Tampilkan postingan dengan label microsoft. Tampilkan semua postingan
Tampilkan postingan dengan label microsoft. Tampilkan semua postingan

Kamis, 08 Januari 2009

Desain Database by Yulius Eka Agung Seputra,ST,MSi

Basis data (database) merupakan kumpulan dari data yang saling berhubungan satu dengan yang lainnya, tersimpan di simpanan luar komputer dan digunakan perangkat lunak tertentu untuk memanipulasinya. Database merupakan salah satu komponen yang penting di sistem informasi, karena berfungsi sebagai basis penyedia informasi bagi para pemakainya. Penerapa database dalam sistem informasi disebut dengan database system. Sistem basis data (database system) ini adalah suatu sistem informasi yang mengintegrasikan kumpulan dari data yang saling berhubungan satu dengan lainnya dan
membuatnya tersedia untuk beberapa aplikasi yang bermacam-macam di dalam suatu organisasi. Tujuan dari desain database adalah untuk menentukan data-data yang
dibutuhkan dalam sistem, sehingga informasi yang dihasilkan dapat terpenuhi dengan baik. Terdapat beberapa alasan mengapa desain database perlu untuk dilakukan, salah satu adalah untuk menghindari pengulangan data.

Adapun metode untuk meminimasi pengulangan data (data redudancy) antara lain dengan :

a. Normalisasi.
b. Dekomposisi lossless.
Diperlukan jika ada indikasi bahwa tabel yang kita buat tidak baik (terjadi pengulangan informasi, potensi inkonsistensi data pada operasi pengubahan, tersembunyinya informasi tertentu) dan diperlukan supaya jika tabel-tabel yang didekomposisi kita gabungkan kembali dapat menghasilkan tabel awal sebelum didekomposisi, sehingga diperoleh tabel yang baik.
c. ERD (Entity Relationship Diagram).
d. Menentukan kardinalitas relasi.

Terdapat beberapa pengertian tentang key sehubungan dengan normalisasi dan ERD, antara lain :

a. Superkey adalah gugus dari sejumlah atribut entiti yang dapat digunakan untuk mengidentifikasi obyek secara unik.
b. Candidate key adalah superkey dengan jumlah atribut minimal dan dapat berdiri sendiri.
c. Primary key adalah superkey yang dipilih oleh desainer atau administrator basis data.

Normalisasi.

Adalah proses yang berkaitan dengan model data relational untuk mengorganisasi himpunan data dengan ketergantungan dan keterkaitan yang tinggi atau erat. Hasil dari proses normalisasi adalah himpunan-himpunan data dalam bentuk normal (normal form). Ada beberapa bentuk normal, yaitu :

a. Bentuk Normal I (First Normal Form / 1-NF).
b. Bentuk Normal II (Second Normal Form / 2-NF).
c. Bentuk Normal III (Third Normal Form / 3-NF).
d. Bentuk Normal IV (Fourth Normal Form / 4-NF).
e. Bentuk Normal Boyce-Codd (Boyce-Codd Normal Form / BCNF).
f. Project-Join Normal I Form (PJNF).
g. Domain-Key Normal I Form (DKNF).
h. Bentuk Normal V (Fifth Normal Form / 5-NF).
Kegunaan normalisasi :

a. Meminimasi pengulangan informasi.
b. Memudahkan indentifikasi entiti / obyek.

Bentuk Normal I (First Normal Form / 1-NF).

Suatu relasi memenuhi 1-NF jika dan hanya jika setiap atribut dari relasi
tersebut hanya memiliki nilai tunggal dalam satu baris atau record.



Tabel 4.1 : Bentuk tidak Unnormalized Form (Non 1-NF table)



Tabel 4.2 : Bentuk 1-NF table

Bentuk Normal II (Second Normal Form / 2-NF).

Suatu relasi memenuhi 2-NF jika dan hanya jika :

a. Memenuhi 1-NF.
b. Setiap atribut yang bukan kunci utama tergantung secara fungsional terhadap semua atribut kunci dan bukan hanya sebagian atribut.

Jika suatu relasi memenuhi 1-NF dan relasi tersebut memiliki tepat satu atribut yang membentuk kunci utama, maka relasi tersebut memenuhi 2-NF.

