0:05 – Michael Gibeault

Hi, I’m Michael Gibeault, and I’m your host for Episode 10: Data Migrations with my colleague Joey Adams who wrote a great article on migrations. Joey is the Director of System Operations at TCDI. Welcome, Joey.

0:21 – Joey Adams

Hello. Hello. Thank you for having me on.

0:24 – Michael Gibeault

Absolutely. Hey Joey, let’s start out with tell me a little bit about your role at TCDI and the experiences you’ve had with eDiscovery and data migrations.

0:33 – Joey Adams

Yeah. So, I am over our Systems Operations department. I have been at TCDI since around 2001, so I’ve been here for a little while. I started out doing software development, did some engineering roles, worked on site with some of our clients for a while. I’ve done a lot of implementation engineering, and now the team that we have now, we’re in charge of our client facing software.

So, we do all the software installations. We do the second level support. We do the database administration. We help with the technical aspects of migrations and things like that. So, we are in the trenches day to day with all the technical work.

1:16 – Michael Gibeault

So, that’s probably where you got a lot of your knowledge and information for the article that you just wrote, and I’ll reference that at the end of the discussion here. So, why don’t we start where, you know…let’s talk about data migrations. You mentioned in the article that it is not just a technical exercise. Can you explain what that means?

1:38 – Joey Adams

Yeah. So, the technical piece, of course, is just getting the data from point A to point B. Making sure that you have, you know, the same number of rows that carry over if you’re expecting the same number of rows, and that the images and the natives and all those things, carry over.

But what’s also important is the context of all of that data and making sure that that’s not lost. So, the things like search histories, reporting, production information, and things like that. Making sure that they’re accounted for and moved over and accessible in the same ways that they were in the original system.

2:15 – Michael Gibeault

Before you even move a single file, you recommend segmenting matters into migrate, archive, or delete. You know, what are some of the common misconceptions legal teams have when making these decisions?

2:30 – Joey Adams

Yeah. I mean, there’s the thought that you would just want to carry over everything. And that is one way to do it. But a lot of times, the time of the migration is also a good time to do some housekeeping – to evaluate what you have.

And when you’re creating that internal database, you can check and see, do we need all of that? Is this something that really is needed, or is it something that, you know, this matter was closed a decade ago? We really just don’t need it anymore. It’s more of a liability than a helpful thing to have.

Or looking at duplicated data and saying, oh, well, this is actually the real copy that we want to have. Not this old backup workspace that we had before that somebody had fiddled with, but now it is no longer needed. So, it’s making calls like that on, you know, like you said, migrate, archive or delete.

3:26 – Michael Gibeault

There are two main methods, to my understanding, right? Migrating eDiscovery data all at once or phased. How do you guide clients in choosing the right approach?

3:38 – Joey Adams

Part of it can be the volume of the data. If it’s just a single workspace, you’re probably going to do that all at once.

If you have a large environment across many workspaces and you have terabytes and terabytes of data, potentially with, you know, decades of work product. You might want to look at more of a phased approach where you can start with, say, one workspace, verify the process, and then work through the remaining workspaces as you go.

The risks there, though, are when you’re migrating a system piecemeal like that, you have to be highly aware of what work is actively going on in the system. And so that once you migrate something over, you either need to make sure that the people working in that workspace are now working in the new system for that subset of the data, or if they’re working in the original system, that they’re not making changes that wouldn’t be propagated to the new system.

And we’ve had situations, with both of those models in the past, and it really just depends on, what work is going on at the time, the volume of the data, and, you know, what business processes are involved.

4:47 – Michael Gibeault

Well, and that takes me to the next question of, you know, what are a few of the operational risks of running two platforms in parallel during the phased migration? You know, what are a few of the risks of doing it all at once?

5:01 – Joey Adams

The risk of doing it all at once is mostly tied to the volume of the data. So, if you have a huge amount of data and you’re trying to do it all at once, then that could result in a large downtime potentially, trying to copy everything over just the amount of processing time required. You know, if that were to estimate out to maybe take a week, a lot of teams can’t afford a week of downtime. That just can’t happen.

And then on the flip side of that, if you’re doing it in a piecemeal fashion – kind of what I said earlier – is you run into the issues of maintaining state between two separate systems. That you have one over here and you’ve copied a piece of it. Now you want to make sure that you’re working in the correct one, otherwise they can start to drift apart.

And now you’ve done the work to migrate the systems. If you’re not careful, you might run into a situation where you need to re-migrate or do extra work to resync them or keep them in sync. And keeping systems in sync is sometimes needed, but never fun and something you want to avoid if possible.

