Tuesday, January 20, 2015

Go: Import cycle not allowed

Level: 1 where 1 is noob and 5 is totally awesome
System: Go

Among the programming languages I use, Go is one of my favorites. Intentionally Go lacks features such as generics, inheritance and some others features, which people might see as standards in a programming language. Some might see this as weaknesses in the language, and some might see it as strengths. In the end, the purpose is the keep the language simple, and to build software with less complex code.

A thing which can be challenging in Go, is object parent-child relations, where a child knows it's parent. Like this code below tries, and fails with an 'Import cycle not allowed' error:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
package child

import "Parent"

type Child struct {
  parent *Parent
}

func (child *Child) PrintParentMessage() {
  child.parent.PrintMessage()
}

func NewChild(parent *Parent) *Child {
  return &Child{parent: parent }
}

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
package parent

import (
  "fmt"
  "child"
)

type Parent struct {
  message string
}

func (parent *Parent) PrintMessage() {
  fmt.Println(parent.message)
}

func (parent *Parent) CreateNewChild() *child.Child {
  return child.NewChild(parent)
}

func NewParent() *Parent {
  return &Parent{message: "Hello World"}
}

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
package main

import (
  "parent"
)

func main() {
  p := parent.NewParent()
  c := p.CreateNewChild()
  c.PrintParentMessage()
}

Cross-refering packages are not allowed in Go. The best thing would be, if Parent could keep track of the children. Then there would be no issues, and some might argue the code would be more clean. But the world is not perfect, and sometimes circumstances enforces design decisions like this.

To make things work, we can use the nice duck typing feature(interfaces) which Go supports. In general, it is good practice to use interfaces. It keeps the code de-couplet and flexible.

So if we make the interface IParent and use it, then everything works out:

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
package child

type IParent interface {
  PrintMessage()
}

type Child struct {
  parent IParent
}

func (child *Child) PrintParentMessage() {
  child.parent.PrintMessage()
}

func NewChild(parent IParent) *Child {
  return &Child{parent: parent }
}

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
package parent

import (
  "fmt"
  "child"
)

type Parent struct {
  message string
}

func (parent *Parent) PrintMessage() {
  fmt.Println(parent.message)
}

func (parent *Parent) CreateNewChild() *child.Child {
  return child.NewChild(child.IParent(parent))
}

func NewParent() *Parent {
  return &Parent{message: "Hello World"}
}

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
package main

import (
  "parent"
)

func main() {
  p := parent.NewParent()
  c := p.CreateNewChild()
  c.PrintParentMessage()
}

Now 'Hello World' is written to output, as intended. Of course, the interface can be placed in a 3rd package. As always, it depends...

Saturday, November 15, 2014

Balancing resources in SQL Server

Level: 3 where 1 is noob and 5 is totally awesome
System: SQL Server

As software developer I believe in decoupling, and I'm certain that decoupling is just as important in database design as well. I really like having things in minimum chunks, and keeping them independent as possible. Independence and decoupling equals scalability. This post is based on a professional experience, which supports my belief.

At my work we are developing a billing platform, and this platform has several clients. Each client use our platform to bill a certain amount of customers. So this set the scene.

Some time ago, we had one database for each client in one instance, which is a easy pattern. But in SQL Server it can easily be an expensive anti-pattern, and it is certainly a pattern which is my last choice. The problem with this pattern, it is only effective, when the databases is equal regarding size and usage.

The buffer pool


The explanation is, there is only one buffer pool pr. instance. SQL Server is an in-memory database, but in most cases, there is not enough memory for a whole database in the memory. And if there is more databases in an instance, then it is almost certain there no room in the memory for all the bases. To solve this issue, SQL server has the buffer pool, which is a cache. When querying, SQL Server will get the result from the buffer pool. If it not available in the buffer pool, it reads from the storage to the buffer pool. If there if no room in the buffer pool, it will flush some of the buffer pool, to make room for the requested data. Everything has to go through the buffer pool.

The process of getting data from storage to the bufferpool, generates PAGEIOLATCH waits. A high count of PAGEIOLATCH waits can be a sign of much data is being loaded from disk to the buffer pool.

Let us say, we have 3 databases in one instance. They respectively supports 25.000, 150.000 and 300.000 customers. The sizes of the databases reflects the number of customer, because more customers generates more data. What will happen, is the database with the most usage, will push out the data from the other bases, out of the buffer pool. This will give the lesser bases more read from the disk, producing more PAGEIOLATCH wait, adding more latency to the lesser bases. Actually, the smallest databases will suffer the most.