Rasionalisasi 2-NF :

a. Memiliki semantik yang lebih eksplisit dari 1-NF.
b. Mencegah beberapa kondisi anomali dalam update data.



Tabel 4.3 : Bentuk 2-NF table (satisfying 1-NF).

Ketergantungan fungsional dilakukan untuk :

a. StudentID => Student, BirthDate (SC1).
b. CourseID => Course, Credit (SC2).
c. StudentID, CourseID => Grade (SC3, SC3A).
d. Grade => Weight (SC3B).



Tabel 4.4 : Tabel yang memenuhi 2-NF.



Tabel 4.5 : Tabel yang memenuhi 3-NF.

Akhirnya semua tabel SC1, SC2, SC3A, SC3B berada dalam kondisi 3-NF, sehingga semua databases mengalami kondisi 3-NF.

Bentuk Normal III (Third Normal Form / 1-NF).

Suatu relasi memenuhi bentuk III (3-NF) jika dan hanya jika :

a. Relasi tersebut memenuhi 2-NF.
b. Setiap atribut bukan kunci tidak tergantung secara fungsional kepada atribut bukan kunci yang lain dalam relasi tersebut.

Suatu relasi yang memenuhi 2-Nf dan hanya memiliki satu atribut bukan kunci selalu memenuhi 3-NF.

Bentuk Normal Boyce-Codd (Boyce-Codd Normal Form / BCNF).

Suatu relasi memenuhi BCNF jika dan hanya jika setiap determinan yang ada pada relasi tersebut adalah kunci kandidat (candidate keys). Determinan adalah gugus atribut dimanaa satu atau lebih atribut lain tergantung secara fungsional.

Model Hubungan atau Relasi Entiti (Entity Realtionship (E-R) Model).

Model relasi entiti didasarkan pada persepsi dunia nyata yang terdiri dari himpunan obyek dasar yang disebut entiti dan relasi antar entiti. Entiti adalah obyek yang dapat diidentifikasi secara unik.Entiti dikarakterisasi dan dipresentasikan dengan suatu gugus atribut. Contoh gugus atribut dari entiti PEKERJA adalah nama, tanggal lahir, NIP,golongan/pangkat.Sekelompok entiti yang memiliki karakterisasi entiti disebut gugus entiti(entity set).Setiap entiti dari gugus tersebut disebut anggota gugus (member of set).Contoh gugus entiti adalah gugus entiti pegawai bank, gugus entiti nasabah bank. Dari beberapa gugus tadi mungkin terjadi suatu relasi, misalnya relasi antara gugus bank dengan gugus nasabah bank.Berdasarkan jumlah gugus yang terlibat maka relasi antar entiti dibedakan menjadi :

a. Relasi biner (binary), yaitu relasi antar 2 gugus entiti.
b. Relasi trio (ternary), yaitu relasi antar 3 gugus entiti.
c. Relasi N-ary, yaitu relasi antar n gugus entiti.

Khusus untuk relasi biner maka relasi antar anggota dari dua gugus yang terlibat (kardinalitas relasi biner) dapat bersifat :

a. Relasi 1-1 (one-to-one relationship).
Adalah satu entiti anggota gugus diasosiasikan dengan tepat satu entiti anggota gugus yang lain.



b. Relasi 1-banyak (one-to-many relationship).
Adalah satu entiti anggota gugus diasosiasikan dengan satu atau lebih entiti anggota gugus yang lain. Sebaliknya satu entiti anggota gugus yang lain tersebut diasosiasikan dengan tepat satu entiti anggota gugus pasangannya.



c. Relasi banyak-1 (many-to-one relationship).
Adalah satu entiti anggota gugus diasosiasikan dengan satu atau lebih entiti anggota gugus yang lain dan berlaku pula sebaliknya.



Menterjemahkan ERD ke Tabel



Tipe file.

Database dibentuk dari kumpulan file. File di dalam pemrosesan aplikasi dapat dikategorikan ke dalam beberapa tipe, diantaranya yaitu sebagai berikut :

1. File induk (master file).
Didalam aplikasi, file ini merupakan file yang penting. File ini tetap terus ada selama hidup dari sistem informasi. File induk dapat dibedakan lagi menjadi :

a. File induk acuan (reference master file), yaitu file induk yang recordnya relatif statis, jarang berubah nilainya. Contoh dari file ini adalah file daftar gaji, file daftar matakuliah.
b. File induk dinamik (dynamic master file), yaitu file induk yang nilai dari record-recordnya sering berubah atau sering dimutakhirkan (updated) sebagai akibat dari suatu transaksi. Contoh file ini adalah file induk persediaan, file induk langganan dan lain sebagainya.

