Closed. This question needs to be more focused. It is not currently accepting answers.
Want to improve this question? Update the question so it focuses on one problem only by editing this post.
Closed 5 years ago.
Improve this question
I need please some clarifications about the BI architecture. According to what I understood, the first step is to gather data from different data sources, clean it, and load it to a data warehouse through an ETL. The Data Schema of the datawarehouse shouldn't be relational, and should support fast business operations (ex. Star schema), then finally we have some reporting tools such as qlick, Tableau ...etc. My question is, what is OLAP and in which step does it come to existence?
thx,
OLAP = online analytical processing, which usually means 'cube' which is usually about reporting at various summaries
This is in contrast to OLTP = online transactional processing which usually refers to a system (usually stored in a relational database) that does a high volume of reads and writes at a detailed level
A cube represents things to users as facts and dimensions.
A data warehouse star schema also represents things as facts and dimensions. In a datawarehouse star schema (which is relational but is not normalised), these are stored in tables
To get a 'grand total' out of a star schema you write a SQL query that runs against the database and adds up all the detail level data into a grand total. Sometimes this takes time
To get a 'grand total' out of a cube (OLAP) you drag and drop the dimensions and measures you want (you usually use a client tool to analyse a cube) and the answer appears much faster because a cube is generally optimised for summaries, (i.e. it usually has summaries pre saved in it, and the storage mechanism is optimised for generating summaries)
A cube is usually built from a star schema but doesn't have to be - it just makes it a lot easier to build it if it is
are'nt Olap cubes represented by the data model in warehouse (star schema for ex.)?
Yes they are represented but they are different things. One stores data in a database. One stores data in a cube. A cube is usually loaded from data, usually from a database.
Related
Closed. This question is opinion-based. It is not currently accepting answers.
Want to improve this question? Update the question so it can be answered with facts and citations by editing this post.
Closed last month.
Improve this question
I have a very large monolithic legacy application that I am tasked with breaking into many context-bounded applications on a different architecture. My management is pushing for the old and new applications to work in tandem until all of the legacy functionality has been migrated to the current architecture.
Unfortunately, as is the case with many monolithic applications, this one maintains a very large set of state data for each user interaction and it must be maintained as the user progresses through the functionality.
My question is what are some ways that I can satisfy a hybrid legacy/non-legacy architecture responsibly so that in the future state all new individual applications are hopelessly dependent on this shared state model?
My initial thought is to write the state data to a cache of some sort that is accessible to both the legacy application and the new applications so that they may work in harmony until the new applications have the infrastructure necessary to operate independently. I'm very skeptical about this approach so I'd love some feedback or new ways of looking at the problem.
Whenever I've dealt with this situation I take the dual writes approach to the data as it mostly a data migration problem. As you split out each piece of functionality you are effectively going to have two data models until the legacy model is completely deprecated. The basic steps for this are:
Once you split out a component start writing the data to both the old and new database.
Backfill the new database with anything you need from the old.
Verify both have the same data.
Change everything that relies on this part of the data to read from the new component/database.
Change everything that relies on this part of the data to write to the new component/database.
Deprecate that data in old database, i.,e. back it up then remove it. This will confirm that you've migrated that chunk.
The advantage is there should no data loss or loss of functionality and you have time to test out each data model you've chosen for a component to see if it works with the application flow. Slicing up a monolith can be tricky deciding where your bounded contexts lie is critical and there's no perfect science to it. Always keep in mind where you need your application to scale and which pieces are required to perform.
Closed. This question needs to be more focused. It is not currently accepting answers.
Want to improve this question? Update the question so it focuses on one problem only by editing this post.
Closed 5 years ago.
Improve this question
This is very elementary question but why does a framework like Rails use ActiveRecord to run SQL commands to get data from a DB? I heard that you can cached data on the Rails server itself, so why not just store all data on the the server instead of the DB? Is it because space on the server is a lot more expensive/valuable than on the DB? If so, why is that? Also can the reason be that you want a ORM in the DB and that just takes too much code to set up on the Rails server? Sorry if this question sounds dumb but I don't know where else I can go for an answer.
