The Riverbed Blog (testing)

A blog in search of a tagline

Bogus WAN acceleration claims

Posted by bobegilbert on July 2, 2008

Peter Sevcik and Rebecca Wetzel recently blogged on Network World about the bogus math and bogus acceleration claims that WAN acceleration vendors are making.  They make a pretty good point about the bad math being used in many cases to calculate what the X improvement is as a result of their acceleration product.  You can read about the math and suggested calculation to use here,  but the net-net is that they believe that vendors are mis-informing potential customers by making a 100X improvement look much more than a 10X improvement when it is actually closer than it appears.  They close by advising customers to ignore speed improvements above 5X and focus on more useful factors when evaluating WAN acceleration vendors.

While I agree to some extent their bad math claims, I believe they are mis-stating the importance of the performance improvement provided by WAN acceleration vendors.  At the end of the day, how important is improving the response time as close to LAN-like as possible?  Is it good enough that it is a bit faster?  I believe this largely depends on the application…

Regardless of the math used to calculate the improvement, reducing the response time for a slow application by minutes or even seconds can be extremely important.  Riverbed has gained a reputation as having the fastest performance and best acceleration results for key applications.  A combination of better deduplication, robust TCP acceleration, and a number of application-specific optimizations is what gives Riverbed industry leading acceleration for CIFS, NFS, Microsoft Exchange, Oracle 11i, Siebel, SharePoint, NetApp SnapMirror, and more.

Let’s take a look at a few of these applications and how Riverbed’s acceleration fares in competitive evaluations and how the performance difference was an important criteria for ultimately choosing Riverbed as the WAN acceleration vendor.

Windows File Sharing (CIFS)

It is well understood that windows file sharing performance over a high-latency, unaccelerated WAN link is one of the most miserable application experiences that one can have.  An operation that takes seconds on a LAN is slowed down to minutes and in some cases hours over a WAN.  Many WAN acceleration vendors provide application-specific acceleration for these CIFS environments and the results can be pretty dramatic.  Many vendors can speed-up the drag and drop of a large file to a file share by orders of magnitude (over 10X).  Using Sevcik’s and Wetzel’s math, the difference between a vendor that demonstrates a 100X improvement vs. a vendor that provides 10X acceleration may indeed not be that significant for a user that is used to the poor performance.  However, for environments that are used to LAN performance, the performance difference may be critical.  This is the vitamin vs. pain-killer premise.  Organizations that have not yet consolidated, but are using WAN acceleration as the tool or vitamin that enables consolidation will most likely have more stringent requirements for acceleration.  For organizations that have already consolidated their file server infrastructure and there is user frustration over poor performance, any level of performance increase might suffice as they don’t have a LAN baseline to go by.  It is also important to note that file sharing is more than just dragging and dropping a file.  Enterprise work flows might include opening a file, working on the file remotely, editing the file, and saving a new version of the file.  Riverbed has optimized its CIFS acceleration to cover a variety of different use cases so more than just drag and drop is LAN-like and overall the experience just feels more "snappy".  Other vendors do not have CIFS acceleration as robust.

Web-based business applications like Siebel, Peoplesoft, Oracle 11i

For web-based business applications, response time is typically measured in seconds and the more seconds you can shave off of the user experience, the more productive the user and potentially the more money saved and the better your business is.  For example, a recent evaluation running Siebel demonstrated that Riverbed reduced a 9 second response time down to 4 seconds where the competition was only able to only go down to 7 seconds.  Using an X improvement or even % improvement does not really tell the story.  Being 3 seconds faster than the competition for a common operation performed throughout the day by thousands of employees results in a significant improvement, productivity increase, and overall business win.

SnapMirror Data Replication performance

Backup windows for replication jobs is often measured in hours or sometimes days.  Reducing a replication job by seconds or even a couple of minutes will typically not provide the justification needed to purchase an WAN acceleration device.  However, reducing a replication job from days to hours or from hours to minutes can have a dramatic impact on the ability for an organization to better meet their RPO or RTO objectives.  A recent competitive evaluation demonstrated that Riverbed was able to deliver over 382Mbps of throughput over a 45Mbps pipe.  Riverbed’s closest competitor delivered 90Mbps, while some competitors had a challenge meeting 45Mbps.  For this customer, the difference between 90Mbps and 382Mbps was not Riverbed’s marketing math, but real-world performance improvement that meant better RPO and RTO than without acceleration or with the acceleration levels provided by the competition.

There are many more examples of different applications and how different acceleration levels can impact the business need.  Sevcik and Wetzel are right that vendors claims about acceleration don’t always use the correct math when they talk about X improvement.  At the end of the day, forget the claims and test out the products for yourself, comparing the specific performance of each vendor and how that performance aligns with your business need.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: