1 |
"Boyd Stephen Smith Jr." <bss03@××××××××××.net> posted |
2 |
200609300330.48229.bss03@××××××××××.net, excerpted below, on Sat, 30 Sep |
3 |
2006 03:30:44 -0500: |
4 |
|
5 |
[on Linksys WRT compatible routers] |
6 |
|
7 |
> I've had good luck running OpenWRT on the older models (v1.1 and v2). The |
8 |
> newer non-L(inux) models [which can have Linux installed on them by the |
9 |
> user, after killing VxWorks] have fewer resources, so DD-WRT might be more |
10 |
> appropriate, but I don't have any experience with that. The newer L(inux) |
11 |
> models come at a premium and may not be carried everywhere, but OpenWRT is |
12 |
> a fine choice for them. |
13 |
|
14 |
If I go that route, I'd either get the Linux model or a compatible unit |
15 |
from someone else. (I know they exist but don't know specifically which |
16 |
ones are and are not compatible with the WRT firmwares, at this time, so |
17 |
if I were buying now, it'd be the Linksys Linux model.) |
18 |
|
19 |
The one I have, one of the first generation routers, a Netgear rt314 (OEMed |
20 |
zyxel p314), that cost me $200 when I bought it years ago. It has been |
21 |
old faithful and would likely continue to be old faithful for many more |
22 |
years, and that was the going price at the time so I've certainly |
23 |
gotten my money's worth, no complaints there, but it's just getting to be |
24 |
too slow -- only 10Mbit Ethernet WAN side, tho 100Mbit LAN side. |
25 |
|
26 |
Anyway, given the $200 I paid for the old one, even the premium $80-ish |
27 |
I'd pay for a Linux version Linksys now doesn't seem unreasonable, given |
28 |
the increased flexibility I'd have with it. |
29 |
|
30 |
> If the load on the LAN will be large, be aware that the WRT hw does have |
31 |
> some issues under heavy load, particularly when the number of TCP |
32 |
> connections being created is high which happens with some P2P clients. |
33 |
|
34 |
It wouldn't be that large, really. It's for personal use, and I tend to |
35 |
put all my money/time/energy into a single machine at once, so I don't |
36 |
tend to have a lot of LAN traffic. However, I /am/ learning skills and |
37 |
techniques I might end up putting to use elsewhere at some point, and even |
38 |
if not, it's better than even money that I'd end up replying to someone on |
39 |
a newsgroup or the like that /was/ using it for something more major -- |
40 |
that always seems to be happening <g>, so I try to keep information |
41 |
appropriate to those possibilities filed away in my head. <g> Thus, |
42 |
regardless of my personal use, it's useful info. Thanks. |
43 |
|
44 |
>> (if it's not, it'll be a full computer running a conventional x86 or |
45 |
>> x86_64/amd64 based Linux kernel), bought with the intent of upgrading it |
46 |
>> to one of the several open source firmware alternatives available. |
47 |
> |
48 |
> If you need GigE speeds for the LAN, be sure to find a main board with |
49 |
> good buses. No PC NICs that I know of support HW level routing at any |
50 |
> layer, so every packet that is routed has to cross your bus twice, once |
51 |
> going to memory and once going back to that (or another) NIC. The |
52 |
> standard PCI bus (not PCI-X or PCI-e) can only support 4 Gbps in the best |
53 |
> of conditions, which means only about 2 GigE clients. |
54 |
|
55 |
As I said, not likely stuff I personally need to worry about, but useful |
56 |
information even so. For me, most traffic will be across the LAN/WAN |
57 |
barrier, not LAN side, and 100Mbit Fast Ethernet should be plenty for |
58 |
several more years anyway, WAN side. As I said, my current router is |
59 |
10Mbit Ethernet WAN side, but Cox just upped the cap to 7Mbit down from 6, |
60 |
here, and the premium service /was/ 9Mbit -- it might be 10 now tho I |
61 |
haven't checked. I know my router was/is handling 6Mbit across the |
62 |
WAN/LAN barrier just fine, but either the 7Mbit hasn't yet taken effect or |
63 |
6Mbit is top of the line for this router. I know it's getting close as |
64 |
7-8Mbit is pretty close to top thruput on 10Mbit Ethernet in any case, |
65 |
even when it's /not/ across a LAN/WAN router barrier. Thus, even if I get |
66 |
lucky and it'll actually do 7Mbit, I don't expect it'll take the next |
67 |
jump, so it's time to upgrade. However, given I was on 608kbit DSL when I |
68 |
got this router, and it has taken me thru 6Mbit anyway, over five plus |
69 |
years, far more than the Linksys (for example) of the same generation did, |
70 |
I figure I've gotten my money's worth and if I do 100Mbit and five-ish |
71 |
years from now have to upgrade as it goes past 70Mbit, I've nothing to |
72 |
complain about there either. =8^) So no, gigabit isn't something I'm |
