The Transfer of the backup data is too long with Cisco WAAS

[Bouchaib EL-GHOREFY]

Hello all,

I configured a 3 WAVE 594 appliances (2 accelerators and 1 central manager) in an inline mode to compress and accelerate the backup trafic which occur every night and take an average of 10 hours. As a result the WAAS compress the trafic and I can see that on the different dashboard on its web interface, and also on the monitoring of the line. But the transfer takes the same delay as above; the time of transfer is the same. I verified all the application accelerators and all of them are enabled and running. 

The WAN line debit: 100 Mbits/s

The amount of backup trafic in the average of 500 Go

Is there a way to decrease the backup transfer?

Thank you.

[finn.poulsen]

Hi Bouchaib,

Many backup systems are limited by really bad design, because they originally were designed to put the data to tape, and this creates a bottleneck at the receiving side, because the device cannot write the received data to the backup media fast enough.

And thus you can do all the WAN optimization you want, you'll just send it faster (and a smaller amount of data), but you'll still have to wait for the device to accept more data.

I'm not saying that this is the case here, but :

• What is the WAN delay between the two sites ?
• Have you tried transferring the same (or lesser) amount of data locally at the receiving site, to verify real transfer speed ?
• Assume that you have checked for duplex errors and data encryption ?

Best regards

Finn Poulsen

[Bouchaib EL-GHOREFY]

Thank you finn.poulsen for the reply,
I didn't try transferring data locally, but I checked the duplex and I configured all the Connected ports to 1 Gig Full-duplex with Enabling Autocense in the WAVEs. The backup servers don't use tapes to record to.
I see the problem coming from the Line capacity between the two sites. 100 Mb/s=12.5 Mo/s
The total amount of data that can pass is 45 Go per hour. so It would be normal to take 10 to 12 hours. Is this right?

Thank you.

 

[finn.poulsen]

Hi,

Your calculation is basically correct, but the 12,5 MBytes/s is a theoretical (and thus unachivable) value, because it doesn't take overhead, retransmissions and ACK waiting into account.

I would use 10 Mbytes/s as a rule of thumb, giving 36 Gbytes/hour and thus for 500Gbytes giving close to 14 hours.

But I definitely think your duplex setup is wrong, if it is configured as above (I assume that you're running with the inline ports) :

If your switches/routers are hardcoded to 1G/Full duplex and you WAVE is running Autonegotiation, the will end up in half duplex, and thus you'll get a lot of errors.

On the switches you'll see "input errors" when doing "show interfaces" and on the WAVE you could issue the command "show interface inlineports <slot#>/<group#>lan or ...wan   and look for  output errors and duplex state.

Best regards

Finn Poulsen

[Bouchaib EL-GHOREFY]

Thank you Finn for the update.

Yes I am running the Inline Ports, ( and I set the AutoSence so as the WAVEs continue working if they change the Switch or the Router where it is connected). On the Devices List from the Manager, I saw Green notifications (meaning the links are negociating at their full capacity as recommended). But I will check for those errors next time.

Thank you.