The company I work for is looking to implement truly real-time file replication with file locking over a WAN link that spans over 2000 miles. We currently have a 16-drive SAN setup in our east coast office. We also have an office out in Colorado that will have the same exact SAN setup. The idea is to have those two SANs contain the same exact data at all times, which will allow us to work with the same data pool, and which will also provide use with an offsite backup solution, should a failure occur on either end. We're running Server 2008.

The objective is to enable users in the east coast office to work on files and have those changes be instantly updated on the Colorado SAN as well. We also need there to be file locking so that there will be no conflicts or overwritten changes if users attempt to work on the same file.

Is this scenario even possible, at speeds that would make the files usable? And if so, what software would we need to pull this off? As I understand it, DFS-R does not provide file locking, so if we used that, we would need to go with a third-party product like Peerlock. But I don't even know if DFS-R is an option. Can it replicate quickly enough over a WAN link? Can any product? It seems that if we were to use synchronous replication, the programs would be unacceptably slow, as every write would have to wait for confirmation from the other end of the link. But if we used asynchronous replication, what kind of latency would we be looking at? There is a product from GlobalScape called WAFS that claims to provide "File coherence with real-time file locking, file release, and synchronization" and says that "As files are modified, changes are mirrored instantly using intelligent byte-level differencing to minimize the impact on network bandwidth". So this sounds like synchronous replication, but that doesn't even seem possible, given physical limitations such as the speed of light.

If anyone has any experience with this kind of setup, or knows whether it's even possible, I'd appreciate your input and suggestions, including recommendations for software that we should check out.

  • What sort of WAN speeds do you have? Have you done any latency measurements between the sites? Does this have to be limited to a file share solution or could it be a web based one such as SharePoint or similar? – Sim Apr 27 '10 at 14:14

Whether it is possible or not depends on two(maybe 3) things:

  1. Rate of change (ie How many blocks are you dirtying per day) and
  2. Link bandwidth between the two sites.
  3. The original data you need to replicate

If your rate of change is too high, no amount of compression of the data will ever help you replicate.

On a large amount of data, with a low rate of change and slow link, It's possible to FedEx drives out with "seed" data to get a good head start on the replication. Other than that, there is no breaking the laws of physics.


Have a look at the Riverbed Steelhead appliances to assist with the latency between offices. They also have some other neat ways to manage data at remote offices. Naturally, there are other parts to your solution but these devices really helped in our scenario. Riverbed

  • Does the Riverbed solution allow for file locking or is it just WAN Acceleration and Caching? – Sim Apr 27 '10 at 14:16
  • Mainly WAN acceleration. Can you keep the "live" data on one SAN and use the WAN acceleration to have users in both offices access that data? The backup runs in whatever way you want and the WAN accelerator cuts down the amount of data and reduces the backup window. The Riverbed device can be "warmed" so tha when it is deployed, you get max acceleration on first request by a user. It really reduces some data requests like CIFS and NFS . BTW, I think they will do a demo for you and setup is usually quite easy. It convinced us. And cut support calls! – Dave M Apr 27 '10 at 14:24

Deploy a Linux server (could be a virtual machine) at each site, and set up Samba, configuring it to use a http://wiki.apache.org/hadoop/MountableHDFS root. HDFS would do the replication and locking.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.