Wan Optimization Support

Technical => Troubleshooting => : mgagliardi September 08, 2018, 06:35:28 PM

: TXP-X : How To Verify?
: mgagliardi September 08, 2018, 06:35:28 PM
I’m running a trial of 4.2.3 and I can see my traffic is being optimized (cool!). I have now enabled TCP-X (TCP Acceleration) and nothing appears to have changed/improved performance-wise. My link has ~40ms (20 Mbps link) latency and without acceleration I’m hitting ~2.7Mbps throughput. With acceleration enabled I’ve actually dropped a bit down to ~2.4Mbps. Is there any way to confirm that traffic is being accelerated? FWIW this is on a single subnet and the Wanos instances are in bridge mode. No routes added, the optimization rule (working) is set up for the subnet and I just have the default “accelerate everything” rule in for TCP-X.

Thanks!
: Re: TXP-X : How To Verify?
: JohnNicholas September 08, 2018, 06:47:48 PM
In the UI look for the session in the Netstat page. If TCP-X is not enabled the session will not show up.

2.7Mbps over a 20Mbps link with 40ms does not sound right. Maybe testing over a bad WiFi link? It seems very low.
Also TCP-X is designed for higher latency, not sure what the result will be at 40ms, but 2.7Mbps is way too low for a 20Mbps link.

To test the performance between the two instances:

Appliance 1:
cd /tce/www/redirect/
dd if=/dev/zero of=speedtest.file bs=1M count=100


Appliance 2:
wget http://appliance-1-ip/speedtest.file


Usually this should at least be around 100Mbps over a 20Mbps 40ms.
: Re: TXP-X : How To Verify?
: mgagliardi September 09, 2018, 02:22:45 PM
Thanks for the suggested test, I did what you'd indicated and got 2.2MB/sec...roughly 17.6Mb/sec on a 20Mb/sec link.  That seems reasonable.  I also did a few other tests...
I'm mystified here...curious about why the WAN link isn't being more fully utilized when optimization is in effect.  FWIW right now this lab is all in place on a single VMware host...no traffic is actually hitting the wire.  All VMs are over-provisioned (ex. the Wanos VMs are 4 vCPUs and 8GB RAM) and the host has plenty of unallocated resources, this isn't an issue of CPU, RAM or network contention.  Is it possible the protocol in use (DICOM) is not well-suited to optimization/acceleration?
: Re: TXP-X : How To Verify?
: ahenning September 09, 2018, 09:31:47 PM
Is Iperf3 running on the test machines or directly on Wanos?

It is worth testing with deduplication set to level 0. This would eliminate some potential RAM and Disk IOPS or Disk latency bottlenecks.

The Optimized Iperf3 results looks as expected. Also perhaps test with TCP-X enabled, but please read the doc first has in bridge mode the network routes much be carefully configured on Wanos to ensure the TCP accelerator can reach all the source and destination subnets.
: Re: TXP-X : How To Verify?
: mgagliardi September 09, 2018, 11:13:07 PM
I've got iPerf3 running on the test machines themselves, transiting the Wanos instances.  Tomorrow I will change the dedup level and see what that brings.  FWIW while I am running in bridge mode the entire setup is on a single subnet (two Wanos instances, each with two NICs, one NIC in the "local" LAN Segment with the test machine, one in the DMZ/WAN LAN Segment).  Traffic is most definitely transiting the Wanos instances as if I power either of them off the clients can't even ping each other.  I saw the warning in the docs about TCP-X and routing but I think I don't think my setup meets the conditions that were cautioned against.
: Re: TXP-X : How To Verify?
: JohnNicholas September 13, 2018, 10:28:27 AM
Yes, in bridge mode all the devices can actually be in the same subnet.

iPerf3 - optimized and accelerated I hit 103Mbps on the LAN Tx, optimized to 8.4Mbps on the WAN Tx.
That looks promising.

Any break through with the Dicom protocol?