To what extent will WPA reduce throughput? and the cause of this reduction?
in [Wireless]
|
Prev: Vista: How to save favorite WLAN connections?
Next: To what extent will WPA reduce throughput? and the cause of this reduction?
From: MingChingMin on 27 Mar 2008 00:03 1. To what extent will WPA reduce throughput? 2. Which factor dominates this reduction, computation loading or bandwidth overhead? With my wireless LAN, I do not feel any degradation to throughput when running WEP64 or WEP128. When I run WPAPSK, the throughput is reduced by 40% on transferring large files. I would be grateful to know if this is a common phenomenon, or it just happens to my site. Then I make attempts to find out what has caused this reduction in throughput. I have gone through the following experiments, but have difficulties in interpreting the results. In my experiments, an 8 MB file is transferred from Computer A to computer B. Computer A is connected by UTP to a port (not the WAN port) of a wireless router. Computer B is equipped with a wireless adapter. Computer B reads the 8 MB file and sends it to Computer A. Computer A receives the file, reports the speed but does not write the file to harddisk. This is done to minimize the impact of disk performance to out experiments. At the mean time, I monitor the networking performance and CPU utilization at "Windows Task Manager". Both computers runs WinXP with hyperthreading CPUs. The following shows my experiment log. (1) Wired connection, B sends to A: Sender: networking: peaks at 62.5% of 100Mbps CPU: peaks at 9% receiver: networking: peaks at 45% of 100Mbps CPU: peaks at 12% Throughput: 94 Mbps It seems that the reading of the file form disk is not a bottleneck in my experiment. (2) Wireless connection without wireless security protocol Sender: networking: peaks at 50% of 54 Mbps CPU: peaks at 16% Receiver: networking: peaks at 28% of 100Mbps CPU: peaks at 11% Throughput: 26Mbps It seems that the wireless adapter is running close to its theoretical maximum. (3) Wireless connection with WEP64: Sender networking: peaks at 50% of 54 Mbps CPU utilization: peaks at 13% Receiver: networking: peaks at 7% of 100 Mbps CPU : peaks at 9% Throughput: 26Mbps It seems that no degradation of throughput is caused by WEP64. (4) Wireless connection with WEP128: Sender: networking: peaks at 50% of 54 Mbps CPU: peaks at 40 Receiver: networking: peaks at 27% of 100Mbps CPU: peaks at 8% Throughput: 26Mbps No degradation to throughput by WEP128 either. (5) Wireless connection with WPAPSK: Sender: networking: peaks at 37.5% of 54 Mbps CPU: peaks at 21% Receiver: networking : peaks at 16% CPU: peaks at 14% Throughput: 16Mbps My difficulties lie here! networking performance is not at the expected 50%. Could it be because the computation work for encryption is not done fast enough to allow data be pumped out at ceiling rate? But CPU utilization is far less than 50%, the ceiling for one hyperthread. Could it be because the encryption is done in the adapter? But it is only an inexpensive USB wireless adapter, even without WPA2 capability. Could it be that actually the wireless raw data throughput has reached ceiling, but Windows Task manager only reports the throughput of data stream before encryption? In that case, the encrypted data is 33% larger than the plantext. But WPA uses RC4 stream cipher so it is unlikely that the ciphertext is 33% larger in size than the plaintext. What has really happened? Please give me some advice. P.S. About my understanding that Windows Task Manager sees a one way traffic at ceiling rate as 50%, I got it from the following experiments. Please tell me if my understanding is wrong. (6) Wired connection, both computers send and receive at the same time: No significant difference between the two streams is observed. Both computers: networking: peaks at 100% (7) Wireless connection, no encryption: both computers attempts to send and receive at the same time: It is observed that the two ways are asymmetric. The traffic from the wireless adapter to the wireless router has significantly higher priority, although the other stream also makes progress. Throughput from B to A is 21Mbps, and that from A to B is 6 Mbps. After the B to A sending has finished its 8 MB file, the A to B traffic resumes its performance at 26 Mbps. Computer A network utilization: peaks 27% of 100Mbps Computer B network utilization: peaks at 50% of 54Mbps Please give me some advice.