2. File transaksi (transaction file).
File transaksi disebut juga dengan nama file input (input file). File ini digunakan untuk merekam data hasil dari suatu transaksi yang terjadi. Misalnya nilai unit suatu barang dapat diketahui dari file induk persediaan. File induk ini hanya menunjukkan status unit akhir dari barang yang dimaksud. Sedang uni akhir ini berasal dari transaksi- transaksi yang pernah terjadi. Untuk melihat transaksi-transaksi yang mempengaruhi nilai di file induk, maka dapat dilihat pada file transaksinya. Contoh file transaksi yang lain adalah file transaksi penjualan yang berisi data tentang transaksi penjualan yang terjadi. Biasanya file transaksi memuat rekaman tanggal dari transaksinya yang menunjukkan kapan transaksi tersebut terjadi.
3. File laporan (report file).
File ini disebut juga dengan file output (output file), yaitu file yang berisi dengan informasi yang akan ditampilkan. File ini dibuat untuk mempersiapkan pembuatan suatu laporan dan biasanya dilakukan bila printer belum siap atau masih digunakan oleh proses yang lain.
4. File sejarah (history file).
File sejarah dibuat judan dengan file arsip (archival file), yaitu file yang berisi dengan data masa lalu yang sudah tidak aktif lagi, tetapi perlu disimpan untuk keperluan mendatang.
5. File pelindung (backup file).
File pelindung merupakan salinan dari file-file yang masih aktif didatabase pada suatu saat tertentu. File ini digunakan sebagai cadangan atau pelindung bila file database yang aktif rusak atau hilang.
6. File kerja (working file).
File kerja disebut juga dengan nama file sementara (temprorary file) atau scratch file. File ini dibuat oleh suatu proses program secara sementara karena memori komputer tidak mecukupi atau untuk menghemat pemakaian memori selama proses dan akan dihapus bila proses telah selesai.
Akses dan organisasi file.

Akses file (access file) adalah suatu metode yang menunjukkan bagaimana suatu program komputer akan membaca record-record dari suatu file.File dapat diakses dengan dua cara yaitu secara urut (sequential access) atau secara langsung (direct access atau random access). Metode akses urut(sequential access method) dilakukan dengan membaca atau menulis suatu record di file dengan membaca terlebih dahulu mulai dari record pertama,urut sampai dengan record yang diinginkan. Metode akses langsung (direct access method) dilakukan dengan cara langung membaca record pada posisinya di file tanpa membaca dari record pertama terlebih dahulu. Organisasi file adalah pengaturan dari record secara logika didalam file dihubungkan satu dengan yang lainnya. File dapat diorganisasikan secara urut (sequential organization) atau secara acak (random organization). Walaupun organisasi file dan pengaksesan file dapat dipandang secara
terpisah, tetapi biasanya pembahasan mengenai organisasi file menyangkut
keduanya, yaitu sebagai berikut :
a. File urut (sequential file) merupakan file dengan organisasi urut
(sequential organization) dengan pengaksesan secara urut (sequential
access).
b. File urut berindeks (indexed sequential file) atau sering disebut dengan
ISAM (indexed sequential access method) merupakan file dengan organisasi
urut (sequential organization) dengan pengaksesan secara langsung
(direct access).
c. File akses langsung (direct access file) atau disebut dengan file alamat
langsung (direct address file) merupakan file dengan organisasi acak
(random organization) dengan pengaksesan langsung (direct access).
Organisasi file seperti ini disebut dengan organisasi file tradisional atau
konvensional, karena telah ada sebelum struktur database dikembangkan.
Organisasi file database dapat berbentuk struktur data berjenjang
(hierarchical data structure), struktur data jaringa (network data structure)
dan struktur data hubungan (relational data structure). Struktur data
hubungan merupakan organisasi file database yang terbaru dan mudah dipahami.
Struktur data hubungan mempunyai karakteristik sebagai berikut :
a. File dalam bentuk tabel yang persis dengan file urut.
b. Hubungan antara record didasarkan pada nilai dari field kunci, bukan
berdasarkan alamat atau pointer.
Struktur data hubungan makin banyak digunakan pada paket-paket DBMS, seperti
misalnya Dbase, Foxbase, Sql dan sebagainya.
Langkah-langkah desain database.

Untuk tahap desain database yang perlu dilakukan adalah mengidentifikasi terlebih dahulu file-file yang diperlukan dalam sistem informasi yang dibangun. File-fila database yang dibutuhkan oleh sistem dapat dilihat pada desain model yang digambarkan dalam bentuk diagram arus data (DFD).
Langkah-langkah desain database secara umum adalah sebagai berikut :

a. Menentukan kebutuhan file database untuk sistem yang baru.
File yang dibutuhkan dapat ditentukan dari DAD sistem baru yang telah dibuat.
b. Menentukan parameter daru file database.
Setelah file-file yang dibutuhkan telah dapat ditentukan, maka parameter dari file selanjutnya juga dapat ditentukan. Parameter tersebut, meliputi:
· Tipe dari file : file induk, file transaksi, file sementara (temporary).
· Media dari file : hardisk, disket, pita magnetik, CD.
· Organisasi dari file : fila sequential, random, berindek.
· Field kunci dari file.

Analis sistem dapat menggunakan formulir berikut untuk mengidentifikasi file database yang akan didesain, sebagai berikut :



Tabel 4.6 : Tabel identifikasi kebutuhan file.



Tabel 4.7 : Tabel identifikasi atribut (field) dalam sebuah file.

Hardware and Software Requirements for Installing SQL Server 2000

The minimum hardware and software requirements for running Microsoft® SQL Server™ 2000 are listed in the following tables.

Hardware Requirements
This table shows hardware requirements for installing Microsoft SQL Server 2000 or SQL Server client management tools and libraries.

Hardware Minimum requirements
Computer Intel® or compatible
Pentium 166 MHz or higher.

Memory (RAM)1 Enterprise Edition and Enterprise Evaluation Edition: 64 MB minimum, 128 MB or more recommended
Standard Edition and Developer Edition: 64 MB minimum

Personal Edition and Desktop Engine (MSDE 2000):

128 MB minimum on Windows XP


64 MB minimum on Windows 2000


32 MB minimum on all other operating systems

Hard disk space2 Enterprise, Enterprise Evaluation, Standard, Developer, and Personal Editions require:
95 to 270 MB of available hard disk space for the database engine; 250 MB for a typical installation.


50 MB of available hard disk space for a minimum installation of Analysis Services; 130 MB for a typical installation.


80 MB of available hard disk space for English Query.
Desktop Engine (MSDE 2000): 44 MB minimum

Monitor VGA or higher resolution
800x600 or higher resolution required for the SQL Server graphical tools

Pointing device Microsoft Mouse or compatible pointing device
CD-ROM drive Required


1 Additional memory may be required, depending on operating system requirements.
2 Actual requirements will vary based on your system configuration and the applications and features you choose to install.


Note Microsoft SQL Server 2000 does not have a hardware compatibility list (HCL). If your computer meets the minimum requirements listed in the preceding table, SQL Server 2000 software works on the hardware certified for use with the Microsoft Windows® operating system. For more information about hardware certified for use with the Windows operating system, see the Microsoft Windows Hardware Compatibility List at Microsoft Web site.

Operating System Requirements
This table shows the operating systems that must be installed to use the various editions or components of Microsoft SQL Server 2000.

SQL Server 2000 edition or component
Operating system requirement
Enterprise Edition and Standard Edition Windows Server 2003, Standard Edition1
Windows Server 2003, Enterprise Edition1
Windows Server 2003, Datacenter Edition1
Windows 2000 Server
Windows 2000 Advanced Server
Windows 2000 Datacenter Server
Microsoft Windows NT Server version 4.0 with Service Pack 5 (SP5) or later
Windows NT Server version 4.0, Enterprise Edition, with SP5 or later
Enterprise Evaluation Edition and Developer Edition Operating systems listed above for Enterprise and Standard Editions
Windows XP Professional
Windows XP Home Edition
Windows 2000 Professional
Windows NT Workstation 4.0 with SP5 or later
SQL Server 2000 Personal Edition2 and SQL Server 2000 Desktop Engine (MSDE 2000) Operating systems listed above for Enterprise, Standard, Enterprise Evaluation, and Developer Editions
Windows Server 2003, Web Edition1 (MSDE only)
Windows 98
Windows Me
Client Tools Support Windows 953
Windows 98
Windows Me
Windows NT Workstation 4.0
Windows 2000 Professional
Windows XP Professional
Windows XP Home Edition
Windows Server 2003, Standard Edition1
Windows Server 2003, Enterprise Edition1
Windows Server 2003, Datacenter Edition1
UNIX, Apple Macintosh, and OS/2 clients require Open Database Connectivity (ODBC) client software from a third-party vendor.
Connectivity Only Windows 95
Windows 98
Windows Me
Windows NT Workstation 4.0
Windows 2000 (all editions)
Windows XP Professional
Windows XP Home Edition
Windows Server 2003, Standard Edition1
Windows Server 2003, Enterprise Edition1
Windows Server 2003, Datacenter Edition1


1 Windows Server 2003 requires SQL Server 2000 Service Pack 3 or later to be applied.
2 SQL Server 2000 Personal Edition is offered for desktop and mobile use. Personal Edition does not contain the full functionality of Standard Edition. Analysis Services, including online analytical processing (OLAP), data mining, and other data warehousing features, are included in Personal Edition but cannot be installed on Windows 98 or Windows Me.
3 Supported for client connectivity only; does not include graphical tools support.


Note Microsoft Windows NT® Server 4.0, Service Pack 5 (SP5) or later must be installed as a minimum requirement for all SQL Server 2000 editions.

SQL Server 2000 is not supported on Windows NT 4.0 Terminal Server.

For installations of SQL Server 2000 Personal Edition or Desktop Engine (MSDE 2000) on Windows 98 computers without a network card, Windows 98 Second Edition is required.

Internet Requirements
This table shows Internet requirements related to using Microsoft SQL Server 2000.

Component Requirement
Internet software Microsoft Internet Explorer 5.0 is required for all installations of Microsoft SQL Server 2000, as it is required for Microsoft Management Console (MMC) and HTML Help. A minimal install is sufficient, and Internet Explorer is not required to be the default browser.
Exception to the Internet Explorer 5.0 requirement: If using the Connectivity Only option and not connecting to a server that requires encryption, Microsoft Internet Explorer 4.01 with Service Pack 2 is sufficient.

Internet Information Services If writing XML applications, see System Requirements for the IIS Virtual Directory Management for SQL Server Utility.


Network Software Requirements
Microsoft Windows XP, Windows 2000, Windows 98, Windows 95, Windows Me, and Windows NT have built-in network software. Additional network software is required only if you are using Banyan VINES or AppleTalk ADSP. Novel NetWare IPX/SPX client support is provided by the NWLink protocol of Windows-based networking.


Note TCP/IP must be enabled at the operating system level before installing SQL Server 2000. For more information, see Network Libraries.

Considerations for Other Microsoft Products
The following Microsoft products require Service Release or Service Packs to operate correctly with SQL Server 2000.

Access 2000
Microsoft Access 2000 requires the installation of either Microsoft Office 2000 Service Release 1 (SR1) or Access 2000 SR1 to operate correctly with SQL Server 2000. If running an earlier version of Access 2000, you cannot test automatic data processing (ADP) applications against SQL Server 2000. You cannot access database diagrams, stored procedures, table designs, or view designs.

Other issues to be addressed in a future Access Service Release:

When you run Access 2000 with SR1, you can test ADP applications. You can also alter database diagrams, stored procedures, table designs, or view designs, but you cannot save any changes. A future Access Service Release will allow limited ability to save changes.


The Access 2000 Create Database Wizard cannot successfully create a SQL Server 2000 database. You can work around this by first creating the database using SQL Server Enterprise Manager, and then creating an ADP for the database using the Project (Existing Database) option on the New dialog box in Access 2000.


The Access 2000 Upsizing Wizard does not support upsizing to SQL Server 2000. You can work around this by using Data Transformation Services in the Enterprise Manager to import your MDB database file into SQL Server. You can then rename your MDB tables and create linked tables to the resulting SQL Server database with the same names as your original MDB table names.
Visual Studio 6.0
When you run Microsoft Visual Studio® 6.0, you cannot access database diagrams, stored procedures, table designs, or view designs in SQL Server 2000. Visual Studio 6.0 Service Pack 4 allows you to alter database diagrams, stored procedures, table designs, or view designs, but you cannot save them. A future Visual Studio Service Pack will allow a limited ability to save changes.

