Hummingbird eDocs - for a geographically distributed architecture

[c***@gmail.com]

Hi guys.

I hope you are able to assist me.

We are considering investing in Hummingbird eDocs as our Document
Management System. The functionality looks good, but it does not meet
our infrastructure requirements.

We have two geographically located offices (London and Dubai). We
would like to implement a full Disaster Recovery and High Availability
solution based on the following requirements:

Both locations should have an exact copy of the SQLServer2005 data
(used by the DM server) and the documents stored in the content
repository (SAN). The replication should be as close to real-time as
possible.

This will allow users in both locations to work on local copies of the
documents, and any changes to either side are replicated to the remote
location. In essence, this is an ACTIVE ACTIVE setup, opposed to an
ACTIVE PASSIVE.

If something serious happens in one of the locations (the server room
burns down for instance) the remote office will have all documents and
the users will be able to continue working on them as per usual (with
a slight lag due to latency).

It has been suggested that what we want is not possible, and that we
should use a centralised architecture, where both offices reference
the same database and document store, and install a caching server
between the remote location and the master server to improve the users
access to the documents. It sounds like an option, but does not give
us a sufficient disaster recovery solution, because if the master
servers died, the offices will not have access to documents.

I would really appreciate any advice or solution you may have to our
requirement.

I look forward to hearing from you.
Regards,
Col

[Milind Joshi]

Post by c***@gmail.com
Hi guys.
I hope you are able to assist me.
We are considering investing in Hummingbird eDocs as our Document
Management System. The functionality looks good, but it does not meet
our infrastructure requirements.
We have two geographically located offices (London and Dubai). We
would like to implement a full Disaster Recovery and High Availability
Both locations should have an exact copy of the SQLServer2005 data
(used by the DM server) and the documents stored in the content
repository (SAN). The replication should be as close to real-time as
possible.
This will allow users in both locations to work on local copies of the
documents, and any changes to either side are replicated to the remote
location. In essence, this is an ACTIVE ACTIVE setup, opposed to an
ACTIVE PASSIVE.
If something serious happens in one of the locations (the server room
burns down for instance) the remote office will have all documents and
the users will be able to continue working on them as per usual (with
a slight lag due to latency).
It has been suggested that what we want is not possible, and that we
should use a centralised architecture, where both offices reference
the same database and document store, and install a caching server
between the remote location and the master server to improve the users
access to the documents. It sounds like an option, but does not give
us a sufficient disaster recovery solution, because if the master
servers died, the offices will not have access to documents.
I would really appreciate any advice or solution you may have to our
requirement.
I look forward to hearing from you.
Regards,
Col

Hi Col,

Technically speaking, what you are asking for is not impossible,
though it could get quite expensive to do it in the way you are
looking to do it.

It looks like disaster recovery is a major required functionality, so
from what you said above, Hummingbird eDocs installed in one location
with caching server does not seem to support the functionality you
require. Explore other ways of implementing disaster recovery, and try
not to mix it up with requiring "fast" access.

One problem with your approach is that "replication" of documents does
not exactly make it clear which version has priority over the other
when 2 or more users simultaneously edit/modify a document. If you
have 2 or more "local" copies being edited simultaneously, then which
one is the "real McCoy" to replicate from/to?

There could be relatively low-tech ways to implement that setup you
want, though the work there may be outside of Hummingbird eDocs - in
fact, the EDMS may not even know that it is being replicated, and an
external name resolution server resolves an address to either this
server or that one based on availability. That is how web-based
services providers implement that... when you go to hotmail.com or
google.com, some server is switching you back and forth between
applications.

Mainly, you may need to define what exactly you mean by real-time
access... and what you mean by disaster recovery, and how fast you can
be up and running after a major outage with different options, and
weigh it against the cost to build out the solution in the ideal case
against the business cost and benefit of taking it down a notch.

What I would do if I was in your place is to see if both locations are
equally important/equally busy/populated, where most of the document
users sit, how often documents are modified, how often documents may
be modified simultaneously, whether an offsite replication with a
certain time lag is acceptable, and whether you can have a disaster
mitigation plan in place of full replication.

Which other EDMS did you consider?

Regards,
Milind Joshi
IDEA TECHNOSOFT INC.
http://www.ideatechnosoft.com

[splitDiff]

Hello Col-

I just wanted to expand on Milind's excellent response a bit.

I'd recommend separating your requirements for disaster recovery and
remote document access. Here is a proposal:

1. Designate one of your offices as the primary data center (if the
user populations are similar, then select the site with the best IT
support).

2. Establish the centralized repository at that location. Build it on
top of a robust SQL server and a speedy SAN. Make sure bandwidth is
not a bottleneck. Follow normal data back-up procedures to protect
against data corruption -- nothing worse than a perfectly-replicated
corrupted database.

3. Create a disaster recovery (DR) data center in that same locality.
Make sure that it's far enough away to avoid common-mode disasters but
close enough so that you can still lease data lines between the two
facilities. Duplicate the hardware found in the primary data center.

4. Establish your near real-time replication between the primary data
center and the DR data center.

5. Bring up the secondary office with the caching server as described.
Use DNS to point the caching server and users local to the primary
data center to your centralized repository.

6. In the event of a disaster or server failure at the primary data
center, simply update the DNS to point to the DR data center.

7. Run periodic DR drills to verify your replication and your
procedures.

Hope that helps

-John

[OnBasePro]

Hi Col,

Implementing a full disaster recovery and high availability ECM
solution like you described cost effectively and without much
complexity can be difficult and in reality it seems like you have two
options: you could explore some alternatives of implementing a
disaster recovery solution with Hummingbird (eDOCS) or you could
explore some other ECM vendors and discuss and evaluate their
approaches. I mention this for a few reasons—one being that while many
products carry the ECM product ‘tag’ just like Hummingbird, there are
many differences (particularly architecturally) that may result in
more satisfactory approaches to DR and high availability. Another
reason is that there is a lot of uncertainty around the future product
direction of Hummingbird’s eDOCS product and as to the amount of
resources and development being put back into the product since
Hummingbird was acquired from Open Text. These concerns and
uncertainty have been expressed openly and publicly from the leading
ECM analysts firms like Gartner, Forrester, CMS Watch, etc and don’t
represent the exclusive opinion of another ECM vendor.

But understanding those concerns and the critical requirements around
designing a satisfactory DR and high availability solution that
haven’t been met, I would recommend evaluating other options alongside
Hummingbird. And although I represent Hyland Software and the OnBase
product for ECM solutions, there are also other enterprise-capable
systems to consider, as well, like IBM/FileNet or EMC/Documentum.

I agree with John’s comment above on separating your Disaster Recovery
requirements and your remote document access requirements, as well.
And with the approach of utilizing two physical locations to support
the high availability of the ECM solution. I know currently we have
just designed a solution proposal for a customer requiring a similar
ACTIVE -ACTIVE setup, which we would refer to as a HOT-HOT setup.
There are inherent capabilities built into OnBase to easily enable our
customers and partners to design high available and completely fault
tolerant solutions for meeting disaster recovery needs that you
referenced.

Regards,

Jim

www.onbase.com