Solved: Does CIFS optimization work without WAFS

snakayama · ‎05-14-2007

Hi everyone,

Now I am testing WAE in my lab with the following environment.

Windows 2003 Server

|

Cisco 2821 with NME-WAE-502-K9 as WAE Core

|

NIST Net WAN Emulation Software

|

Cisco 2821 with NME-WAE-502-K9 as WAE Edge

|

PC

Cisco 2821:

12.4(9)T3

enabled CEF, WCCP v2 with TCP promiscuous mode service (WCCP services 61 and 62)

NME-WAE-502-K9:

4.0.3

enabled WCCP v2

NIST Net parameters:

Bandwidth: 1.5M

Latency: 300ms

Packet loss: 0.5%

Test Tools/Methods on PC:

Cisco WAFS Benchmark Tool for Microsoft Office Application downloaded from CCO

for CIFS/SMB application

TCP Replay utility downloaded from CCO for HTTP application

FTP for FTP application

In this environment, I can get good result for HTTP (by using TCP Reply utility) and FTP, that is,

speed for downloading file by HTTP and FTP is improved over WAN (NIST Net).

However I can not see the effectiveness of WAE for CIFS/SMB traffic by using WAFS Benchmark Tool.

I do not configure WAFS (Wide Area File Services), because I can not configure it due to lack of

DRAM on WAE Core.

2GB DRAM is needed to get WAE Core to be worked as WAFS core cluster but both WAE have

just 1GB DRAM.

So I am testing how WAE efficiently work for WAFS Benchmark Tool without WAFS.

However the result of testing is not indicating WAE efficiency for WAFS Benchmark Tool.

That is, Speed is not different, in other words Speed does not change so much in the following

three cases.

Native WAN:

Remove WCCP v2 configuration from Cisco 2821

Then PC accesses Windows 2003 server by using WAFS Benchmark Tool

Cache Miss:

Enable WCCP v2 with TCP promiscuous mode service (WCCP services 61 and 62) on Cisco

2821 and WAE and restart WAE to clear any cache information

Then PC accesses Windows 2003 server (it is a first time) by using WAFS Benchmark Tool

Cache Hit:

In the case of PC has been accesses Windows server in the past

PC accesses Windows 2003 server (it is a second time) by using WAFS Benchmark Tool

And unfortunately, when "wccp cifs-cache" command is configured or not on both WAE,

the result is the same, that is, Speed does not change so much.

I am confusing about optimization for CIFS/SMB traffic.

According to "Default Application Policies" in the Cisco Wide Area Application

Services Configuration Guide,

http://www.cisco.com/en/US/customer/docs/app_ntwk_services/waas/waas/v403/configuration/guide/apx_apps.html

WAAS, by default, handles CIFS/SMB traffic with LZ, TFO and DRE.

Classifier: CIFS-non-wafs

WAAS Action: LZ+TFO+DRE

Destination Ports: 139, 445

So I was thinking WAE optimize CIFS/SMB traffic even if no WAFS configured...

Do I need to configure WAFS in other words WAFS must be configured to achieve

the effect of WAE for CIFS/SMB traffic ?

or

Do I need to configure additional command to get WAE work fine for CIFS/SMB traffic

without WAFS ?

or

Is my idea wrong ? that is, WAFS Benchmark Tool is not appropriate tool in this

environment (no WAFS) ?

Your assistance would be appreciated.

Best regards,

tblancha · ‎05-17-2007

Configs are good and with F's you have an optimized session. So, you should be getting good optimization. Do a clear dre cache and answer yes. Then reboot the PC and restart the connection to the server. Instead of using the benchmark tool, try to cut & paste from the server. After the first cut & paste, issue a show dre statistic dre and make sure the DRE is not zero percentage. Is this really a server or is it just a PC that is sharing a drive?

View solution in original post

tblancha · ‎05-15-2007

In this situation, you will only get TFO+DRE+LZ. This is because the 302 cannot be a core nor an edge device. Without a core/edge concept you cannot manually configure the connectivity directive which will allow the CIFS protocol to be optimized at L4.

tblancha · ‎05-15-2007

Actually, the 502 can be an edge but not a core. The result is same as previously posted. If you do a show connection tfo summ you will see the end devices or you will see something on port 4050. That is good and should have all F's in the last column. Do you not get a vast improvement when having WAAS do optimization?

snakayama · ‎05-16-2007

Thank you very much for your reply and assistance.

