Back in September of last year NetApp announced the availability of network compression for SnapMirror (See more here). Ever since then we’ve been getting questions about how SnapMirror compression compares to WAN optimization products and if they can be used together.
To date, those of us in technical marketing and solutions marketing have answered questions as they came in. Lately though the tide of questions seems to be increasing so we figured it was time to address these questions in a larger community. Specifically:
- What’s the difference between SnapMirror compression and WAN optimization devices?
- Can I use them together?
So let’s get to it. SnapMirror compression works by running data through a compression engine and then transmitting the data to a target system(s) where it is uncompressed, reordered and written to disk. This delivers two primary benefits for those of us that own storage and it’s availability:
- Maintains your RPO level in the face of data growth and/or increasing change rates
- Improves your functional RPO over the current network infrastructure
This also has benefits for the network infrastructure team:
- Postpones / eliminates the need for network upgrades or leasing of additional bandwidth
- Reduces potential for network contention which can impact production app performance
The specific savings from SnapMirror compression will depend on the data set. During internal testing of three data set types – Exchange DB, Oracle DB, and Home Directory data sets, we have seen savings as high as 70%. (Srinath has a great Tech On Tap article here that goes into more depth.)
WAN optimization devices works in a similar fashion to SnapMirror compression, compressing as well as eliminating redundant data within the packet stream. The major difference being that WAN optimization devices sit outside the storage so they can handle traffic from multiple applications concurrently. One other difference is that WAN optimization devices are designed to handle issues with high latency, poor line quality between sites.
If your goal is bandwidth reduction on clean (low latency) lines, both SnapMirror compression and WAN optimization devices will typically result in similar benefits and similar % savings. As a result, there isn’t a compelling use case at this time for using the two together.
But let’s face it (and finish answering question #1 at the same time) – SnapMirror compression only minimizes and accelerates SnapMirror traffic. WAN optimization devices minimize and accelerate a wide variety of application traffic – including SnapVault traffic which does currently support compression. And SnapMirror compression is not designed to provide ‘optimization’ for low quality lines. Which leads us to a few key rules of thumb (and answers question 2).
- There isn’t a benefit today to using a WAN optimization device with SnapMirror compression. All you would be doing is creating overhead on both the FAS and the WAN optimization device.
- If you don’t a have WAN optimization devices and are only looking to optimize SnapMirror bandwidth utilization – use SnapMirror compression.
- If your network links have high latency / packet loss – use a WAN optimization device. Then you get compression and network ‘stabilization.’
- If you are looking to optimize a lot of different application streams and already have a WAN optimization device use it. Unless of course the additional traffic would overload the box.
- For those of you tuning for maximum speed: If your available network bandwidth exceeds the maximum compression throughput limits for a given FAS platform – use a WAN optimization device. SnapMirror compression is primarily intended for low bandwidth links and may not be able to complete saturate very high bandwidth links. Refer to TR3446 (SnapMirror Async Best Practices Guide) for more details.
Well, there you have it a few quick rules of thumb and a description of the different technologies. Hope that helps you understand the technologies and determine the best way to save on network costs.
Nathan Moffitt and Srinath Alapati