6:06 – Michael Gibeault

So, let’s talk about execution, and you know, the real-world challenges. So, once the planning and strategy is complete, what are some of the biggest technical hurdles teams typically face during execution?

6:20 – Joey Adams

I mean – there are issues around just data volumes, figuring out best ways to transfer the data. You know, are you mailing drives? Are you doing it, you know, over a secure link over the internet, and, you know, what are the bandwidth limitations in those scenarios?

Looking at compatibility between the two systems: how are you exporting data from the one system, from the source system? How are you importing the data into the destination system and making sure that the teams responsible for both sides are communicating and coming up with a plan for how those two systems are going to interact?

And then, you know, having a plan for those different context pieces like the searches and the reporting and the production histories and user permissions and all of those sorts of items.

And then – also even thinking about new data coming into the system and making sure that any new processing platforms, if there was a processing platform migration at the same time, making sure that things like the metadata extraction and hashing is compatible with the existing data that was migrated over. And if it’s not, then you’re looking at a situation where you may have to rehash data or something like that – which isn’t always a deal breaker, but it’s work that needs to be done and something that needs to be accounted for.

7:52 – Michael Gibeault

Let’s focus in for a minute on deduplication and hashing, alright? So, this can take us into the weeds quickly, but can you explain, you know, briefly how platform differences can throw a wrench into things if not accounted for in this process?

8:08 – Joey Adams

So, a hash is a string of numbers and characters that represent the data in the document, and it’s going to be unique per document. It’s a nice way to compare whether documents are the same without having to look at the entire document. You can just compare this little value, and if they match, the documents are the same. If they’re different, the documents are different.

And, generally, that’s done by looking at the full text of the document and then some associated metadata. So, for an email you might look at who it was from, where it was sent to, the subject and the body, and maybe some other things.

Now where it gets tricky is that two different processing systems might not look at the exact same fields. Or maybe when they extract out the email addresses from the email, they might include extra space between the name and the email address, or they might use a comma and a space or something like that.

And depending on what gets fed into that hashing algorithm, you’re going to get a different value. So, you might have the same email, but because one thing looked at something slightly different than the other system, you’re going to get a different hash out of that. And it’s going to say they’re not the same when they actually are. And so that can be a tricky thing.

9:24 – Michael Gibeault

So TCDI has handled migrations involving petabytes of data. Can you walk us through what makes those high stakes projects different, and how you manage the associated risks with those?

9:38 – Joey Adams

Planning is, of course critical, because whether you’re dealing with 100 records or 100 million records, everything needs to be accounted for. Every document, every image, every native. You know, one missing image is not acceptable in either situation.

And so, as there gets to be more and more data, planning becomes more and more critical. Working closely with the client becomes more and more critical. The ongoing effort to organize the project and make sure that that staged approach is going correctly and is on target.

And that everyone knows, the communication aspect of making sure that folks know as the migration is ongoing, are you working in the right location? Are there safeguards in place for if people make mistakes? You know, what happens, because when you’re dealing with enough people and enough data, mistakes will eventually happen. So, what then, and what have you done to prepare for that situation and get ahead of the game?

And then of course QC after the fact. You know, if you’re migrating 10,000 records, you can put eyes on most of those. When you’re migrating 100 million records, you can’t put eyes on every single one, and you have to leverage automation to QC all of that. And so, what do those, what does that QC look like? What reports are being given to the client to give them assurances that this migration has been successful?

So, really it just boils down to that extra effort in planning and communication and that due diligence.

11:18 – Michael Gibeault

Okay. Well, what advice would you give a legal department who are considering a migration, and they’re not sure, like how or where to even start?

11:30 – Joey Adams

Well, come talk to the experts at TCDI, of course.

But, beyond that, you know, at the risk of sounding like a broken record, it’s planning, planning, planning and documenting, documenting, documenting. And making sure that you have a thorough accounting of everything that you have and a thorough accounting of everything that gets migrated over and making sure that the numbers agree. Or at least, you know, if you’re archiving or deleting that you know those numbers, when you subtract them out, agree with what you end up with at the end point.

12:06 – Michael Gibeault

Well, thanks for joining us. So, if you’d like to learn more about how to prepare for an eDiscovery data migration, check out Joey’s articles: eDiscovery Data Migration Challenges and How to Overcome Them and Developing a Successful Data Migration Strategy. We’re going to have a link next to the video.

Thanks for joining us!

12:17 – Joey Adams

Thank you.