MongoDB CEO Claims They're Luring Customers From Oracle (diginomica.com)

"MongoDB is increasingly encroaching on Oracle's database lead -- with enterprises becoming more and more confident with the maturing NoSQL technology," according to Diginomica, citing this new interview with CEO Dev Ittycheria: 30% of our business is migration off existing workloads to us. Two years ago it was 5%. Ditching Oracle and others, but mainly Oracle... one of the nice benefits of being in this market is that Oracle has done a great job of alienating its customer base... if there are performance reasons, regulatory reasons, developer demand -- [people] will change... We have grown business by 2.5X over last two years. And our employee base has pretty much doubled.
One reason he cites is Oracle's higher prices on their top-line products, saying MongoDB's new customers include "a large bank, whose logo you would recognize instantly [with] a very sophisticated equities trading platform." Ittycheria says MongoDB is now a nine-figure business, and after they launched their new database-as-a-service product Atlas last June, "the growth in that business has been off the charts."

Comments

ACID doesn't matter...

[tlambert]

"ACID doesn't matter... your first install is free!"

could just be the beginning

[sribe]

From my experience, I'd guess that about 90% of Oracle installations do not need Oracle. They're all ripe for migration to PostgreSQL or MongoDB. (Granted the 10% of installations that are big enough to need Oracle will be way more than 10% of Oracle's database revenue, but it's still a nice market segment, ripe for the taking.)

Re:could just be the beginning

[fustakrakich]

If Mango can produce enough revenue, Oracle will make them an offer they can't refuse. In the meantime, they can let them (Mango) do all the dirty work of gathering clients.

Re:could just be the beginning

[Anonymous Coward]

Mongo just pawn in game of life.

Re:could just be the beginning

[Tablizer]

But orgs don't want the extra expense of staffing and training between multiple brands, and migrating back and forth between them as needs change, such as a small database growing large and vice versa. One-stop-shopping simplifies all this.

I'd suggest picking no more than two brands for an org: one high-end and one low-end. You'd probably want the option of ACID for both, which makes no-sql solutions a problem.

Re:could just be the beginning

[Kjella]

From my experience, I'd guess that about 90% of Oracle installations do not need Oracle.

True, but many Oracle customers turn into all-Oracle shops because the DBAs claim it's the "enterprise quality" solution that they can get for a pittance more on top of the already expensive contract - not to mention a vested self interest - and the executives see the costs of managing an Oracle database environment and fear that they'll be hit with another huge bill. And that drives a lot of software to support installation on Oracle and so the circle is complete.

I've worked with Oracle and when it works "right" it's a beast. But I often found that in complex SQL it's a system that wants it "the Oracle way" or need handholding with indexes and execution hints even though there's several ways to achieve the same results. MS SQL and PostgreSQL would usually do something roughly right performance-wise as long as it was logically right. It's okay if you would have to make those kinds of optimizations anyway, but a giant pain in the ass when that's not a priority.

Re:could just be the beginning

[lrichardson]

From my experience, I'd guess that about 90% of Oracle installations do not need Oracle.

I'll go one step beyond that: in my experience, 99% of Oracle installations could be replaced with SQLite, MySQL, Firebird, even Derby. (Possibly Excel, in some cases)

Virtually every Oracle DB I have encountered has used the POWER of Oracle (TM) as an excuse to skip putting together a decent schema. Massive duplication of data. Joining dozens of tables to get commonly needed data. Tables with far too many fields.

I'm currently dealing with one that works ... just. Minor issues, like virtually every table can be dropped by at least two orders of magnitude in size, the actual Oracle DB supporting the application uses ~140 tables, when it needs ten, and there is a ton of data stored that is inaccessible using keys.

If management would spend a fraction of the amount they spend on hardware on a decent DBA, then they wouldn't need to spend millions on monstrosities (in terms of overkill) like Oracle ... and the hardware needed to run it on. Have similar feelings towards Hadoop ... yay, it's great, sooo scalable! Whadya mean I can't get the data out in a usable form?

I do have a grudging respect for OLAP, in one regard: put together a decent schema, and new elements can generally be added by inserting one row in the description table, not by changing the schema itself. That really lends itself to ease of maintenance. But, again, it does require *some* up-front design work.