The poor solution  


I'm tempted to call this, the poormans solution, but it can be very expensive. The solution is to extend the buffer pool, with more memory. But there is a physical and economical limit to this. SQL Server 2014 has a new feature called buffer pool extensions, where it is possible to extends the buffer with storage(preferable SSD drives). I guess extending the buffer pool, will only cache more of the database with most uses. Besides, we have tried this feature, and have yet to see some great results. I would really like to hear, if some has used this feature with success.

The better solution


A easy way to balance databases resourcing over an instance, is by using the resource governor feature in sql server. It works excellent, but it is really a doubled edged sword. Used wrong it can really downgrade performance. It would also require some nursing, which I prefer to keep to a minimum.

The best solution


I find the best solution is to have more instances. If I have 3 clients, I will assign an instance to each, make it thier domain. Each instance, should have hard assigned a certain amount of memory, and their databases would never be able to inflict each others performance. 

This was the solution we did end up with, and it gave us better performance overall. 

  


Wednesday, October 22, 2014

Smart and Useful are not implicit the same

This special topic is something, I do philosophy on often. Why do we see, the smartest solution looses to the less smart solution? Like why did Betamax loose to VHS? Amiga to PC? HD-DVD to Blu-ray? And so on. My conclusion is; people in generel did find the looser less useful, no matter how smart they where compared to the winner.

Our job as developers(I assume you are a developer), no matter sort of developer, is to make the life easier for somebody or some, usually in an organization. At least, making life easier a.k.a. creating value, should be rule number one, for every developer.

I believe most developers understand the business domain, for the job they are hired to do, but I don't believe it is enough. By knowing the business domain, a smart solution can be developed, but is it useful? And what are the difference of smart and useful? I can state it, by making an example of financial data which need to be presented for a CEO.

Let's say the financial data is produced with the finest practices. The code behind is clean code, full maintainable and maybe even very innovative so the data can be produced in half the time as normal. The numbers are spot on error free. The data is presented in XML, so everybody with an understanding of XSLT, can transform it any way the like. It is smart!!! At least from a developer's point of view. But is it useful from a CEO's point of view. I don't think so. For being useful to, the data could have been sent, as an Excel file to the CEO's inbox. Everyone in such position know some degree of Excel, and done completely useful, the CEO should be able to present the report for his peers without further changes. XML and transformation should not be considered CEO knowledge, but Excel could.

My point is, to develop something smart and useful, you should not only know the business domain, you should also know the recipient(s) of your solution.

A good way to know a recipient is to talk with the recipient, and recognise the person's workflow. See if the solution you create, can be easily integrated into the recipient's workflow with little impact as possible, while still solve the recipient's problem and make the life easier for this person. That would be useful. Thing you could consider in advance, could be:


  • If your recipient, are using a coperate computer, then avoid things which means something has to be installed. Sometimes installation of components, is not even possible. Often components often involves regular involvement from support. Try to target you solution to something already on the your recipient(s) computer. E.g. Microsoft office. It will give you less fuzz and frustration.

  • If possible and depending of the type of the recipient, try to push the result of your solution to the recipient. As an example, again consider a report to a CEO. He shouldn't use time on digging out a report from a system you build, it should just be mailed to him. It would fit his busy workflow better. What would be even worse, would be if the system didn't use his OS credentials for logging in as single sign on, but if he had to use a 2nd credentials for this. Then the digging would be much harder. It would have even more negative impact on his workflow.

  • Optimise your result if possible. A report example again, instead of pages full of numbers, try to optimise with graphs in the start of the report. This would point out the important faster. Numbers could be added as appendix if details are needed. Optimise the report, so it is ready for be shown at meetings. CEO it save time, his life is more easy. 

  • Maybe most importantly. As a developer you are often a minority, compared to the number of recipients of your solution. They very seldom have the knowledge and technical insight on your solution as you. They will see your solution from another perspective, and you should respect this and learn from it. It is the key to make your solution more useful.

