laughable when the RT-N66U was set up).

All the way to the RT-AX86U and the GT-AX6000 I use today (with full 2.5GbE LAN connectivity) is because I'm chasing the lowest latency possible for my network.

If/when I have a faster ISP option, I'm sure I will do that too, for the same reasons.

The math problem you're describing is merely the unwillingness to spend a few dollars more to see if you see a benefit for your use cases. Almost any decision you make can be reversed. But theoretical/number analysis won't give you the full picture either.

The cost of doing this has been low(er) for me because I can sell the previous equipment quickly, and at a fair price (to me and the buyer). The benefits are immense though. And noticed even when I'm not on my network (while working elsewhere). But, anyone else can do the same.

A network, like the internet, is not a stagnant part of our world (it never has been). It is in our best interests to keep them updated and current. If we want to truly enjoy, rather than just endure them.

The biggest impact on my network has been running an unbound DNS cache server on my APU2. This is how I get fast load times locally.

 

[Tech Junky]

You mean like the AX86S? In theory you should be able to get 1.2g with 2x2 at 80 with WIFI 6. But that’s practically impossible, so the vendor is probably like why bother with a faster nic.

But it also doesn’t make sense that the other routers only have one 2.5g either. It should be all of them.

For the masses at a price point they'll pay is the game in the consumer arena.

Higher speed ports generate more heat in the plastic case that doesn't have active cooling. The OEM/Asus doesn't want to price out the cheap people by adding more heat.

If you step up to smb level gear you get performance because stupid stuff like noise usually isn't a concern when the background noise levels drown it out.

My 5ge nic ports idle at 130F and require fans to keep them cool but, the system is dead silent with proper cooling. If you do it right then it's not an issue.

 

[bdub76]

The lowest latency possible is my use case.

I don't have time to wait an extra second for the next 'page' in an article to appear. Nor for the updates my computers need daily, weekly, and monthly. Not to mention the cloud backup I use for certain files too.

Yes, speed has a cost. But unlimited time, I do not have (so that cost seems less than negligible, to me).



The network I've been building since I started using RMerlin firmware has been all about lower latency (whether I knew what that was or not, back then).

Airport Extreme (= molasses network), Netgear (huge step up from the airport, but laughable when the RT-N66U was set up).

All the way to the RT-AX86U and the GT-AX6000 I use today (with full 2.5GbE LAN connectivity) is because I'm chasing the lowest latency possible for my network.

If/when I have a faster ISP option, I'm sure I will do that too, for the same reasons.

The math problem you're describing is merely the unwillingness to spend a few dollars more to see if you see a benefit for your use cases. Almost any decision you make can be reversed. But theoretical/number analysis won't give you the full picture either.

The cost of doing this has been low(er) for me because I can sell the previous equipment quickly, and at a fair price (to me and the buyer). The benefits are immense though. And noticed even when I'm not on my network (while working elsewhere). But, anyone else can do the same.

A network, like the internet, is not a stagnant part of our world (it never has been). It is in our best interests to keep them updated and current. If we want to truly enjoy, rather than just endure them.

How are you mitigating latency? What’s your target?

For the masses at a price point they'll pay is the game in the consumer arena.

Higher speed ports generate more heat in the plastic case that doesn't have active cooling. The OEM/Asus doesn't want to price out the cheap people by adding more heat.

If you step up to smb level gear you get performance because stupid stuff like noise usually isn't a concern when the background noise levels drown it out.

My 5ge nic ports idle at 130F and require fans to keep them cool but, the system is dead silent with proper cooling. If you do it right then it's not an issue.

I looked at some 2.5G switches, and the biggest complaint has been heat. My HP Procurve Switch is a brick. But it’s been reliable

The other factor is to keep power usage down. That’s why I use the APU2 instead of a full blown tower.

I prefer business gear, but keeping it updated is far more painful than necessary.

 

[Tech Junky]

I looked at some 2.5G switches, and the biggest complaint has been heat. My HP Procurve Switch is a brick. But it’s been reliable

The other factor is to keep power usage down. That’s why I use the APU2 instead of a full blown tower.

I prefer business gear, but keeping it updated is far more painful than necessary.

Heat is always something that comes into play no matter the speed. You could take the apu2 and throw it in a bigger case and keep the benefits of your current setup and just add the additional gear needed to spin up a NAS.

If you only need a couple of devices to have higher speed to the data then size it that way and just use the 1ge for everything else.

I run my system as the router as well since before switching from a gig ISP connection it allowed binding two ports and getting the full bandwidth from them which was another 300-400mbps.

I combined the router, switch, NAS, dvr, and some other appliances into a single box this way and dropped the electrical from 5-6 devices to the single PC.