From: Jack (MVP-Networking). on 27 Mar 2008 13:50 Hi WEP indeed reduces the Bandwidth. WPA should not, unless you have a badly designed hardware. Try to optimize the TCP/IP stack on all your computers. Optimizing the TCP/IP Stack - http://www.ezlan.net/Internet_Speed.html Jack (MVP-Networking). <MingChingMin(a)yahoo.com.hk> wrote in message news:362a66aa-ff0e-4cfc-95ea-10a868fb4881(a)e10g2000prf.googlegroups.com... > 1. To what extent will WPA reduce throughput? > 2. Which factor dominates this reduction, computation loading or > bandwidth overhead? > > With my wireless LAN, I do not feel any degradation to throughput when > running WEP64 or WEP128. When I run WPAPSK, the throughput is reduced > by 40% on transferring large files. I would be grateful to know if > this is a common phenomenon, or it just happens to my site. > > Then I make attempts to find out what has caused this reduction in > throughput. I have gone through the following experiments, but have > difficulties in interpreting the results. > > In my experiments, an 8 MB file is transferred from Computer A to > computer B. Computer A is connected by UTP to a port (not the WAN > port) of a wireless router. Computer B is equipped with a wireless > adapter. Computer B reads the 8 MB file and sends it to Computer A. > Computer A receives the file, reports the speed but does not write the > file to harddisk. This is done to minimize the impact of disk > performance to out experiments. At the mean time, I monitor the > networking performance and CPU utilization at "Windows Task Manager". > Both computers runs WinXP with hyperthreading CPUs. The following > shows my experiment log. > > (1) Wired connection, B sends to A: > Sender: > networking: peaks at 62.5% of 100Mbps > CPU: peaks at 9% > receiver: > networking: peaks at 45% of 100Mbps > CPU: peaks at 12% > Throughput: 94 Mbps > It seems that the reading of the file form disk is not a bottleneck in > my experiment. > > (2) Wireless connection without wireless security protocol > Sender: > networking: peaks at 50% of 54 Mbps > CPU: peaks at 16% > Receiver: > networking: peaks at 28% of 100Mbps > CPU: peaks at 11% > Throughput: 26Mbps > It seems that the wireless adapter is running close to its theoretical > maximum. > > > (3) Wireless connection with WEP64: > Sender > networking: peaks at 50% of 54 Mbps > CPU utilization: peaks at 13% > Receiver: > networking: peaks at 7% of 100 Mbps > CPU : peaks at 9% > Throughput: 26Mbps > It seems that no degradation of throughput is caused by WEP64. > > (4) Wireless connection with WEP128: > Sender: > networking: peaks at 50% of 54 Mbps > CPU: peaks at 40 > Receiver: > networking: peaks at 27% of 100Mbps > CPU: peaks at 8% > Throughput: 26Mbps > No degradation to throughput by WEP128 either. > > > (5) Wireless connection with WPAPSK: > Sender: > networking: peaks at 37.5% of 54 Mbps > CPU: peaks at 21% > Receiver: > networking : peaks at 16% > CPU: peaks at 14% > Throughput: 16Mbps > My difficulties lie here! networking performance is not at the > expected 50%. Could it be because the computation work for encryption > is not done fast enough to allow data be pumped out at ceiling rate? > But CPU utilization is far less than 50%, the ceiling for one > hyperthread. Could it be because the encryption is done in the > adapter? But it is only an inexpensive USB wireless adapter, even > without WPA2 capability. Could it be that actually the wireless raw > data throughput has reached ceiling, but Windows Task manager only > reports the throughput of data stream before encryption? In that case, > the encrypted data is 33% larger than the plantext. But WPA uses RC4 > stream cipher so it is unlikely that the ciphertext is 33% larger in > size than the plaintext. What has really happened? Please give me some > advice. > > P.S. > About my understanding that Windows Task Manager sees a one way > traffic at ceiling rate as 50%, I got it from the following > experiments. Please tell me if my understanding is wrong. > > (6) Wired connection, both computers send and receive at the same > time: > No significant difference between the two streams is observed. > Both computers: > networking: peaks at 100% > > (7) Wireless connection, no encryption: both computers attempts to > send and receive at the same time: > It is observed that the two ways are asymmetric. The traffic from the > wireless adapter to the wireless router has significantly higher > priority, although the other stream also makes progress. Throughput > from B to A is 21Mbps, and that from A to B is 6 Mbps. After the B to > A sending has finished its 8 MB file, the A to B traffic resumes its > performance at 26 Mbps. > Computer A network utilization: peaks 27% of 100Mbps > Computer B network utilization: peaks at 50% of 54Mbps > > > Please give me some advice.