  • The purpose of education of recipients is to have a win win situation, you have less to do while your recipients would get more value. But educating people is hard, and choose you battles wisely here. E.g. Learning business people reading XML, compared to learning them a new functionality in Excel is tedious. You will often find the urge to learn people something, when they don't have time for it. E.g. When a business man who really need some data, and you have it in XML. The issues here are, his mind is on getting data, not on learning XML, He can't speak XML with his peers, He is not interested in XML at all and so on. On the other, if you extend a tool he already knows or show a feature, so you can provide him silent with XML. Then he would properly be more keen to learn about the tool, and then the education would both help you and him.  
It was few example, but is easy to come up with more.

My examples has primarily been with reporting, but this can be transferred to what ever. The golden rule is basically to know you audience and their habits. 

Thursday, October 2, 2014

SQL Server, SSIS or SSAS: Fastest and easiest way to create a time dimension

Level: 2 where 1 is noob and 5 is totally awesome
System: SQL Server 2005 and above

One of the most vital components of a dataware house, is the time dimension. It also seems to be the part which people to have most difficulties to generate. I seen all from black magic in MS Excel to big complex SQL scripts and even fancy SSIS solutions, just to generate time dimensions.

I do believe, I have found a better solution.

My solution is this script, I have written(inspired from a script I did found on Stack Overflow). You set a start and end date, plus the granularity. My recommendation is to generate times for at least a decade. A decade with the granularity of one hour, takes less than 5 Mb of data storage and 87601 records with this script as default. A decade is generated in a few seconds.


 DECLARE @StartDate DATETIME = '2010-01-01';   
 DECLARE @EndDate DATETIME = '2019-12-31';   
   
 WITH DateTimeGenerator AS  
 (  
  SELECT @StartDate AS DateAndTime  
  UNION ALL  
  SELECT DATEADD(HOUR, 1, DateAndTime) -- Change datepart to ajust granularity, see http://msdn.microsoft.com/en-us/library/ms186819.aspx for more  
  FROM  DateTimeGenerator    
  WHERE  DateAndTime + 1 < @EndDate  
 )  
 SELECT ROW_NUMBER() OVER (ORDER BY DateAndTime) As Id,   
 DateAndTime,   
 DATEPART(HOUR, DateAndTime) AS [Hour],  
 DATEPART(DAY, DateAndTime) AS [Day],  
 DATEPART(WEEK, DateAndTime) AS [Week],  
 DATEPART(MONTH, DateAndTime) AS [Month],  
 DATEPART(QUARTER, DateAndTime) AS [Quarter],  
 DATEPART(Year, DateAndTime) AS [Year]  
 INTO DimTime  
 FROM  DateTimeGenerator  
 OPTION (MAXRECURSION 0)  

Enjoy

Tuesday, June 24, 2014

SQL 2014: Getting started with OLTP, OLAP, SSIS, SSAS and SSRS

As passionate developer, I come in touch with many technologies, but only a few of them has been so tough to crack as SSAS and SSRS. While learning SSIS, I felt i was learning a lot about datawarehousing, and I couldn't really imagine there could be more to it. At the same time I knew there must be more to it, as SSAS and SSRS was a part of the SQL Server suite.

Now I know that SSIS, SSAS and SSRS can be translated to data gathering, data modelling and data visualisation.

And I have written this post to give a swift explenation of SSIS, SSAS and SSRS so it might be easier for you to get started.

My dream database setup 


Starting with a greenfield project, is often difficult. There no experiences and no knowledge, thereby there is paved no road, to go by when developing the system. In my opinion, the 1st mistake which the most make at this point, is to put it all data in same database. This leads to a monolith system, which is hard to maintain and expand. The 2nd mistake is not to divide the database setup in OLTP and OLAP. This lead to analysis with heavy querying in production and passive historic data also stored in production.

OLTP is an acronym for Online Transaction Processing. In other words, it is the database, which is in the line of business. You can also call it, the database with live data. In my dream setup for at system, i would have a bunch micro services, all supported by their own database. Not necessarily relational databases, I would have the database which fitted the job best. The data in these databases, would be on a journey, which ended in a OLAP system. This journey could span from instant to the lifetime of the system.

OLAP is an acronym for Online analytical Processing. OLAP is almost always, analysed for financial reasons, and therefor it should be the last stop for data. It must not return to OLTP, because it would mean, that history could be re-written which is a problem when it comes to finance.

The process of getting data from OLTP to OLAP, would be data gathering, which is the perfect intro for next topic: SSIS.