73 |
even worried about at this point, tho obviously it'd be nice for |
74 |
flexibility if I can get it. |
75 |
|
76 |
Anyway, some of the boards now come with two or more separate PCI buses. |
77 |
I know my current Tyan dual Opteron comes with three plus the AGP, two |
78 |
slots each on two dual PCI-X, off the AMD north-side PCI-X chip (8131), and |
79 |
an old 5v compatibility PCI off the south-bridge (8111). While I'd not be |
80 |
getting anything /that/ high end for my router, presumably someone wanting |
81 |
to run four or five gig-E interfaces off the same mobo/cpu /would/ need |
82 |
something with that sort of PCI layout, and since I have it on my main |
83 |
machine mobo, it's certainly available. |
84 |
|
85 |
The thing I'm debating now, is if I choose to go full computer anyway, why |
86 |
not go lowest end amd64 I can buy, and run Gentoo on it the same as on my |
87 |
main system, in which case I can share at least /some/ packages, the ones |
88 |
without desktop specific USE flags that I want on both systems, anyway. |
89 |
In theory, I could even run a distcc client on it to help with compiling, |
90 |
altho my coming upgrade to dual dual-core Opterons (285s, most likely) |
91 |
would mean I'd not get /that/ much benefit out of it, and it'd break the |
92 |
rule of not putting stuff like gcc on a firewall purposed system. I |
93 |
figure low end bare-bones, with a smallish <100GB hard drive set hardware |
94 |
write-only mode after installation) and using an extra half-gig RAM stick I |
95 |
already have, would run ~$300-ish. |
96 |
|
97 |
So... anybody have any opinions on this? Should I go straight 32-bit or |
98 |
64-bit Gentoo? If I went 32-bit, I'd probably go with a pre-built router |
99 |
distribution instead of bothering with trying to keep up with Gentoo on |
100 |
it, altho I might change my mind on that after I get the dual Opteron 285s |
101 |
in my main system. Anybody else running such things, either Gentoo or |
102 |
other Linux or BSD? Why did you choose what you did? |
103 |
|
104 |
See, this thread /did/ come back around to amd64! <g> |
105 |
|
106 |
>> Likewise, my next mp3 player, which |
107 |
>> will be my first hard drive based unit, will be purchased with the |
108 |
>> intent of upgrading it to rockbox or a similar alternative, as well as |
109 |
>> upgrading the hard drive to a 120 gig or so model. |
110 |
> |
111 |
> While the HD isn't upgradable, the iRiver H10 line will run Rockbox. |
112 |
> IIRC, the older H300 models do as well. I've be very happy with my |
113 |
> iFP-799, although I do use a hacked firmware, which improves the range |
114 |
> of Vorbis bitrates the player supports. (The native vorbis support of |
115 |
> the iFP line was the main reason I purchased from iRiver.) IIRC, |
116 |
> Rockbox is also being ported to the iFP line, but that port is not |
117 |
> complete. Once I feel installing RockBox is safe, I'll probably upgrade |
118 |
> my iFP to it. |
119 |
|
120 |
I've been looking at the H10 line, and I /think/ some of them might |
121 |
actually be HD upgradable, now. They are running 1.8" hard drives, which |
122 |
at present top out at 60 gig, but new ones have been announced, and 120 |
123 |
gig shouldn't be far away. |
124 |
|
125 |
Do you know for sure that the current 20 gig (or maybe 30 gig) aren't hard |
126 |
drive upgradeable? |
127 |
|
128 |
Anyway, I figure it'll be 1H2007 before the 120 gig 1.8" drives are |
129 |
decently available, so I had some time to wait, for both that and rockbox |
130 |
H10 support to develop further. |
131 |
|
132 |
The other thing I was figuring was that iPods will probably be announced |
133 |
with 120 gig 1.8" drives shortly after they become available as well. |
134 |
While rockbox iPod support for that generation would lag a bit, it's a |
135 |
fair bet they'd come out with it. It's also a fair bet I could buy a |
136 |
smaller capacity one that already has rockbox support and upgrade the |
137 |
drive. However, again, we're looking at probably first half next year |
138 |
before the drives are widely available, and in any case I expect I'll be |
139 |
doing the dual Opteron 285 upgrades first, later this year, and will be |
140 |
waiting several months after that before getting serious on the mp3 player |
141 |
front, which would again take me to March or later of next year, so it |
142 |
looks like everything will come together nicely at about the same time. <g> |
143 |
|
144 |
-- |
145 |
Duncan - List replies preferred. No HTML msgs. |
146 |
"Every nonfree program has a lord, a master -- |
147 |
and if you use the program, he is your master." Richard Stallman |
148 |
|
149 |
-- |
150 |
gentoo-amd64@g.o mailing list |