From: MingChingMin on 27 Mar 2008 21:55 On Mar 27, 5:22 pm, Barb Bowman <b...(a)nospam.com> wrote: > what wireless router are you using? > > On Wed, 26 Mar 2008 21:03:59 -0700 (PDT), MingChing...(a)yahoo.com.hk > wrote: > > > > > > >1. To what extent will WPA reduce throughput? > >2. Which factor dominates this reduction, computation loading or > >bandwidth overhead? > > >With my wireless LAN, I do not feel any degradation to throughput when > >running WEP64 or WEP128. When I run WPAPSK, the throughput is reduced > >by 40% on transferring large files. I would be grateful to know if > >this is a common phenomenon, or it just happens to my site. > > >Then I make attempts to find out what has caused this reduction in > >throughput. I have gone through the following experiments, but have > >difficulties in interpreting the results. > > >In my experiments, an 8 MB file is transferred from Computer A to > >computer B. Computer A is connected by UTP to a port (not the WAN > >port) of a wireless router. Computer B is equipped with a wireless > >adapter. Computer B reads the 8 MB file and sends it to Computer A. > >Computer A receives the file, reports the speed but does not write the > >file to harddisk. This is done to minimize the impact of disk > >performance to out experiments. At the mean time, I monitor the > >networking performance and CPU utilization at "Windows Task Manager". > >Both computers runs WinXP with hyperthreading CPUs. The following > >shows my experiment log. > > >(1) Wired connection, B sends to A: > >Sender: > >networking: peaks at 62.5% of 100Mbps > >CPU: peaks at 9% > >receiver: > >networking: peaks at 45% of 100Mbps > >CPU: peaks at 12% > >Throughput: 94 Mbps > >It seems that the reading of the file form disk is not a bottleneck in > >my experiment. > > >(2) Wireless connection without wireless security protocol > >Sender: > >networking: peaks at 50% of 54 Mbps > >CPU: peaks at 16% > >Receiver: > >networking: peaks at 28% of 100Mbps > >CPU: peaks at 11% > >Throughput: 26Mbps > >It seems that the wireless adapter is running close to its theoretical > >maximum. > > >(3) Wireless connection with WEP64: > >Sender > >networking: peaks at 50% of 54 Mbps > >CPU utilization: peaks at 13% > >Receiver: > >networking: peaks at 7% of 100 Mbps > >CPU : peaks at 9% > >Throughput: 26Mbps > >It seems that no degradation of throughput is caused by WEP64. > > >(4) Wireless connection with WEP128: > >Sender: > >networking: peaks at 50% of 54 Mbps > >CPU: peaks at 40 > >Receiver: > >networking: peaks at 27% of 100Mbps > >CPU: peaks at 8% > >Throughput: 26Mbps > >No degradation to throughput by WEP128 either. > > >(5) Wireless connection with WPAPSK: > >Sender: > >networking: peaks at 37.5% of 54 Mbps > >CPU: peaks at 21% > >Receiver: > >networking : peaks at 16% > >CPU: peaks at 14% > >Throughput: 16Mbps > >My difficulties lie here! networking performance is not at the > >expected 50%. Could it be because the computation work for encryption > >is not done fast enough to allow data be pumped out at ceiling rate? > >But CPU utilization is far less than 50%, the ceiling for one > >hyperthread. Could it be because the encryption is done in the > >adapter? But it is only an inexpensive USB wireless adapter, even > >without WPA2 capability. Could it be that actually the wireless raw > >data throughput has reached ceiling, but Windows Task manager only > >reports the throughput of data stream before encryption? In that case, > >the encrypted data is 33% larger than the plantext. But WPA uses RC4 > >stream cipher so it is unlikely that the ciphertext is 33% larger in > >size than the plaintext. What has really happened? Please give me some > >advice. > > >P.S. > >About my understanding that Windows Task Manager sees a one way > >traffic at ceiling rate as 50%, I got it from the following > >experiments. Please tell me if my understanding is wrong. > > >(6) Wired connection, both computers send and receive at the same > >time: > >No significant difference between the two streams is observed. > >Both computers: > >networking: peaks at 100% > > >(7) Wireless connection, no encryption: both computers attempts to > >send and receive at the same time: > >It is observed that the two ways are asymmetric. The traffic from the > >wireless adapter to the wireless router has significantly higher > >priority, although the other stream also makes progress. Throughput > >from B to A is 21Mbps, and that from A to B is 6 Mbps. After the B to > >A sending has finished its 8 MB file, the A to B traffic resumes its > >performance at 26 Mbps. > >Computer A network utilization: peaks 27% of 100Mbps > >Computer B network utilization: peaks at 50% of 54Mbps > > >Please give me some advice. > > -- > > Barb Bowman > MS-MVPhttp://www.microsoft.com/windowsxp/expertzone/meetexperts/bowman.mspxhttp://blogs.digitalmediaphile.com/barb/- Hide quoted text - > > - Show quoted text - The wireless router used: Manufacturer: Level One Model: WBR-3406TX It supports 802.11b/g The USB wireless adapter used: Manufacturer: Level One Model: WNC-0301USB It supports 802.11b/g In these experiments, the wireless adapter is placed within 10 cm from the wireless router. Thank you.
