Research projects in a small company

2 February 2010 #

Thanks guys.  We are pretty excited with the idea.  Really frees up the creative juices to think out of the box for a bit of other ways to solve a problem rather than only in one context.

I try really hard to be as transparent as I can with our customers.  It is a fine line between freaking people out because you are thinking outside the box, and a well thought out strategic decision.  

Hopefully people will understand the intent of the above post.  We do have a plan, and I hope some other small business owners look at this for ideas as well.

js_vistadb

3 February 2010 #

Some of the RedGate tools are in our sites at this point.  Theirs only work with SQL Server, anything we build will work with VistaDB and SQL Server at a minimum, others if possible.  And they don't offer very many APIs for their products.  If you do buy their API it costs per machine you deploy to for runtime fees as well.

We think the data generation system we are working on will meet the needs of developers for data generation far better because of the API.  If you want to build known databases for things like unit testing then you have to be able to talk to the generators through an API.

js_vistadb

4 February 2010 #

All statements of things we are thinking about are always forward thinking and subject to change.  I publicly discuss ideas and concepts with the community.  Doesn't mean each of them will ever come to fruition.  All roadmaps are always forward thinking for companies.  I have always tried to be open with people about our processes, but we never commit to shipping anything until we are 100% certain we can do so.  

The server version is not feasible to my satisfaction without a lot of additional research first. This is research and changes we have been doing for 2 years.  The same algorithms that work on the desktop don't scale to multi user server scenarios, streaming data, scale for writing to disk, multi user security, etc.  All of these require additional research and development.  

No one is willing to pay more than SQL Server for a lightweight server of VistaDB.  So how do you pay for the R&D costs to make that happen?  This is our way of finding ways to pay for R&D without the price of VistaDB skyrocketing to cover that research.

Our internal cost estimates to build a server that does what a "sql server" product should do weigh in at over $1 million USD.  That is in addition to the amount we have already spent building the current engine.  Small companies cannot afford a 3-5 year recoup time for that cost, because it comes out of my pocket as we build it.

People don't pay for R&D, they want finished products.  So how does a small business pay for these large products that need huge amounts of R&D?  Either you take VC money (something I have never done), or you bootstrap.  The way we are going to bootstrap is by paying for R&D through additional products.  Each of them are small compared to a database engine, but solve real world needs.  As long as you solve a real world need, then it should generate some revenue to help that R&D funding process.

This is all just part of the reality of being a small business with limited resources.  You have to find creative ways to make things happen.  I am sharing that process with everyone, because over half of our customers are also small business owners.  I am sure they feel the same pain we do at these types of decisions.

js_vistadb

5 February 2010 #

In order to change something in the database engine like the SQL Parser there is about 6 months of work that will not "show" anything to users.  This happens a lot more than you might expect.  Major subsystems require huge amounts of work to redesign and implement.  If you take that component out of the engine and put it into something else you can worry about the implementation directly, and not how it fits into VistaDB until later.  You will actually be more productive, and not spend as much time trying to keep the core engine stable.  It doesn't matter for this subsystem - just test that part.  Now once that part is done, you need something real world to test it.

An example - We are working on new FTS and text indexing systems.  We have implemented them into an ASP.NET Site Search component.  We needed a site wide search, it is our exact domain of problem knowledge, and by productizing it you will get a much wider testing of scenarios.  

And to be honest, people don't want us to release so often.  We are constantly getting complaints from companies that we release too frequently.  They want a 12-18 month cycle (some even longer).  The current plan is to probably change to a quarterly release cycle.  Fewer updates mean less integration time for clients, and each release we do costs us a LOT of hours in testing, documentation, installs, etc.  I think we will probably find that just by slowly our release cycle we will have a lot more time to do these research projects.

Customers have also told us repeatedly that they don't want "new" features (fear of change).  They want it to be stable, stable, stable.  So by doing these risky things outside the core product we can get feedback on techniques and algorithms without worry of freaking out those more conservative users.

The problem with small business (especially today), is that you can't just hire more staff.  It is not viable.  But that is an entirely different blog post about people not renewing subscriptions.

js_vistadb

5 February 2010 #

I should also point out that we have always done "sideline" projects here at VistaDB.  Customers who needed design reviews, custom versions of the engines for partners, consulting work for special one off features for large customers, etc.  I guess a lot of people don't know about that stuff because we don't talk about it publicly very often, but we have always done non product work.  We have even done custom programming for customers to solve specific problems they had totally unrelated to VistaDB as consulting agreements.

I think that is just life in a small business.  You can't dedicate 100% of each employee to a single task, just not going to happen.

js_vistadb