[Eug]

I am hardwiring one of my rooms to another. One room has an iMac with GigE. The other has a MacBook with GigE.

I am playing with the idea of getting a couple of GigE switches so that the iMac and (Penryn) MacBook can communicate with each other at Gigabit speeds... while residing on a network with a 100 Base-T router, with the following topology:

iMac
|
|
GigE switch #1 ---- 100 Base-T router
|
|
GigE switch #2
|
|
MacBook

This should give me GigE speeds from the MacBook to the iMac. If correct, what should I be looking for in a Gigabit switch? Should I just forget about jumbo frames? I'm thinking that enabling jumbo frames could bork everything, since the other hardware on the system will all be Fast Ethernet, not Gigabit.

[Big Mac]

WIFI not fast enough?

[Eug]

Originally Posted by Big Mac 

WIFI not fast enough?

My WiFi is 802.11g, and even if I were to spend the money on 802.11n, the range and consistency is not good enough.

(WiFi also adds latency, but I don't actually game much online with my Xbox 360 now so that's less important.)

[Big Mac]

I think Jumbo Frames would only work properly if all your equipment were Gigabit enabled.

[Simon]

Umm, why the need for two Gigabit switches?

[Eug]

Originally Posted by Big Mac 

I think Jumbo Frames would only work properly if all your equipment were Gigabit enabled.

Thx. That's what I figured.

Originally Posted by Simon 

Umm, why the need for two Gigabit switches?

In both rooms, there are multiple clients. I need a gigabit switch at both ends in order to maintain the gigabit link between the MacBook and iMac and at the same time be able to split the link for the rest of clients in those rooms (albeit at 100 Base-T speeds).

So, in fact, this is the true topology:

iMac
|
|
GigE switch #1 <---- 100 Base-T DSL router with integrated 802.11g access point ----> Multiple clients
|
|
GigE switch #2 ----> 100 Base-T switch and integrated 802.11g access point ----> Multiple clients, including an Xbox 360
|
|
MacBook

I could keep the existing trunk in the MacBook room independent from the MacBook, and just use the gigabit link for the MacBook as a second trunk. However, that room is connected by powerline networking, and (since I got some electrical work done on the house) I'm only getting 10 Base-T speeds. That's fine for internet access and gaming, but I'd like to get 100 Base-T speeds for the Xbox 360, for media streaming.

BTW, gaming is great on powerline networking, since you don't get the latency that you get with wireless. However, speed for video streaming can be a problem, depending on the wiring.

Here is a map of my existing network topology. The iMac is in the Office, and the MacBook is in the Garage.

[mduell]

You can only use jumbo frames if all the devices on the subnet support jumbo frames; it's probably not worth the effort and complexity to put the iMac and MacBook on a different subnet.

That said, in the consumer/soho space even the cheapest decent gigabit switches (I've had a good experience with D-link DGS-220x which are about $5/port) support jumbo frames so you'll be ready to use them should your entire subnet be jumbo-capable in the future.

[Eug]

OK. I've ordered a couple of D-Link DGS-2205 units. We'll see how it goes.

I'll forego the subnet idea, so no jumbo frames. Hopefully I will still get 200+ Mbps out of it without jumbo frames though. I don't actually need GigE for my media streaming since Fast Ethernet should be good enough, but it would be nice to move a 350 MB data file in 10-15 s. Right now it takes me around 40 s to do that.

Now I just need my gel-filled outdoor CAT5e to arrive, so I can actually lay the cable.

[ghporter]

Outdoor Cat5? You're wiring an outbuilding or something? The "conventional wisdom" is to use Cat6 with gig, but it's unlikely that high-end Cat5e will give you measurably lesser performance than standard Cat6.

[Eug]

The easiest route from my home office to my garage is outside. (Actually, it's about the only reasonable route, without ripping multiple walls apart.) My garage is actually now a home theatre, and my home office is where all my media files are.

