Network Latency over WAN

Question

Let me take a minute to explain the background of the challenge I am facing. My company has offices in US and Asia. The users work on large CAD files. The office in US has NAS (Network Attached Storage) where all the files are stored. The US users have no difficulty in opening the files however the users in Asia have to wait for a long time. Sometimes it takes up to 30 minutes to open a file. This affects their productivity. The issue is caused by Network Latency (300 ms). There are close to 300 users in Asia.

Following are some of the options that were explored.

1. WAN Optimization software - There is already a WAN Optimization Software installed but that does not seem to cache the files more than a day.
2. Remote Desktop - This will reduce the time taken to open the file but every operation in the application is slow.
3. Cloud Storage - Risk of storing data with 3rd Party Vendor.
4. File Replication software like Microsoft DFS Replication. The main drawback of this technology is it does not lock the file when it is being edited.

I am open to other ideas as well. However I would like to get your opinion and suggestion of softwares that lock the file and sync data on a real time basis. If someone has already implemented or using such software please suggest the pros and cons.

Answer 1

You already have main options:

• get (a copy of) the data closer/faster to the users and deal with locking, replication (conflicts), caching, pre-seeding caches, network optimalisation etc.
• get the users closer to the data with Citrix, remote desktop, virtual desktops etc.

When dealing with larger data files the latter has been giving us the best results, with the added benefit that it also allows better control of our data and improved the work from home possibilities for the local staff as well.

Citrix's ICA protocol has proven to deal well with relatively high latency / low bandwidth links and still provide an acceptable user experience for our users.

The newer virtual desktop solutions are even better at providing stuff as video acceleration as well which might be needed in CAD.

Answer 2

You're describing the reason the firm RIVERBED was invented -

"The issue is caused by Network Latency (300 ms)"

Do some research - you have what is called a "long fat network".

we see this all the time - even with a "1Gbps Cogent link" between New York and San Francisco.

the users never get more than like 8Mbps per second, per tcp connection (with windows 2008 R2, it was even lower with 2003).

You guys should get some Riverbed's and be done with it... they are $120,000 a pair per site, but what's this network worth to you?

Answer 3

Due to time difference you may be able to avoid the file conflict you are concerned about. DFS definitely is not made for collaboration in that fashion. However, there are ways you can set the namespace to point users to his/her local site. This way they will always open the file closer to them. Once they close the file DFS will replicate to the other site. In case of conflict (two users in different sites made changes to the same file) the file saved last will take precedence, the looser file will be put under Conflict and Deleted folder, so that the work will not be completely lost.

To me this is much better than centralizing files and live on the mercy of network availability, WAN optimizing appliances etc. those should be used to supplement the DFS solution.

With that said, DFS replication requires Windows servers on both ends. So the NAS in US will need to be fronted by a Windows server for DFS replication to replicate with the Windows box in Asia.

Answer 4

Reading your post it seems latency is the problem.

Are you sure the network throughput between your two sites is not involved at all? How good is it? If really latency is the only culprit, I would try to use another protocol than CIFS (I assume that's what you use as you considered dfs). CIFS is a really chatty protocol that creates a lot of back and forth communication which makes it a pain to use on high latency network. Can you consider using things like NFS instead?

A proper ECM light help you share those files more efficiently. take a look at Alfresco (open source and free community version or paying enterprise) it has support for many protocols, locking and file transfer/replication.

Is 300ms the normal latency for your network or is it caused some bottleneck? if this is because of a bottleneck and you can't afford a link upgrade, some QoS on the link will help you prioritise flows.

regarding the options you posted:

1: is mandatory. beside caching look at QoS.

2: RDP is usually bad on high latency networks as well.

3: if you can afford putting data on the cloud... but make sure you won't end up with happier Asian users but displeased us users

4: still an option if you're stuck to CIFS between sites.