First of all, why I would like to confirm whether CIFS optimization works or not in case of WAFS not configured is that,

I would like to know whether CIFS traffic can be optimized/accelerated by just using WAAS TFO, DRE and LZ, not using WAFS.

The following is the result of our testing how NME-WAE handle CIFS traffic.

Test environment is same as previously posted and I still used NME-WAE on both core and edge.

Test method executed were as follows:

Test 1:

On the Native WAN environment which means WCCP v2 configuration from both Cisco 2821 were removed that is CIFS traffic was not redirected to NME-WAE

PC accessed Windows 2003 server by using WAFS Benchmark Tool

Benchmark Tool open, save and close 2MB power point file on the server

Test 2:

Cache Miss: (this expression might be wrong though ...)

Enable WCCP v2 with TCP promiscuous mode service (WCCP services 61 and 62) on both Cisco 2821 and NME-WAE and clear cache information by "clear cache dre"

PC accessed Windows 2003 server (it was a first time) by using WAFS Benchmark Tool

Benchmark Tool open, save and close 2MB power point file on the server

Test 3:

Cache Hit: (this expression also might be wrong though ...)

PC accessed Windows 2003 server again (it was a second time) by using WAFS Benchmark Tool

Benchmark Tool open, save and close 2MB power point file on the server

Measured speed (spent time to open, save and close) by Benchmark Tool are as follows.

Test 1 (Native WAN):

File Open(s) Save(s) Close(s)

2M.ppt 22.583 23.503 0.06

(1918.5KB)

Test 2 (Cache Miss):

File Open(s) Save(s) Close (s)

2M.ppt 31.487 22.642 0.541

(1918.5KB)

Test 3 (Cache Hit):

File Open(s) Save(s) Close(s)

2M.ppt 24.225 26.077 0.04

(1918.5KB)

I can not see the effectiveness of NME-WAE for CIFS/SMB traffic by using WAFS Benchmark Tool.

Because speed does not change so much between Test 1, 2 and 3.

And the output of "sh statistics dre" and "sh tfo conn sum" on edge NME-WAE while Test 2 and 3 were being executed. Same information also got on the core NME-WAE however that information is similar as edge, so I do not post it here.

Please note that I post the output of "sh statistics dre" and "sh tfo conn sum" on edge NME-WAE while Test 2 and 3 were being executed as next post message because I can not it to attach due to Server Error.

Following are my question I would like to know.

[Question 1]

You said "Actually, the 502 can be an edge but not a core".

Does it means that I can not use NME-WAE-502-K9 as WAE core even if I do not configure WAFS ?

[Question 2]

According to my testing, it indicates NME-WAE handle CIFS traffic with Full optimization, however measured speed does not so much.

So I guess CIFS traffic can not be optimized/accelerated so much without WAFS even if NME-WAE apply Full optimization (TFO+DRE+LZ) to the CIFS traffic.

Is my understanding is correct ?

[Question 3]

Or if I use Appliance WAE (WAE-512 or WAE-612) on the core, does CIFS traffic get optimized/accelerated from this ?

I understand I have to same test in case of WAE-512 or 612 as core to confirm it, however unfortunately I can not prepare WAE-512 or 612 now.

Your any assistance would be greatly appreciated.

Best regards,

snakayama · ‎05-16-2007

=== Before Test 2 is executed ===

WAE-SiteA#sh tfo conn sum

WAE-SiteA#sh statistics dre

Cache:

Status: Usable, Oldest Data (age): 0s

Total usable disk size: 47527 MB, Used: 0.00%

Hash table RAM size: 189 MB, Used: 0.00%

=== While Test 1 is being executed ===

WAE-SiteA#sh tfo conn sum

Optimized Connection List

Policy summary order: Our's, Peer's, Negotiated, Applied

F: Full optimization, D: DRE only, L: LZ Compression, T: TCP Optimization

Local-IP:Port Remote-IP:Port ConId PeerId Policy

192.168.2.2:1031 192.168.1.1:445 2 00:16:9d:38:8a:5d F,F,F,F

WAE-SiteA#sh statiWAE-SiteA#sh statistics dre

Cache:

Status: Usable, Oldest Data (age): 4m14s

Total usable disk size: 47527 MB, Used: 0.00%

Hash table RAM size: 189 MB, Used: 0.00%

Connections: Total (cumulative): 1 Active: 1

Encode:

Overall: msg: 426, in: 2120 KB, out: 1558 KB, ratio: 26.47%

DRE: msg: 426, in: 2120 KB, out: 1843 KB, ratio: 13.04%

DRE Bypass: msg: 0, in: 0 B

LZ: msg: 374, in: 749 KB, out: 464 KB, ratio: 38.00%

LZ Bypass: msg: 52, in: 1094 KB

Avg latency: 0.808 ms

Message size distribution:

0-1K=69% 1K-5K=8% 5K-15K=7% 15K-25K=6% 25K-40K=5% >40K=1%

Decode:

Overall: msg: 397, in: 101158 B, out: 2196 KB, ratio: 95.50%

DRE: msg: 396, in: 162 KB, out: 2195 KB, ratio: 92.61%

DRE Bypass: msg: 1, in: 89 B

LZ: msg: 338, in: 67992 B, out: 130 KB, ratio: 48.95%

LZ Bypass: msg: 59, in: 33166 B

Avg latency: 0.045 ms

Message size distribution:

0-1K=65% 1K-5K=9% 5K-15K=8% 15K-25K=4% 25K-40K=11% >40K=0%

=== While Test 3 is being executed ===

WAE-SiteA#sh tfo conn sum

Optimized Connection List

Policy summary order: Our's, Peer's, Negotiated, Applied

F: Full optimization, D: DRE only, L: LZ Compression, T: TCP Optimization

Local-IP:Port Remote-IP:Port ConId PeerId Policy

192.168.2.2:1031 192.168.1.1:445 2 00:16:9d:38:8a:5d F,F,F,F

WAE-SiteA#sh statiWAE-SiteA#sh statistics dre

Cache:

Status: Usable, Oldest Data (age): 12m5s

Total usable disk size: 47527 MB, Used: 0.01%

Hash table RAM size: 189 MB, Used: 0.00%

Connections: Total (cumulative): 1 Active: 1

Encode:

Overall: msg: 862, in: 4223 KB, out: 1616 KB, ratio: 61.72%

DRE: msg: 862, in: 4223 KB, out: 1948 KB, ratio: 53.88%

DRE Bypass: msg: 0, in: 0 B

LZ: msg: 726, in: 822 KB, out: 491 KB, ratio: 40.27%

LZ Bypass: msg: 136, in: 1125 KB

Avg latency: 0.584 ms

Message size distribution:

0-1K=70% 1K-5K=8% 5K-15K=7% 15K-25K=6% 25K-40K=6% >40K=1%

Decode:

Overall: msg: 803, in: 167 KB, out: 4392 KB, ratio: 96.18%

DRE: msg: 802, in: 269 KB, out: 4392 KB, ratio: 93.86%

DRE Bypass: msg: 1, in: 89 B

LZ: msg: 674, in: 100 KB, out: 203 KB, ratio: 50.29%

LZ Bypass: msg: 129, in: 68670 B

Avg latency: 0.024 ms

Message size distribution:

0-1K=66% 1K-5K=9% 5K-15K=7% 15K-25K=4% 25K-40K=11% >40K=0%

tblancha · ‎05-16-2007

Full optimization refers to TFO+DRE+LZ and is seen in the show tfo conn summ as all F's in last column. These optimization occur at layer3. CIFS optimization is a higher-layer protocol aware optimization IN ADDITION to the layer 3 optimizations. In your case, you will get full optimization only and your speeds should be substantially better than what you are showing. I don't think you have this set up correctly - please post config and/or open TAC case to have someone review the configurations.

snakayama · ‎05-16-2007

Thank you very much again for your assistance.

First, I got both NME-WAE backed to factory default configuration by executing "restore factory-default" command to initialize them, then added necessary configuration into them.

And I did Benchmark Tool to measure speed, however measured speed did not change so much as well as previously posted. At that time both NME-WAE handled CIFS traffic with full optimization as below.

Local-IP:Port Remote-IP:Port ConId PeerId Policy

192.168.2.2:1031 192.168.1.1:445 2 00:16:9d:38:8a:5d F,F,F,F

Attached are the configuration of both Core and Edge NME-WAE.