SQL Server Integration Services (SSIS)


SSIS has become one of my favourite SQL Server tools. With few drag and drops, it is possible to setup data copy or transformation. Compared to scripting(as a developer I usually prefers scripting), it is so much faster to get something up and running, and also easier to structure. Enough praising.

Referring to OLTP about having several different databases, and SSIS being able to read from different source. It is the perfect tool to support this solution.

DQS (Data Quality Service) and MDS (Master Data Service) are often mention together with SSIS. These tool works only with SQL Server, but if you are all-in with SQL Server these tools are quite good. DQS is for correcting data, as they are gathered. MDS is sort of an authority which distribute a data model through out a system to the connected SQL Server database. It makes sure data integrity is kept.

There is other suppliers of ETL tool for SQL Server, but I only use SSIS.

SQL Server Analysis Services (SSAS)


Even thou data is arranged by stars, and it browsable through excel, it can be better. While a star is easy to imagine, it can be more difficult to imagine an cube, and in the beginning even to realize the difference. A cube is like a dimensional system, as we now it from math. Most of us can imagine 1st, 2nd, 3rd dimensional system. Some of us can even imagine a 4th dimensional system, by collapsing 3 dimensions and add a 4th dimension, so it look 2 dimensional again.

Each dim table in a star, would be a dimension in a cube, but they are still not the same. In principle you could take a star, print out all tables, put the papers on a table and then draw lines for each relation from the dimension tables to the fact table. Where a cube is arranged (as mentioned before) an a virtual dimensional system, it is really not printable, but fast for analysing/browsing.

SSAS Helps you building cubes, so that would be data modelling. It will also assist you to do some data mining. I'll will refrain to speak of data mining with SSAS, because I haven't touch that topic yet.

Working with cubes, is a way in SSAS which called dimensional mode. SSAS has two more modes: PowerPivot and Tabular. It is other ways to model data, and which one is best? It depends. 

But with there is no point in having data modelled, if you can't deliver their information in a readable way.

SQL Server Reporting Services (SSRS)


As you might have guessed SSRS is the visualisation part of the SQL Server suite. There not much to say. It easy to design reports with the editor, you can render reports to various formats and make them available from a reporting server or a share point site.


Take advantages of having SSIS, SSAS and SSRS available


SSIS, SSAS and SSRS are available from SQL Server standard, BI and Enterprise. Not using these tools, when having SQL Server licenses which includes them is as failure. Often have I seen, and have properly in past done myself, custom made reports and based on some analysis directly from the production environment. 

Nobody finds it funny to create custom reports and nobody finds it funny to maintain them. So using SSIS, SSAS and SSRS is a win-win, because it would mean full value from licenses, and developers would use less time on reports and more time on funny stuff.


Getting started 


Getting started with SSAS and SSRS, was a bit hard for me, but fortunately I did stumble in the book: Microsoft SQL Server 2014 Business Intelligence Development: Beginner’s Guide by Raza Red. It is real page turner, which is nice, if you reads a lot of books per year.

It is very comprehensive and filled examples, with good explanations, which I like a lot.

Saturday, April 5, 2014

SQL Server 2014: SQL Server Data Tools(SSDT) and Visual Studio 2013 Challenges in Training Kit 70-463 and in general

Level: 2 where 1 is noob and 5 is totally awesome
System: SQL Server 2014

Notice: I'm using the SQL Server 2014 Developer edition, so if you are using a different edition, things might be different.

Some History


I'm currently studying for the exam 70-463 Implementing a Data Warehouse with Microsoft SQL Server 2012, and I'm using the 70-463 training kit for the exam. I have decided to use SQL Server 2014 for the practical training(regarding to SSDT, I now know it is the best to use SQL Server 2014). It has just been released and regarding to exam 70-463 there is no change. People who took the exam for SQL Server 2012 is also certified for SQL Server 2014. For more see FAQ here at http://www.microsoft.com/learning/en-us/sql-certification.aspx.

When using the training kit, you sooner or later realize, it has become a bit outdated. Which is quite natural, because it was published 25. Dec. 2012, and lot have happened since. There has been new releases and updates of Visual Studio, and we even have a new version of SQL Server. Which, all in all is nice.

The first sign of you might be challenged when using the training kit with SQL Server 2014, is in the beginning. The exam kit provides a list of features you have to install, and last on this list is SQL Server Data Tools, also known as SSDT. As you might discover, this feature is not shipped with SQL Server 2014, and it is the most important component for SSIS and the exam 70-463.