Oracle != Mongo

[Anonymous Coward]

Mongo is NoSQL, it can't steal business from Oracle at a large scale as processes would have to be completely rewritten to use NoSQL....and while the new setup would be faster, it would lose transactions and the ability to roll them back, among other things.

Re: Oracle != Mongo

[_merlin]

You've completely sidestepped the question. How do you use MongoDB and its ilk in situations where you need transactional consistency or fine-grained permissions?

If all you have is a hammer...

[snookiex]

I won't waste my time explaining you why NoSQL databases are suitable for many use cases, as your post is a flamebait, however I will tell you a bit about my first-hand experience: I'm one of the developers of an open source network inventory application (see signature). We started out by using an RDBMS, but one of the requirements was to provide a dynamic data model and using a schema-based database brought a lot of problems to the table. Besides, we quickly realized that the best way to represent a network was not a bunch of linked tables but a graph, because that's what telecommunications networks are.

Re:If all you have is a hammer...

[haruchai]

I think I'll give your Kuwaiba a try. I've started to evaluate some network inventory applications for when we ditch our managed services provider in the next couple years

Re:If all you have is a hammer...

[fluffernutter]

NoSQL to me seems both uninteresting and interesting at the same time. I big part of the justification of it seems to be 'I don't want to work with a set schema'. Well there are reasons for schemas and just because it seems inconvenient, usually that isn't a reason for doing anything and you're just going to shoot yourself in the foot in the end. The more interesting aspect of nosql is its ability to easily scale across multiple small servers, as opposed to an SQL database which is a big monolithic thing that is difficult to separate.

Re:If all you have is a hammer...

[Parker Lewis]

using a schema-based database brought a lot of problems to the table

Pun intended?

Guess that means they really are Web Scale!!

[haruchai]

  https://www.youtube.com/watch?...

Re:Why not postgres?

[Voyager529]

OK, I'm not a DBA (IANADBA? Hmm, I like the sound of that, 'yanadba', which syllable to put the accent on though.)

But, really, why do corporations not use postgres? Is it some inherit deficiency in the product? A general antipathy to Open Source? Lack of publicity and marketing on the part of Postgres? Nobody from the company to hold the customer's hand when they first get it? (In that case, maybe there needs to be a Red Hat Postgres) Or something else?

Honestly, it may be a bit of everything, but I'll share my own personal anecdote...

I used to provide desktop and network support to a relatively small insurance company (about a hundred employees). Their line-of-business application that handled everything from claims to brokers to billing was coded and maintained by a firm who built the product on an Oracle backend. It's a relatively small Oracle installation - $90,000/year was the number I remember hearing, which is obviously peanuts for an Oracle install.

The coding company was mortified of change. Virtualizing the old Solaris boxes was months of meetings and performance metrics ("virtualized servers will be slower" - we were migrating from servers old enough to still have "Sun Microsystems" logos on them to a high end Poweredge blade chassis...). The people coding for it knew Oracle and Java, and genuinely believed that "write once, run everywhere" was a promise that Java delivered - it took over a year to convince them that "running in IE compatibility mode was not a long-term solution"; the product was indeed written for IE6, though to be fair they did indeed update it to run properly in Webkit and newer iterations of Trident.

Checkbox Compliance is a very, very powerful thing. No matter how much PostrgeSQL and EnterpriseDB promises to be a drop-in replacement, and no matter how true that actually is, the fact that it isn't "Oracle" scared the crap out of the developers and the people who supported them. MariaDB was able to overcome this to an extent because it was literally the exact same code at the beginning, and answered to the same SQL commands (still using 'mysql' where appropriate). The first time something goes wrong, no matter whose fault it is, it's blamed on the recent change. It very well could be - and probably is - someone else's fault...but when you're dealing with management - even remotely competent management - and saying that something they asked for isn't going to work right because Postgres != Oracle, or the upstream vendor keeps saying "we don't support PostgreSQL" no matter what the issue is (...the website isn't compatible with Edge...), migrating to Postgres becomes a headache for 100% PEBKAC reasons.

Isn't it web scale

[goombah99]

I heard MongoDB is webscale.