There is a lot of confusion when it comes to WAN optimization and wide area file systems (WAFS). The most frequently asked questions are about the differences, advantages and weaknesses.
WAN optimization and WAFS are technologies designed to significantly increase the effective data throughput on IP WANs. Each of these technologies attacks TCP/IP WAN limitations from very different perspectives.
WAFS is designed for the serverless remote office where the servers are consolidated at a centralized data center. It makes the centralized applications appear as if they are local to the branch office. It does this two ways. For files, it caches the last generation of a file in an appliance at the branch office. It caches on disk or in memory. It also liberally uses common sequence data reduction. This is another way of saying that it only transmits blocks of data it has not transferred before. For "chatty" applications such as CIFS, Exchange (MAPI), Notes, SQL Server, HTTP, HTTPS and others, it spoofs the protocol as well, so that non-data commands do not have to go roundtrip on the WAN.
The advantage to WAFS is that it enables centralization of branch office applications, file servers and mail servers. And it does this while mitigating the performance/response issues most users typically have when the application is remote. Production results have ranged from a 300% to 7,800% increase in effective data throughput. Centralization also reduces the total cost of ownership for the organization because of economies of scale. And it eliminates technical expertise requirements in the remote office.
One highly touted WAFS benefit is that it eliminates remote data protection expertise, skills, software and equipment because it is all handled centrally.
There is, in fact, some downsides to WAFS. When their network is disrupted, the branch is cut off from their applications. Desktop applications may still have access to files in cache; however, remote application access is down. When a centralized application is disrupted or taken down, the productivity loss is felt at ALL of the branch offices. This does increase productivity risk to a greater degree (all your eggs in one basket mentality).
And finally, WAFS does not solve the data protection requirements for desktops, laptops and servers that cannot be moved centrally because the application does not lend itself to this technology.
If the users are disciplined in putting all of their important data on the centralized data servers, then data protection is only an issue for "bare-metal-restores" and the applications that can't be moved. There are a number of agentless and agent-based distributed data protection products that can fill this gap (a future blog).
WAN optimization operates under completely different baseline assumptions. It attacks the TCP/IP performance limitations directly to improve effective data throughput between any two sites. It assumes that data is moving between applications at each site and is completely transparent to the applications. WAN optimization varies; however, in general it performs similar functions. WAN Optimization eliminates TCP latency by terminating the sessions locally and using proprietary protocol enhancing proxies between the optimizers. It mitigates packet loss events by using selective acknowledgements, retransmitting only those packets that are lost and reordering the packets upon delivery. It increases bandwidth utilization by providing dynamic window sizing. And it increases throughput beyond the rated bandwidth by compressing the data using common sequence reduction.
The combinations of these WAN Optimization technologies significantly reduce the amount of data going across the WAN while dramatically increasing the effective data throughput. Production results have shown increases of 300% to 2,000%.
There are two types of WAN optimization: One is for branch office to central or DR site and is call branch data replication optimization (B-DRO); the other is for data center to data center or DR site data replication optimization (D-DRO). D-DRO typically has orders of magnitude more data transfer requirements than B-DRO.
B-DRO is optimized for low bandwidth -- less than 2Mbps. D-DRO is optimized for high bandwidth -- up to OC12 or 622Mb/s. D-DRO is a much different technology challenge than B-DRO and utilizes very different techniques. D-DRO can typically scale down to the lower bandwidth of the branch office. B-DRO can rarely scale up to the D-DRO.
The downside to WAN optimization is troubleshooting or tuning when there is a problem. (Some products and vendors are better than others.)
There are a couple of final notes to consider. Just like the EPA and mileage estimates, WAN optimization and WAFS "effective data throughput" vary significantly. Test it in your environment before buying and compare at least two suppliers.
I will provide greater detail on each technology in future blogs, specifically covering where and when to use them. Send me your questions, and I will make sure they get answered.
About the author: Marc Staimer is president and CDS of Dragon Slayer Consulting in Beaverton, Oregon. He is widely known as one of the leading storage market analysts in the network storage and storage management industries. His consulting practice of six+ years provides consulting to the end-user and vendor communities. You can reach Marc at firstname.lastname@example.org