I couldn't find outdoor-rated CAT6 locally, and gel-filled outdoor CAT6 is even harder to find. I didn't feel like doing the eBay thing either. However, I did find an electrical distributor here that sells gel-filled outdoor CAT5e. Outdoor cable is has a thicker plastic coat, is UV rated, and has better resistance to cold temps. Gel-filled outdoor cable is also resistant to water, and can even be directly buried. I'm not going to bury it, but it will be near ground level, and will be covered in snow during the winter months. Thus, I figured better safe than sorry. Apparently, it's a bit gooey to work with, but I'm told that it otherwise crimps like any other CAT5e. BTW, I crimped very first CAT5e plug this week.

As for GigE, I'm only going to be using consumer grade equipment... with standard consumer hard drives, so as long as I can get over 200 Mbps I'll be happy. 300 mbps would be preferable, but that's just gravy.

P.S. After getting some CAT6 patch cables, and checking out some CAT6 at the local electronics store, I'm convinced it would be more annoying to work with. There are a lot more twists internally, so the cable is less flexible, and it's bigger. Plus, some of the CAT6 out there is 23 gauge, which apparently doesn't work well with CAT5e plugs since the latter are built for 24 gauge or smaller. CAT6 RJ45 plugs are fairly expensive too, but at least you can find them. I haven't seen CAT6 keystone jacks anywhere locally, just CAT5e.

[Simon]

Originally Posted by Eug 

As for GigE, I'm only going to be using consumer grade equipment... with standard consumer hard drives, so as long as I can get over 200 Mbps I'll be happy.

But isn't that roughly what a pair of decent powerline adapters would achieve?

[Eug]

Originally Posted by Simon 

But isn't that roughly what a pair of decent powerline adapters would achieve?

No.

Powerline will give you a theoretical max of 200 Mbps, which means more like 100+... on the same electrical circuit assuming your wiring was good. If you're across the house it is much slower.

Benchmarks in cnet's setup:

In our Labs' tests, the Netgear Powerline HD Ethernet Adapters clocked an impressive 52.7Mbps throughput rate--faster than that of the Panasonic adapters, which posted a 35.4Mbps rate.

More benchmarks:



Here are my own benchmarks, before the electrical work was done.



I had about 30 Mbps on powerline, until I had that electrical work done (new circuits, A/C, baseboard heating in the garage, etc). Unfortunately, now I've dropped to below 9 Mbps average. That's actually not bad since my office and garage are actually on separate breaker panels, but it's not good enough for media streaming. It's fine for internet access and gaming though (no lag). Since it's still OK for internet access, I'm gonna keep the rest of the house on power line networking (with attached 802.11g access points).

[Simon]

Ugh. I had no idea powerline adapters were that crap. I once tried an inexpensive pair and saw short of 100Mbps but that was on brand new and very well done wiring. I'm glad I stuck with a combination of wired Gigabit and 802.11n routers at my place.

[Eug]

Well to be fair, as you know it's not necessarily the powerline adapters per se that are crap. It's the interference on the wiring, even if the wiring is good. Stick an air conditioner, etc. on the same breaker panel and watch your speeds go down.

The other option that should be out soon is Ethernet over coax. We've come full circle. Coax went away 10 years ago, and now it's back again... with a twist. You can get real-life speeds well over 100 Mbps, while still being able to watch HD TV and access internet broadband on the same cable line. However, I didn't want to wait, and I didn't want to pay over $100 each for two of these.

[Eug]

I've been trying to run iperf on OS X, and that necessitates installing Darwin Ports 1.60 on a Core 2 Duo iMac. However, I never get past the install process. It fails during the install.

Is this a known issue? I can't seem to find a direct reference to it.

Finally I gave up and just deleted the opt subdirectory (which is the Darwin Ports install directory).

[Eug]

Booyah!!

Installed the new CAT5e cable today. Wow this outdoor gel-filled stuff is EXTREMELY stiff. The outer shell is very thick and stiff, and of course the innards are completely full of goop. It's almost comparable in terms of (lack of) flexibility as coaxial cable for cable TV. I crimped a standard no-name RJ-45 connector at one end, and put a Leviton CAT5e keystone jack at the other end.

I also got my two $19.99 refurb D-Link DGS-2205 GigE switches. The RJ-45 end is directly connected to one switch, and the keystone end is attached to the other switch via a CAT6 patch cable.