From: MingChingMin on 27 Mar 2008 22:17
On Mar 27, 12:03 pm, MingChing...(a)yahoo.com.hk wrote: > 1. To what extent will WPA reduce throughput? > 2. Which factor dominates this reduction, computation loading or > bandwidth overhead? > > With my wireless LAN, I do not feel any degradation to throughput when > running WEP64 or WEP128. When I run WPAPSK, the throughput is reduced > by 40% on transferring large files. I would be grateful to know if > this is a common phenomenon, or it just happens to my site. > > Then I make attempts to find out what has caused this reduction in > throughput. I have gone through the following experiments, but have > difficulties in interpreting the results. > > In my experiments, an 8 MB file is transferred from Computer A to > computer B. Computer A is connected by UTP to a port (not the WAN > port) of a wireless router. Computer B is equipped with a wireless > adapter. Computer B reads the 8 MB file and sends it to Computer A. > Computer A receives the file, reports the speed but does not write the > file to harddisk. This is done to minimize the impact of disk > performance to out experiments. At the mean time, I monitor the > networking performance and CPU utilization at "Windows Task Manager". > Both computers runs WinXP with hyperthreading CPUs. The following > shows my experiment log. > > (1) Wired connection, B sends to A: > Sender: > networking: peaks at 62.5% of 100Mbps > CPU: peaks at 9% > receiver: > networking: peaks at 45% of 100Mbps > CPU: peaks at 12% > Throughput: 94 Mbps > It seems that the reading of the file form disk is not a bottleneck in > my experiment. > > (2) Wireless connection without wireless security protocol > Sender: > networking: peaks at 50% of 54 Mbps > CPU: peaks at 16% > Receiver: > networking: peaks at 28% of 100Mbps > CPU: peaks at 11% > Throughput: 26Mbps > It seems that the wireless adapter is running close to its theoretical > maximum. > > (3) Wireless connection with WEP64: > Sender > networking: peaks at 50% of 54 Mbps > CPU utilization: peaks at 13% > Receiver: > networking: peaks at 7% of 100 Mbps > CPU : peaks at 9% > Throughput: 26Mbps > It seems that no degradation of throughput is caused by WEP64. > > (4) Wireless connection with WEP128: > Sender: > networking: peaks at 50% of 54 Mbps > CPU: peaks at 40 > Receiver: > networking: peaks at 27% of 100Mbps > CPU: peaks at 8% > Throughput: 26Mbps > No degradation to throughput by WEP128 either. > > (5) Wireless connection with WPAPSK: > Sender: > networking: peaks at 37.5% of 54 Mbps > CPU: peaks at 21% > Receiver: > networking : peaks at 16% > CPU: peaks at 14% > Throughput: 16Mbps > My difficulties lie here! networking performance is not at the > expected 50%. Could it be because the computation work for encryption > is not done fast enough to allow data be pumped out at ceiling rate? > But CPU utilization is far less than 50%, the ceiling for one > hyperthread. Could it be because the encryption is done in the > adapter? But it is only an inexpensive USB wireless adapter, even > without WPA2 capability. Could it be that actually the wireless raw > data throughput has reached ceiling, but Windows Task manager only > reports the throughput of data stream before encryption? In that case, > the encrypted data is 33% larger than the plantext. But WPA uses RC4 > stream cipher so it is unlikely that the ciphertext is 33% larger in > size than the plaintext. What has really happened? Please give me some > advice. > > P.S. > About my understanding that Windows Task Manager sees a one way > traffic at ceiling rate as 50%, I got it from the following > experiments. Please tell me if my understanding is wrong. > > (6) Wired connection, both computers send and receive at the same > time: > No significant difference between the two streams is observed. > Both computers: > networking: peaks at 100% > > (7) Wireless connection, no encryption: both computers attempts to > send and receive at the same time: > It is observed that the two ways are asymmetric. The traffic from the > wireless adapter to the wireless router has significantly higher > priority, although the other stream also makes progress. Throughput > from B to A is 21Mbps, and that from A to B is 6 Mbps. After the B to > A sending has finished its 8 MB file, the A to B traffic resumes its > performance at 26 Mbps. > Computer A network utilization: peaks 27% of 100Mbps > Computer B network utilization: peaks at 50% of 54Mbps > > Please give me some advice. The wireless router used in this experiment: Manufacturer: Level One Model: WBR-3406TX Supports 802.11b/g The USB wireless adapter used in this experiment: Manufacturer: Level One Model: WNC-0301USB Supports 802.11b/g In these experiments, the wireless adapter is placed within 10 cm from the wireless router. Thank you. |