SSDT is a Visual Studio plugin, for developing SQL Server Integration Service(SSIS) packages. SSDT is a replacement from Business Intelligence Development Studio, also known as BIDS.Among things, SSDT delivers SSIS packages templates to Visual Studio. SSIS 2012 depends on Visual Studio 2010, Like SSIS 2014 does, but SQL Server 2012 installs the Visual Studio 2010 IDE if none is present. SQL Server 2014 has failed to do so, everytime I have tried to install it. If there is no SSDT in the SQL Server 2014 installation, it might make sense. Also it is okay, because I prefers to use Visual Studio 2013, and if you are a developer like me, I guess you prefer the same.

And Now the Tricky Part


I discovered Microsoft had developed SSDT (SSDT-BI which it is called now) for Visual Studio 2012, it the most easy package to find, and I thought it was the latest. I installed it, and was hit hard in chapter 3 of the training kit 70-463. Because I has installed SSDT-BI 2012 for Visual Studio 2012.

The problem was I have created a SSIS package with the SQL Server Import and Export tool, the version which comes with SQL Server 2014. It stores it SSIS Packages in the SSIS Package format version 8, while the SSDT 2012 stores in the SSIS Package format version 6. The SSIS Package format is apparently not forward compatible. 

The pain I experienced, in chapter 3 of the training kit, was I had to create a package with the SQL Server Import and Export tool (version 2014 in my case) and add it to a SSIS Project created with SSDT 2012. The because the version numbers didn't match, result was this:



Fortunately I succeeded to find SSDT-BI 2014 well hidden at 

It saves SSIS Packages in SSIS Package format version 8, like the rest of the SQL Server 2014 Suite, plus it works with Visual Studio 2013. I'll expect it will work it with the rest of the training kit 70-463, or else I'll provide a solution on this blog. As bonus info, you have to select "Perform a new installation of SQL Server 2014", or you will get some kind of architecture error. Despite the option name, it only installs SSTD-BI 2014.


Enjoy




 

Tuesday, March 18, 2014

Migrating from HaveBox 1.6.1 to 2.0.0

It's here, the lastest version of HaveBox. HaveBox 2.0.0. A bump in the major vision number means there are some breaking changes, when migrating from a prior version of HaveBox.

The breaking changes was needed, to get HaveBox in the direction I wanted. My goal is still to keep HaveBox simple, it doesn't mean less features in the future, but on the other hand I will not bloat HaveBox with features. HaveBox 2.0.0 now consist of a core with the basic features build in. Auxiliary features like scanners, config injection or xml configuration, will be added via sub-configs. Sub-config is the key word to extend HaveBox from now on. It should now be possible for everybody, to write a sub-config to customize HaveBox.

The Breaking Changes


Scanners

I have removed the scanner concept from config. I deprecated scanners in version 1.6.0, and from now on, scanner has to be defined sub-configs. HaveBox still comes with a scanner called SimpleScanner, but now it is a sub-config which have to be merged in to the config. See more in the documentation: HaveBox Scanners

Config injection

Config injection has to be enabled by merging in the Config Injection sub-config into the config.  See more in the documentation: Config Injection

Spawning Child Containers

The Spawn methods has been removed. Now you have to configure the life-time, of the singletons in the config section and call CreateChildContainer to create a child container. In this way, it is possible to combine the functionality of the spawn methods. See more in the documentation: Custom life-times

The New Features


TryGetInstance

It now possible to try-getting an instance. It is the same try-get functionality known from the rest of the .NET frameworks. If it succeeds the method returns true, and there will be an instance in the out parameter, else it return false, and there will be NULL in the out parameter.

How About Performance?


HaveBox 2.0.0 has also been optimized for performance. The ilasm has been cleaned and reduced, plus I have written a hashtable to replace Dictionary. The new hashtable is almost twice as fast compared to Dictionary(which also is a hashtable, but less optimal written, check the .NET reference sources.).

The concept app, is showing some nice improvement in performance, but I urge you download HaveBox and do your own testing.

That's all for now


HaveBox 2.0.0 can be downloaded from Nuget: http://www.nuget.org/packages/HaveBox/
The project page: https://bitbucket.org/Have/havebox/downloads

And don't forget to checkout HaveBoxJSON and HaveBoxStates.

Enjoy