by Marc Ferreria, Practice Leader & Senior Networking Consultant at Taos
Where does it make sense—cost-wise and performance-wise—to invest in WAN Optimization equipment? This is a question some clients will have. There are a few of them on the market, and they all claim to speed up application performance and save money on bandwidth. How do we in good conscience advise a client on whether they should invest the money in WAN Optimization equipment, and at which of their sites? Ideally you would do some testing with demo equipment, or on a try-and-buy basis, to obtain real-life data for your recommendations. Before then you can look at the specs on the brand of Optimizer you are considering and you can get a good prediction on how it should perform.
I had the fortunate opportunity to have a client who had deployed 40 WAN Optimizers 3 years ago. The client wanted to know if they should continue investing in them, and if so at which sites or should they phase them out because of no real ROI?
I suspected a general rule of thumb is that the users will not see any noticeable performance increase unless they are at least 15ms away (this would be Los Angeles to the client’s Data Center in the Sacramento area) and the vendor agreed.
The existing WAN Optimizer deployment was showing:
- 46% average optimization in their primary Data Center
- 28% average optimization in their DR Data Center (these two would be expected to reverse in the case of a fail-over)
- 76% average optimization in their smaller San Francisco Data Center
- 60-70% at their Corporate sites
- 55% average at their typical remote office
These numbers represent the amount of bandwidth removed from the WAN for the traffic which is being optimized. There is other traffic which is not optimized, so if a remote site is seeing 55% optimization this does not mean they are saving 55% of their WAN bandwidth necessarily.
None of their applications are being optimized because they all use SSL/TLS for access and the WAN Optimizers are not currently decrypting this traffic. The above numbers can be expected to increase noticeably once this feature is enabled.
I examined these criteria in the analysis:
- Perceivable performance improvement: For performance, I used only CIFS/SMB as a gauge since their apps are not optimized. Using a 35MB Visio file, CIFS file transfers to/from LA (15ms away) decreased from 8 seconds (no optimization) to 2 seconds (with optimization). To/from the East coast transfer time decreased from 16 seconds to 8 seconds. Each test was performed 3 times and averaged. I was unable to gather an accurate test from a site closer than LA because any test I did would saturate the WAN link, severely skewing the results.
- Impact on nightly backup jobs: Impact on nightly backups was basically none, because these are either already compressed, de-duplicated and/or encrypted. In these cases the WAN Optimizer will actually slow down the backup job and sometimes cause it to fail. It is best to install a pass-through rule for this type of traffic.
- Bandwidth savings (WAN capacity upgrade vs. cost of WAN Optimizer): Number 3 is the test which surprised me. I was certain the cost of a WAN Optimizer strictly to avoid upgrading circuit bandwidth would never result in a meaningful ROI. However, it turns out that in the case of this environment, the cost of WAN Optimization equipment is actually one third the cost of upgrading a pair of WAN circuits by 10Mb (e.g. from 10Mb to 20Mb or 20Mb to 30Mb). Each of their sites has two MPLS WAN circuits and two routers and most sites have 10Mb circuits. The cost of upgrading bandwidth is about $700/mo. for the two circuits combined. The cost of a 20Mb WAN Optimizer (upgradable to 30Mb), amortized over 5 years is about $134/mo plus $100/mo for maintenance. So $700/mo. for circuit upgrades compared to $234/mo. for a WAN optimizer is considerable savings. A smaller WAN Optimizer (10Mb) could conceivably be used at these 10Mb sites but I wanted a worst-case comparison.
The most dramatic improvement reported in this environment was from an office in Hawaii where employees were waiting 40 minutes to embed the link to a file stored in San Francisco into an e-mail. This was still faster than downloading the file and attaching it to the e-mail. With a WAN Optimizer employed and encrypted MAPI supported the time to embed a file link reduced down to 1-2 minutes.
It is also worth noting that these Optimizers are supporting more and more applications as time goes on, and you can improve the users’ experience more and more as these additional features are employed.