1 |
Linux GNUbie wrote: |
2 |
|
3 |
>On Sun, 2005-04-24 at 08:32 +0200, Daniel Armyr wrote: |
4 |
> |
5 |
> |
6 |
>>I am not too familiar with VPN, but as I understand it, it is pretty much only a convenient way to tunnel all traffic through an encrypted pipe. Would a benchmark using an ssh-tunnel give sufficiently relevant results? |
7 |
>> |
8 |
>> |
9 |
> |
10 |
>I have friends that told me that the hardware specs of the Linksys |
11 |
>BEFSR41 or WRT54G cannot support the VPN client service simply because |
12 |
>of the encryption/decryption process that needs more processing power |
13 |
>and physical memory. |
14 |
> |
15 |
|
16 |
The WRT54GS has 8MB of flash, btw, and (IIRC) 32MB of RAM (I have one on |
17 |
the desk next to me, would have to check.) The extra flash makes it a |
18 |
bit roomier to hack on. I'm (slowly) getting it set up to replace a |
19 |
machine at work that we use to bridge a couple of networks. |
20 |
|
21 |
I did some googling and came up with: |
22 |
|
23 |
http://martybugs.net/wireless/openwrt/openvpn.cgi |
24 |
|
25 |
----(snip)---- |
26 |
Performance Testing |
27 |
Network Architecture |
28 |
This WRT is connecting to an 802.11b Minitar MNWAPB access point, and |
29 |
hence is restricted to 802.11b 11Mbps speeds. |
30 |
|
31 |
The throughput was measured by using wget to retrieve a 3MB file over |
32 |
the wireless link. |
33 |
|
34 |
Initial tests were performed during setup, when the WRT was physically |
35 |
located close to the Minitar access point, so the WRT was associated to |
36 |
the Minitar with a link rate of 11Mbps. The tests were repeated once the |
37 |
WRT was installed at the client site, with similar results. |
38 |
|
39 |
Throughput Without VPN |
40 |
Throughput over the wireless link between the WRT and the Minitar was |
41 |
tested at approximately 600 kbytes/sec (ie, typical for an 802.11b |
42 |
wireless link). |
43 |
|
44 |
Throughput With VPN |
45 |
Once the VPN tunnel was established, and all traffic routed through it, |
46 |
the tests were repeated. Throughput dropped to approximately 300 kbytes/sec. |
47 |
|
48 |
The major cause of this slow-down is the CPU in the WRT, as it needs to |
49 |
encrypt and decrypt all the traffic that is passing through the VPN |
50 |
tunnel. This can be observed by monitoring the CPU usage on the WRT |
51 |
while transferring large amounts of traffic through the VPN tunnel - the |
52 |
OpenVPN process consumes 99% of the CPU during this time. |
53 |
|
54 |
The slow-down caused by the VPN tunnel is acceptable in the situation |
55 |
I'm using the WRT. If this isn't the case, the throughput of the VPN |
56 |
tunnel can be increased by moving the VPN termination from the WRT onto |
57 |
a faster device (ie, a linux router) behind the WRT. |
58 |
----(snip)---- |
59 |
|
60 |
So there's at least one test, but as usual YMMV. |
61 |
|
62 |
-P |
63 |
|
64 |
-- |
65 |
gentoo-embedded@g.o mailing list |