"Core_Config.txt" is the output of "sh ver" and "sh run" on Cisco 2821 and NME-WAE at Core

Edge_Config.txt is the output of "sh ver" and "sh run" on Cisco 2821 and NME-WAE at Edge

Again, thank you very much for your assistance.

Best regards,

tblancha · ‎05-17-2007

Configs are good and with F's you have an optimized session. So, you should be getting good optimization. Do a clear dre cache and answer yes. Then reboot the PC and restart the connection to the server. Instead of using the benchmark tool, try to cut & paste from the server. After the first cut & paste, issue a show dre statistic dre and make sure the DRE is not zero percentage. Is this really a server or is it just a PC that is sharing a drive?

snakayama · ‎05-17-2007

Thank you very much for your continuous assistance.

I tested how many time is required to copy a 2MB power point file from Server to PC by using DOS "copy" command on PC instead of Benchmark Tool.

The result were as follows.

##### Native WAN: without WAE #####

Begin: 10:42:35.87

End: 10:43:21.63

Required time: 45 sec

##### Cache Miss: with WAE and WAE did not any cache information #####

Begin: 11:11:35.21

End: 11:12:10.20

Required time: 35 sec

Output of "sh statistics dre" while copy was being executed

Cache:

Status: Usable, Oldest Data (age): 21h

Total usable disk size: 47527 MB, Used: 0.02%

Hash table RAM size: 189 MB, Used: 0.00%

Connections: Total (cumulative): 1 Active: 1

Encode:

Overall: msg: 68, in: 5371 B, out: 1816 B, ratio: 66.19%

DRE: msg: 68, in: 5371 B, out: 5985 B, ratio: 0.00%

DRE Bypass: msg: 0, in: 0 B

LZ: msg: 68, in: 5985 B, out: 1816 B, ratio: 69.66%

LZ Bypass: msg: 0, in: 0 B

Avg latency: 0.015 ms

Message size distribution:

0-1K=100% 1K-5K=0% 5K-15K=0% 15K-25K=0% 25K-40K=0% >40K=0%

Decode:

Overall: msg: 121, in: 43345 B, out: 1445 KB, ratio: 97.07%

DRE: msg: 120, in: 62962 B, out: 1445 KB, ratio: 95.75%

DRE Bypass: msg: 1, in: 89 B

LZ: msg: 72, in: 23046 B, out: 42839 B, ratio: 46.20%

LZ Bypass: msg: 49, in: 20299 B

Avg latency: 0.843 ms

Message size distribution:

0-1K=25% 1K-5K=25% 5K-15K=14% 15K-25K=10% 25K-40K=23% >40K=0%

#### Cache Hit: with WAE and WAE had cache information from previous CIFS traffic #####

Begin: 11:13:34.87

End: 11:14:10.23

Required time: 35 sec

Output of "sh statistics dre" while copy was being executed

Cache:

Status: Usable, Oldest Data (age): 21h

Total usable disk size: 47527 MB, Used: 0.02%

Hash table RAM size: 189 MB, Used: 0.00%

Connections: Total (cumulative): 2 Active: 0

Encode:

Overall: msg: 164, in: 12086 B, out: 3947 B, ratio: 67.34%

DRE: msg: 164, in: 12086 B, out: 13421 B, ratio: 0.00%

DRE Bypass: msg: 0, in: 0 B

LZ: msg: 164, in: 13421 B, out: 3947 B, ratio: 70.59%

LZ Bypass: msg: 0, in: 0 B

Avg latency: 0.006 ms

Message size distribution:

0-1K=100% 1K-5K=0% 5K-15K=0% 15K-25K=0% 25K-40K=0% >40K=0%

Decode:

Overall: msg: 307, in: 111 KB, out: 3908 KB, ratio: 97.14%

DRE: msg: 306, in: 155 KB, out: 3908 KB, ratio: 96.03%

DRE Bypass: msg: 1, in: 89 B

LZ: msg: 170, in: 61275 B, out: 103 KB, ratio: 42.04%

LZ Bypass: msg: 137, in: 53240 B

Avg latency: 0.371 ms

Message size distribution:

0-1K=18% 1K-5K=26% 5K-15K=18% 15K-25K=12% 25K-40K=22% >40K=1%

It indicates that measured time was improved in case of using WAE than not using WAE and DRE percentage also increased between them.

I think Benchmark Tool might be intended in case of using WAFS ???

Thank you very much for your assistance.

Best Regards,