What if some other program/person wants to access this data and for some reason cannot use your rails application? What if in future you decide to discontinue using rails and decide to go with some other technology for front end but want to keep the data? In these cases having a separate database helps. Also could you run complex join queries on cached data on Rail Server?
databases hold a substantial amount of advantages against other types of databases. Some of them are listed below:
Data integrity is maximised and data redundancy is minimised, as
the single storing place of all the data also implies that a given
set of data only has one primary record. This aids in the maintaining
of data as accurate and as consistent as possible and enhances data
reliability.
Generally bigger data security, as the single data storage location
implies only a one possible place from which the database can be
attacked and sets of data can be stolen or tampered with.
Better data preservation than other types of databases due to
often-included fault-tolerant setup.
Easier for using by the end-user due to the simplicity of having a
single database design.
Generally easier data portability and database administration. More
cost effective than other types of database systems as labour, power
supply and maintenance costs are all minimised.
Data kept in the same location is easier to be changed, re-organised,
mirrored, or analysed.
All the information can be accessed at the same time from the same
location.
Updates to any given set of data are immediately received by every
end-user.
Closed. This question does not meet Stack Overflow guidelines. It is not currently accepting answers.
We don’t allow questions seeking recommendations for books, tools, software libraries, and more. You can edit the question so it can be answered with facts and citations.
Closed 7 years ago.
Improve this question
Using Delphi XE2: I have used AbsoluteDB for years with good success for smallish needs, but it does not scale well to large datasets. I need recommendations for a DB engine for the following conditions:
Large (multi-gigabyte) dataset, few tables with lots of small records. This is industrial-equipment historical data; most tables have new records written once a minute with a device ID, date, time and status; a couple tables have these records w/ a single data point per record, three others have 10 to 28 data points per record depending on the device type. One of the single-data-point tables adds events asynchronously and might have a dozen or more entries per minute. All of this has to be kept available for up to a year. Data is usually accessed by device ID(s) and date window.
Multi-user. A system service retrieves the data from the devices, but the trending display is a separate application and may be on a separate computer.
Fast. Able to pull any 48-hour cluster of data in at most a half-dozen seconds.
Not embedded.
Single-file if possible. Critical that backups and restores can by done programatically. Replication support would be nice but not required.
Can be integrated into our existing InstallAware packages, without user intervention in the install process.
Critical: no per-install licenses. I'm fine with buying per-seat developer licenses, but we're an industrial-equipment company, not a software company - we're not set up for keeping track of that sort of thing.
Thanks in advance!
I would use
either PostgreSQL (more proven than MySQL for such huge data)
or MongoDB
The main criteria is what you would do with the data. And you did not say much about that. Would you do individual queries by data point? Would you need to do aggregates (sum/average...) over data points per type? If "Data is usually accessed by device ID(s) and date window", then I would perhaps not store the data in individual rows, one row per data point, but gather data within "aggregates", i.e. objects or arrays stored in a "document" column.
You may store those aggregates as BLOB, but it may be not efficient. Both PostgreSQL and MongoDB have powerful objects and arrays functions, including indexes within the documents.
Don't start from the DB, but start from your logic: which data are you gathering? how is it acquired? how is it later on accessed? Then design high level objects, and let your DB store your objects in an efficient way.
Also consider the CQRS pattern: it is a good idea to store your data in several places, in several layouts, and make a clear distinction between writes (Commands) and reads (Queries). For instance, you may send all individual data points in a database, but gather the information, in a ready-to-use form, in other databases. Don't hesitate to duplicate the data! Don't rely on a single-database-centric approach! This is IMHO the key for fast queries - and what all BigData companies do.
Our Open Source mORMot framework is ideal for such process. I'm currently working on a project gathering information in real time from thousands of remote devices connected via Internet (alarm panels, in fact), then consolidating this data in a farm of servers. We use SQLite3 for local storage on each node (up to some GB), and consolidate the data in MongoDB servers. All the logic is written in high-level Delphi objects, and the framework does all the need plumbing (including real-time replication, and callbacks).
Closed. This question is opinion-based. It is not currently accepting answers.
Want to improve this question? Update the question so it can be answered with facts and citations by editing this post.
Closed 8 years ago.
Improve this question
I'm using Core Data in my app and I have many "categories" of data. Some are smaller and some are bigger, and I thought to divide some of the categories to different entities.
So I wanted to ask, if there are any advantages or disadvanteges to using multiple entities even if not mandatory, and if I should also create an entity for smaller "categories" of data? Thanks!
Here is the rule that I use:
If incorporating data into the entity itself with attributes does not result in excessively repetitive data, I would prefer to add attributes rather than new entities.
