Wan Optimization Support

Technical => Features => : aels May 10, 2017, 06:46:17 AM

: Peer Detect
: aels May 10, 2017, 06:46:17 AM
Hi A Henning

I want to ask about peer detect, based on this statement "Peers detect each other with TCP Option 76. To trigger peers to come online, start a new TCP session."

Are this mean new TCP session is "New Connection State?". Why not use ping/icmp to trigger peers,
I has testing in lab, when I open network sharing windows and status come UP but when status back idle/down, I reload few times sharing folder but status not come UP, I must close window and open link share \\PC-02 again to get status UP.
This mean TCP connection state for 'related-connection' cannot use to trigger peer?

I Think using ping (icmp) to trigger from another wanos the best way to make sure peer still UP so wanos not passthrough the traffic or maybe there is something option in wanos to auto send TCP option 76 from one wanos to another peer with interval where users can config it.

: Re: Peer Detect
: JohnNicholas May 10, 2017, 08:49:44 AM
In my experience this is the same behaviour as the riverbed steelhead RiOS

(config) # tcp connection send pass-reset source-addr source-port 45344 dest-addr dest-port 445

Silver Peak is the same as well. When testing in the lab the TCP sessions needs to be resetted when enabling Boost.
: Re: Peer Detect
: ahenning May 10, 2017, 11:14:26 AM
This is new in v4 since we now auto detect which sessions to optimize and which to leave alone. If the TCP session is seen on both appliances, then it is eligible to be optimized. If a session is only seen on one appliance but not the other, then it is bypassed. Hence, MultiSite has been removed and configuring Traffic Policies are now optional. This is a big step forward considering that these are often miss-configured in v3.

Wanos has the same TCP reset capability as Silver Peak and Riverbed Steelhead:

Configure -> Reset -> Reset TCP-X

If TCP-X is not enabled: Enable TCP-X, then disable TCP-X again

As like the other vendors, this is only expected to be used in a lab/test environment. Under normal production environments it would not be normal to reset TCP sessions in order to force a TCP reconnect to start optimization for a particular application like CIFS/SMB.

As for the ping idea, it is not really workable. The remote peer could be removed from the inline position but still respond to ICMP and even wanos control traffic like PLR and RTT measurements may still flow between the sites. This would cause the peers to remain "Active" causing the device that is inline to keep optimization alive. This would cause a blackout for optimized traffic. Better to be safe and only optimize while valid traffic is seen. When one of the peers are removed or moved where traffic flow is not correct, the peers should go into Idle state to avoid traffic blackout.

Test in a production network. If you still see peers Idle occasionally but want to force them to "Active" increase the peer timeout value under settings.
: Re: Peer Detect
: aels May 12, 2017, 08:53:30 AM
Hi A Henning

Thanks for your explained, one more question, are Router mode has removed from V4 ?
I see only Bridge and Tunnel mode, how I can mix for 2 device, where one device have only 1 NIC and other have 2 NIC.

And, can I modify default policymap port, I want to optimize RDP (3389), does this work if I add rule #99 to allow RDP trafict to optimize?

: Re: Peer Detect
: JohnNicholas May 12, 2017, 09:05:09 AM
Router mode is not supported in Express.

Read the section in the admin guide about RDP: