WAAS Design and Implemention questions

Unanswered Question
Feb 29th, 2008

Q1: What is the bandwidth range ( lowest and highest )in which WAAS is effective?


Q2: What is the latency value below which WAAS WAN optimization is NOT effective? 80ms ?


Q3: If the Peer relationship is automatic through auto-discovery, will the branch to branch traffic will also be optimized?


Q4: What is the best method of interception interms of performance, configuration and failover? In-line or out-of-path?


  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Loading.
Zach Seils Fri, 02/29/2008 - 20:49

A1: The current WAE platforms are rated for WAN speeds up to 1Gbps.


A2: There really isn't minimum value. I usually advise customers to take a closer look at what you are trying to accomplish when the RTT latency is below 10 msec. That being said, we do have customers who use WAAS on very low latency links to take advantage of the WAN bandwidth reduction.


A3: Yes.


A4: It really depends on the deployment. WCCP and inline are the two most popular interception methods.


Regards,

Zach



natathyblr Fri, 02/29/2008 - 22:30

Thanks Zach for your valuable feedback.


For the question and answer relating latency,


Based on certain test figures gathered from internet, I understand that the latency higher than 80ms has great impact on bandwidth usage for bandwidths above 6 mbps or more. It reduces the effective bandwidth that can be used, thus not allowing customer fully utlise the WAN pipe.


So based on that, I assume for bandwidths over 6 Mbps and for latency over 80ms ths optimization is effective.


Secondly irrespective of bandwidth size if the latency is very less, say, 20 or 30 ms

will it still make sense for the customer to deploy WAAS?. If, Yes, Am I right to say that the selling point in this case to the customer can only be WAN bandwidth usage reduction, rather than application performance due to poor latency?.


With that thought I am summarizing the possible BW & latency options


1) High bandwidth above 6 Mbps and high latency above 80 ms

Csutomer Solution: WAN BW optimization + Application acceleration


2) High Bandwidth above 6 Mbps and low latency ( 20 ms or less )

Customer Solution : WAN bandwidth optimization only. Customer won't appreciate application performance improvement in this case.


3) Low Bandwidth below 6 Mbps and high latency above 80 ms

Cusomter Solution: WAN BW optimization + Application acceleration


4) Low bandwidth below 6 Mbps and low latency

Customer Solution : WAN bandwidth optimization only. Customer won't appreciate application performance improvement in this case.


Is that right?

Zach Seils Tue, 03/04/2008 - 06:15

Natarajan,


The bandwidth delay product determines how much data can be in transit at any given point in time. You can find a description here:


http://en.wikipedia.org/wiki/Bandwidth-delay_product


Based on this calculation and the configured TCP buffers on the host systems in your environment, you can determine whether or not the hosts can "fill the pipe".


However, bandwidth utilization is only one aspect of the solution. You also need to consider the impact of latency on application performance. For instance, in most cases, the performance of the CIFS protocol is latency-bound. So even in cases of relatively low latency, CIFS performance across the WAN can be adversely affected. Simply adding bandwidth does nothing to improve the end-user response time.


For latency-bound applications, application-specific acceleration is required. These technologies vary by protocol, but are typically focused on latency mitigation through local acknowledgement, message batching, pipelining, etc.


So in options 1-3 you provided, both WAN optimization and application acceleration will provide benefits. The benefit provided really depends on the specific application protocol behavior. For option 4, application acceleration may not have as big an impact, but this should be evaluated on a case-by-case basis.


Regards,

Zach



Actions

This Discussion