This is a subtle tradeoff in which you have to consider
Performance
Complexity of the data model
Code legibility.
If you consider these factors carefully with the guidance above, I am sure you can make good decisions on when to create new entities rather than using attributes.
For example, if you have an entity like Story that has attributes like title, text, date, etc. and would like to add a category, in most cases it would make sense to create a to-one or even to-many relationship to a Category entity rather than using a string attribute. Presumably, there will be hundreds of stories and dozens of categories, and the flexibility to have more than one category is a definite advantage.
On the other hand, if you have an entity Story which is always one of not more than three types, i.e. "report", "analysis" or "opinion", you would be better off with an enum type of attribute rather than a relationship to a new entity.
I'm not overly sure what you mean by categories but, obviously the bigger the entity the more memory your app will take up when you load the entities into your NSManagedObjectContext. However there's no real point splitting entities up into two when you're just going to load them both anyway.
In terms of performance Core Data docs mention that even 10,000 objects is a pretty small database. Just be careful with BLOBs, dont load data you dont need into memory and release data when you can using NSManagedObjectContext's reset method.
Closed. This question needs to be more focused. It is not currently accepting answers.
Want to improve this question? Update the question so it focuses on one problem only by editing this post.
Closed 9 years ago.
Improve this question
I am working on a project to implement an historian.
I can't really find a difference between an historian and a data warehouse.
Any details would be useful.
Data Historian
Data historians are groups of tables within a database that store historical information about a process or information system.
Data historians are used to keep historical data regarding a manufacturing system. This data can be changes in state of a data point, current values and summary data for these points. Usually this data comes from automated systems like PLCs, DCS or other process controlling system. However some historian data can be human entered.
There are several historians available for commercial use. However, one of the most common historians have tended to be custom developed. The commercial versions would be products like OsiSoft’s PI or GE’s Data Historian.
Some examples of data that could be stored in a data historian are items (or tags) like:
- Total products manufactured for the day
-Total defects created on a particular crew shift
-Current temperature of a motor on the production line
-Set point for the maximum allowable value being monitored by another tag
-Current speed of a conveyor
-Maximum flow rate of a pump over a period of time
-Human entered marker showing a manual event occured
-Total amount of a chemical added to a tank
These items are some of the important data tags that might be captured. However, once captured the next step is in presentation or reporting of that data. This is where the work of analysis is of great importance. The data/time stamp of one tag can have a huge correlation to another/other tag(s). Carefully storing this in the historians’ database is critical to good reporting.
The retrieval of data stored in a data historian is the slowest part of the system to be implemented. Many companies do a great job of putting data into a historian, but then do not go back and retrieve any of the data. Many times this author has gone into a site that claims to have a historian only to find that the data is “in there somewhere”, but has never had a report run against the data to validate the accuracy of the data.
The rule-of-thumb should be to provide feedback on any of the tags entered as soon as possible after storage into the historian. Reporting on the first few entries of a newly added tag is important, but ongoing review is important too. Once the data is incorporated into both a detailed listing and a summarized list the data can be reviewed for accuracy by operations personnel on a regular basis.
This regular review process by the operational personnel is very important. The finest data gathering systems that might historically archive millions of data points will be of little value to anyone if the data is not reviewed for accuracy by those that are experts in that information.
Data Warehouse
Data warehousing combines data from multiple, usually varied, sources into one comprehensive and easily manipulated database. Different methods can then be used by a company or organization to access this data for a wide range of purposes. Analysis can be performed to determine trends over time and to create plans based on this information. Smaller companies often use more limited formats to analyze more precise or smaller data sets, though warehousing can also utilize these methods.
Accessing Data Through Warehousing
Common methods for accessing systems of data warehousing include queries, reporting, and analysis. Because warehousing creates one database, the number of sources can be nearly limitless, provided the system can handle the volume. The final result, however, is homogeneous data, which can be more easily manipulated. System queries are used to gain information from a warehouse and this creates a report for analysis.
Uses for Data Warehouses
Companies commonly use data warehousing to analyze trends over time. They might use it to view day-to-day operations, but its primary function is often strategic planning based on long-term data overviews. From such reports, companies make business models, forecasts, and other projections. Routinely, because the data stored in data warehouses is intended to provide more overview-like reporting, the data is read-only.