DB2 is a family of relational database software solutions offered by IBM. It includes standard Db2 and Db2 Warehouse editions, either deployable on-cloud, or on-premise.
$0
TensorFlow
Score 7.7 out of 10
N/A
TensorFlow is an open-source machine learning software library for numerical computation using data flow graphs. It was originally developed by Google.
I have primarily used it as the basis for a SIS - but I have migrated more than a few systems from there database systems to DB2 (Filemaker, MySQL, etc.). DB2 does have a better structural approach, as opposed to Filemaker, which allows for more data consistency, but this can also lead to an inflexibility that can sometimes be counterintuitive when attempting to compensate for the flexibility of the work environment as Schools tend to have an all in one approach.
TensorFlow is great for most deep learning purposes. This is especially true in two domains: 1. Computer vision: image classification, object detection and image generation via generative adversarial networks 2. Natural language processing: text classification and generation. The good community support often means that a lot of off-the-shelf models can be used to prove a concept or test an idea quickly. That, and Google's promotion of Colab means that ideas can be shared quite freely. Training, visualizing and debugging models is very easy in TensorFlow, compared to other platforms (especially the good old Caffe days). In terms of productionizing, it's a bit of a mixed bag. In our case, most of our feature building is performed via Apache Spark. This means having to convert Parquet (columnar optimized) files to a TensorFlow friendly format i.e., protobufs. The lack of good JVM bindings mean that our projects end up being a mix of Python and Scala. This makes it hard to reuse some of the tooling and support we wrote in Scala. This is where MXNet shines better (though its Scala API could do with more work).
Theano is perhaps a bit faster and eats up less memory than TensorFlow on a given GPU, perhaps due to element-wise ops. Tensorflow wins for multi-GPU and “compilation” time.
The DB2 database is a solid option for our school. We have been on this journey now for 3-4 years so we are still adapting to what it can do. We will renew our use of DB2 because we don’t see. Major need to change. Also, changing a main database in a school environment is a major project, so we’ll avoid that if possible.
You have to be well versed in using the technology, not only from a GUI interface but from a command line interface to successfully use this software to its fullest.
I have never had DB2 go down unexpectedly. It just works solidly every day. When I look at the logs, sometimes DB2 has figured out there was a need to build an index. Instead of waiting for me to do it, the database automatically created the index for me. At my current company, we have had zero issues for the past 8 years. We have upgrade the server 3 times and upgraded the OS each time and the only thing we saw was that DB2 got better and faster. It is simply amazing.
The performances are exceptional if you take care to maintain the database. It is a very powerful tool and at the same time very easy to use. In our installation, we expect a DB machine on the mainframe with access to the database through ODBC connectors directly from branch servers, with fabulous end users experience.
Easily the best product support team. :) Whenever we have questions, they have answered those in a timely manner and we like how they go above and beyond to help.
Community support for TensorFlow is great. There's a huge community that truly loves the platform and there are many examples of development in TensorFlow. Often, when a new good technique is published, there will be a TensorFlow implementation not long after. This makes it quick to ally the latest techniques from academia straight to production-grade systems. Tooling around TensorFlow is also good. TensorBoard has been such a useful tool, I can't imagine how hard it would be to debug a deep neural network gone wrong without TensorBoard.
DB2 was more scalable and easily configurable than other products we evaluated and short listed in terms of functionality and pricing. IBM also had a good demo on premise and provided us a sandbox experience to test out and play with the product and DB2 at that time came out better than other similar products.
Keras is built on top of TensorFlow, but it is much simpler to use and more Python style friendly, so if you don't want to focus on too many details or control and not focus on some advanced features, Keras is one of the best options, but as far as if you want to dig into more, for sure TensorFlow is the right choice
By using DB2 only to support my IzPCA activities, my knowledge here is somewhat limited.
Anyway, from what I was able to understand, DB2 is extremely scallable.
Maybe the information below could serve as an example of scalability.
Customer have an huge mainframe environment, 13x z15 CECs, around 80 LPARs, and maybe more than 50 Sysplexes (I am not totally sure about this last figure...)
Today we have 7 IzPCA databases, each one in a distinct Syplex.
Plans are underway to have, at the end, an small LPAR, with only one DB2 sub-system, and with only one database, then transmit the data from a lot of other LPARs, and then process all the data in this only one database.
The IzPCA collect process (read the data received, manipulate it, and insert rows in the tables) today is a huge process, demanding many elapsed hours, and lots of CPU.
Almost 100% of the tables are PBR type, insert jobs run in parallel, but in 4 of the 7 database, it is a really a huge and long process.
Combining the INSERTs loads from the 7 databases in only one will be impossible.......,,,,
But, IzPCA recently introduced a new feature, called "Continuous Collector".
By using that feature, small amounts of data will be transmited to the central LPAR at every 5 minutes (or even less), processed immediately,in a short period of time, and withsmall use of CPU, instead of one or two transmissions by day, of very large amounts of data and the corresponding collect jobs occurring only once or twice a day, with long elapsed times, and huge comsumption of CPU
I suspect the total CPU seconds consumed will be more or less the same in both cases, but in the new method it will occur insmall bursts many times a day!!
