March 12, 2011 at 6:39 pm
Comments posted to this topic are about the item NoSQL is Not the Answer
March 12, 2011 at 9:57 pm
NoSQL types of databases may, in fact, be the answer. Now, lets figure out what the question is. 😛
If you need a database that is essentially nothing but a place to store and retrieve data to and from, like some lightweight desktop application or perhaps a smart phone application or simple inventory for a refrigerator application (yeah, they're here), the NoSQL database types may, in fact, be the way to go.
Change your requirements to something that has to support thousands of users on a system that has mega-row reference tables and giga-row main storage tables and, you're correct, NoSQL probably isn't going to hack it.
Like everything else in our business, "It Depends". 🙂
--Jeff Moden
Change is inevitable... Change for the better is not.
March 14, 2011 at 2:49 am
I think most seasoned DBA's learned long ago that there is no such thing as an instant one size fits all out of a glossy shrink wrapped box when it comes to capturing, storing, retrieving and presenting data, information and intelligence. If there were, almost every vendor on the planet would be out of business. SQL has provided us with an answer to a very wide range of issues and given us a "standard" and a "platform" everyone can work with and perhaps more importantly, that we can "live" with day on day ... for the longer term. NoSQL has yet to provide anything comparable we can get a handle on but, it may well have merit and like everything else that comes along, is worthy of looking into for what we may be able to use it for.
Regards.
Ted Smith
March 14, 2011 at 4:17 am
It is definitely not the answer to most common real-world full-business data requirements. BUT it is certainly a candidate for some niche problems where either a full RDBMS is overkill or where an RDBMS's complexity is not required and sits actually in the way of performance/scale (wouldn't want to run Google's search engine on a MS-SQL platform).
Where an RDBMS is merely overkill, you could look at whether the maintenance-overhead to support 3 platforms (document file-system, relational transaction data AND one or more NoSQL databases) instead of 2 (the documents with relational database) is worth it (yes, the business WILL still need an RDBMS for many business-processes). Or whether it's just easier to add the additional database(s) into the already existing (and maintained/supported) RDBMS.
For those niche problems NoSQL will bring a lower-cost solution than scaling up/out of an existing RDBMS, and so by all means use it.
If traditional RDBMS and NoSQL platforms can be seamlessly integrated into a single platform and the strengths of each applied to the appropriate problems without adding maintenance/cost overheads over a one or the other solution I might look into NoSQL in my field (mostly automating/integrating complex business processes).
Although: I'm still waiting for seamless integration of a document file-system with the transactional database, although I might be out of a job when off-the-shelf solutions would become that easy.
Yes, I know that NoSQL could provide an interesting alternative for file-shares for document-storage, but it would still require maintenance/integration of 2 platforms without much benefit over the current 2 platform solutions, but that may change in the not-so-distant future, so I'll keep taps on this new technology.
March 14, 2011 at 7:51 am
I agree with everyone who's posted so far. I think the sensible approach is to see where NoSQL fits in and where SQL fits in and use the best tool for the job. The current battle of comparisons where people take extreme views seem unproductive. My guess is that the future will see some combination of SQL and NoSQL, but that it would be a mistake to abandon SQL entirely, especially for well-normalized databases. The thought that went into that normalization probably includes lots of safeguards for data integrity that would be put at risk if those databases were rashly converted into NoSQL with no way to roll back from them. At the same time, NoSQL could work well with data extracted from SQL databases to improve performance in many different scenarios (web sites, reports, etc.).
I hope cooler heads prevail and sensible compromises win out. Just my two cents.
- webrunner
-------------------
A SQL query walks into a bar and sees two tables. He walks up to them and asks, "Can I join you?"
Ref.: http://tkyte.blogspot.com/2009/02/sql-joke.html
March 14, 2011 at 7:52 am
Just because a database platform is good enough for Facebook or Google, that doesn't mean it's good enough for online banking or e-commerce. I believe that financial and operational data, where every transaction has a dollar amount attached to it, is held to a higher degree of scrutiny than a page full of web search results or some teenagers guest book.
NoSQL databases may be the answer for a write one / read many datamart serving a content management application, where all you are doing is bulk loading and bulk reading, and the end result isn't audited by accountants. I'd bet that even Google and Facebook are running their own internal CRM and financial database applications on SQL Server or Oracle.
"Do not seek to follow in the footsteps of the wise. Instead, seek what they sought." - Matsuo Basho
March 14, 2011 at 8:21 am
NoSQL DB's like Cassandra are good at scaling out rather than up, more so than SQL Server. It allows a dept to move away from expensive big-iron SANs and just scatter your data across dozens of commodity servers. In theory, one could save significant money with the licensing costs (zero for Cassandra vs. ~25-30K per socket for SQL Ent.) and on hardware (no SANs, no wasted servers on MS clustering). But, Cassandra is not ACID compliant so data integrity is not implied nor guaranteed. It may also require more manpower to manage as with most open-source applications unless you employ only rock stars. As far as the volume it can handle, I don't have a clue.