Did file transfer tests 363.5 MB, or 3.812 million bytes, or 3049.3 million bits, from my C2D iMac to my C2D white MacBook. Patch cable to iMac is CAT5e. Patch cable to MacBook is CAT6.

Without the Gigabit switches on newly installed cable: 34.2 seconds = 89 Mbps
With the Gigabit switches on newly installed cable: 7.6 seconds = 401 Mbps <-- I assume jumbo frames are OFF, but I can't find the setting anywhere.

With only one GigE switch and other CAT5 (not CAT5e) cable): 7.6 seconds = 401 Mbps <-- I'm surprised.
With only one GigE switch and other CAT5 cable, and a cheap patch cable to MacBook: 33.8 seconds = 90 Mbps <-- I threw out that patch cable

BTW, that works out to 48 MB/s. I didn't even know my laptop hard drive could write that fast. I'll have to try another computer at some point, because I suspect my laptop's hard drive could even be limiting the speed. I'd try iPerf, but I can't get Darwin Ports to install.

I knew I had a good chance of hitting 300+ Mbps, but I never thought I'd hit 400 Mbps.

EDIT:

iperf tells me I get 895 Mbps.

[Simon]

If you're sure you have quality cables, 400 Mbps definitely sounds like a HDD bottleneck. At work I have seen Gigabit speeds just short of 900 Mbps across switches.

From what I gather the only notebook HDD to come even close to that would be the recently released Momentus 7200.4. Before that, the fastest 2.5" HDDs would see sustained read/writes no higher than about 600 Mbps.

[Eug]

Yeah, I mentioned above that iperf is giving me 895 Mbps.

iperf



I believe what it does is send data for 10 s, and measures the throughput. I was shocked to see I transferred over 1 GB in that 10 s, not only because I wasn't confident my n00b crimping skillz with CAT5e would be up to the task, but also because I have strictly consumer hardware: iMac, MacBook, $19.99 switches, and many reports out there have consumer hardware giving speeds more in the 300-400+ range. I assume I don't have jumbo frames active either (but I don't even know how to set that anyway).

[mduell]

ifconfig | grep mtu will let you determine if you're using jumbo frames or not.

Glad to hear it worked out with the switches.

[Eug]

Normal frame size of 1500:

Code:
Eug-iMac-C2D:~ eug$ ifconfig | grep mtu
lo0: flags=8049<UP,LOOPBACK,RUNNING,MULTICAST> mtu 16384
gif0: flags=8010<POINTOPOINT,MULTICAST> mtu 1280
stf0: flags=0<> mtu 1280
en0: flags=8863<UP,BROADCAST,SMART,RUNNING,SIMPLEX,MULTICAST> mtu 1500
en1: flags=8823<UP,BROADCAST,SMART,SIMPLEX,MULTICAST> mtu 1500
fw0: flags=8802<BROADCAST,SIMPLEX,MULTICAST> mtu 2030
en2: flags=8822<BROADCAST,SMART,SIMPLEX,MULTICAST> mtu 1500
en3: flags=8822<BROADCAST,SMART,SIMPLEX,MULTICAST> mtu 1500
Eug-iMac-C2D:~ eug$

I'm not entirely sure what en2 and en3 are.

[Simon]

Do you use virtualization?

[Eug]

I use Parallels, but it wasn't actually running at the time.

[Simon]

Parallels creates additional network interfaces when you install it.

[Eug]

OK thx, that makes sense.

P.S. After doing some more testing... I was lying when I said powerline has no lag. It's just a lot less lag than what you usually get with wireless. With some wireless, the lag sometimes gets into the perceptible range during gaming. Not so for powerline unless you're dropping packets or something.

However, for my new CAT5e line, the lag is now only 0.3 ms. Nice.

Oh and those D-Link switches seem to run very cool, and that so-called dlink green power saving doesn't seem to affect functionality in any way. I'm impressed. I wonder how much power it actually saves though. Probably marginal.

[mduell]

I'm extremely skeptical of the power savings of the "green power" switches, but they work, so I don't care.