Re: WEP and performance?


From: Jouni Malinen (jkmaline_at_cc.hut.fi)
Date: 2002-05-02 16:19:20 UTC



On Thu, May 02, 2002 at 08:02:42AM -0700, Brad Colbert wrote:

> Will enabling WEP reduce my overall throughput performance?

Yes, but not necessarily very much. WEP frames have 8 bytes of extra data (IV + ICV), so no matter how fast the encryption/decryption is done, there will be some overhead.

However, larger performance difference can be seen when WEP processing is the bottleneck, which is the case in at least many of the older Prism2 cards. Newer cards (at least Prism2.5-based PCI cards I have been testing) seem to have sufficiently fast WEP processing so that the difference of using WEP is not that noticeable.

I just happened to finish implementing host decryption that can be used to perform decryption on the host driver (similar operation is already done for encryption in Host AP mode). Assuming that the host computer has sufficiently fast CPU, this can be used to overcome low performance WEP processing on some Prism2 cards.

To give an example, I testing UDP performance from Prism2.5-based PCI card to Compaq WL200. Without WEP the throughput was 6.36 Mbps, with Compaq WL200 doing 40-bit WEP decryption the throughput was 2.87 Mbps and with the driver performing 40-bit WEP decryption throughput was 6.29 Mbps. In other words, Compaq WL200 has a clear bottleneck on WEP processing, but when using host decryption (and encryption), there is not much difference in performance whether WEP is used or not.

-- 
Jouni Malinen                                            PGP id EFC895FA


This archive was generated by hypermail 2.1.4.