March 14, 2011 at 9:05 am
One reason for using NoSQL is when there is no need to be careful about how safe the data is. NosSQL can be the "down and dirty," or "git 'r' done" mode at times. MS SQL, on the other hand, is so careful not to loose data, but sometimes the data's not worth much to begin with. In these situations, it's either easy to recreate or doesn't matter much to begin with.
The greatest enemy of knowledge is not ignorance, it is the illusion of knowledge. - Stephen Hawking
March 14, 2011 at 9:29 am
Good comments, and very true. NoSQL doesn't necessarily work for many of the highly transactional systems we have. However it might be nice for others. Maybe DSS, maybe some of the warehouse/query systems where there are scale issues. Maybe for some corporate sites that are doing more publishing of data and aren't so transactional.
It's worth learning about and making an evaluation. Lots come into play, including supporting multiple platforms, learning tech, etc.
March 14, 2011 at 10:07 am
We are currently evaluating a NoSQL solution (HyperTable) for a "datamart" type of application. We have gigabytes of weather observations that need to be searchable by location, date, and observation type. The indexing options are rudimentary at best, but the highly distributed nature and ability to handle gigantic datasets seems like it would be worth the trouble of working the rest of the way through the implementation.
I figure we will probably use a RDBMS for the more complex queries, which will return a key range of sorts for the actual data retrieval from the NoSQL tables. I'd like to hear more about other hybrid solutions that others have tried.
March 14, 2011 at 11:14 am
The primary trouble I've seen with the NoSQL movement so far is that a large portion of those who are migrating to it are using it because they don't understand relational databases.
We all probably know the type: young people, a few years out of school. They read all the tech blogs, speak of sprints and patterns and can't be bothered with anything as stodgy and ancient and hidebound as data design. They chafe at the restrictions, and create tables that look like spreadsheets and use columns of type XML and refuse to understand what a surrogate key is.
THESE guys are the ones who are jumping all over NoSQL, in large part because it means they don't have to think about hard stuff like consistency and isolation.
NoSQL databases have their uses. People at Google, for example, understand relational databases very well and they designed something to get round the limitations that relational databases have in a specific set of circumstances. But nobody at Google claims that BigTable will replace a relational database.
The fundamental problem is that in order to understand what you are giving up, you have to understand databases and at least a medium amount of database theory, but the people who are leaping all over NoSQL as the Next Big Thing and making architecture decisions around data storage 1. generally don't and 2. generally won't talk to someone who does.
And so as with everything else, a lot of bad NoSQL databases are going to be built before this is over.
March 14, 2011 at 11:26 am
And so as with everything else, a lot of bad NoSQL databases are going to be built before this is over.
That might not be a Bad Thing, if it means that far fewer bad RDBMS databases are going to be built.... 😀
March 14, 2011 at 2:15 pm
I think most developers most of the time don't think about consistency in their application. SQL is so robust and ubiquitous that they take it's guarantees for granted even when working on platforms that don't provide ACID guarantees, including both NoSQL, and non ACID compliant sql databases.
Of course that breaks both ways. Alot of the time SQL is actually overkill for simple object persistence. But I'd rather have it than not until you need to start giving up consistency for scalability.
March 14, 2011 at 3:20 pm
As a frequent visitor of the procedural developer forums I can tell you that a good number of the developers in support of the NOSQL DB's do so because they like the idea of not having to learn/use that
stupid/archaic/hard as Latin/<put your favorite negative descriptor here>
and so they see the NoSQL as the answer to their dreams. Instead of having to learn how to do Relational DB right they can cheat by moving to a NoSQL system. Now granted some of these guys have very valid arguments and reason for these kinds of NoSQL systems and their scenario warrants something like it. however far too many try very hard to do everything they can to avoid learning how to use SQL and to use it the right way.
The move from a procedural mind set to a set based one is for many a developer, the equivalent of kryptonite to Superman and the first snake oil NoSQL salesman to come along cons them easily into buying into the "NoSQL-Fits-All & Does-All" line.
I used to work at a sofware company that had little to no interest on doing the SQL side of te
their application right. Whatever worked, worked and that was good enough. It didn’t matter whether it was the most efficient way of working so long as it worked. Now with many clients moving into the hundreds of gigabytes sixed DB's the good ole "Just so long as it works" no longer works due to performance problems. And so what is the solution? Well its not to learn how to do RBBMS & SQL the right wy but to use a scale-out NoSQL system.
But I digress.
Kindest Regards,
Just say No to Facebook!March 14, 2011 at 4:24 pm
A good guide to the capabilities of the NoSQL environments is the diagram on Nathan Hurst's blog. http://blog.nahurst.com/visual-guide-to-nosql-systems
As I continue to work with computers and in the industry, the more I believe that truly the answer is "it depends". So the question then becomes "What does Product X do for me?"
If some of the NoSQL solutions can bring a defined benefit to a project I'm involved with, then I want to explore that.
Being closed is simply not being open, and I want to never be closed minded.
Viewing 15 posts - 1 through 15 (of 45 total)
You must be logged in to reply to this topic. Login to reply