Database Design by Yulius Eka Agung Seputra,ST,MSi

Part 1: An overview of index-tuning tools

As part of my current job at Microsoft, I work with SQL customers worldwide, ranging from fortune 500 financial institutions to online retailers to small partner outfits. I am engaged primarily to conduct performance tuning of their SQL Server systems.

When a system has a performance problem, and time is short (as it inevitably is), it's tempting just to "kill it with iron" – add more memory, more CPU horsepower, bigger and better disk subsystems, and so on. The problem is that if your design is fundamentally flawed then there is a strong chance that your application will not scale: doubling your horsepower will not double the performance. You might end up throwing a lot of hardware (and a lot of money) at the problem in order to reach your performance target.

Instead, your first port of call should be with the design of the application, and probably the "lowest-hanging fruit" here – the option that produces maximum gains with least impact on existing systems and processes – is to examine your indexing strategy.

When you run a query, SQL Server query optimizer's job is to find the best possible execution plan (based on internal costing algorithms) to provide the best response time. If the indexes required for implementing this optimal execution plan are not present, then the query optimizer is forced to choose a sub-optimal execution plan, for which indexes do exist. This could cause SQL Server to do more I/O and result in slower query performance. The need to have the right indexes in place cannot be overemphasized.

However, the task of identifying the right indexes is not necessarily straightforward. It requires a sound knowledge of the of the sort of queries that will be run against the data, the distribution of that data, and the volume of data, as well as an understanding of what type of index will suit your needs best. While this understanding is still essential, SQL Server 2005 does offer a helping hand in the form of some new tools – principally the Database Engine Tuning Advisor – that can help you determine, tune and monitor your indexes.

In this article, I will demonstrate how to use these tools to get answers to the following questions:

Which indexes do I need for my queries?
How do I monitor index usage and their effectiveness?
How do I identify redundant indexes that could negatively impact performance of my DML queries (insert, updates and deletes)
As workload changes, how I do I identify any missing indexes that could enhance performance for my new queries?
Finding the right indexes for your workload
Determining exactly the right indexes for your system can be quite a taxing process. For example, you have to consider:

Which columns should be indexed (based on your knowledge of how the data is queried)
Whether to choose a single-column index or a multiple column index
Whether you need a clustered index or a non-clustered index
Whether or not (in SQL 2005) you could benefit from an index with included columns to avoid bookmark lookups
How to utilize indexed views (which the optimizer might access instead of the underlying tables to retrieve a subset of your data)
All of these options increase the complexity of identifying the right indexes for your system. You might be tempted to drop indexes in everywhere "just in case", but one has to remember that with the introduction of every new index for improving the performance of the SELECT workload, the query optimizer has to do more work when running DML statements (insert, update or delete), as now it has to reorganize this index to accommodate for the newly updated data in the table. There is a balance to be struck here.

Even once you've determined the prefect set of indexes, your job is not finished. Your workload will change over time (new queries are will be added, older ones removed from the application) and this might warrant revisiting existing indexes, analyzing their usage and making adjustments (modifying/dropping existing indexes or creating new ones). Maintenance of indexes is critical to ensuring optimal performance in the long run.

SQL 2005 provides the following tools and features to help you find the right indexes for your queries, and then monitor and tune them:

Database Engine Tuning Advisor (DTA)
STATISTICS XML output
Dynamic Management Views (DMVs)
I will describe each one of them in the subsequent sections.

Using Database Engine Tuning Advisor (DTA)
The DTA is a new tool in SQL 2005 which replaces the Index Tuning Wizard in earlier versions of SQL Server. DTA can analyze both OLTP and OLAP workloads. You can either tune a single query or the entire workload to which your server is subjected. Based on the options that you select, you can use the DTA to make recommendations for several Physical Design Structures (PDS), which include:

Clustered indexes
Non-clustered indexes
Indexes with included columns (to avoid bookmark lookups)
Indexed views
Partitions
Let's walk through the steps involved in using DTA effectively.

Generating the DTA workload
The first step is to collect a workload for DTA to analyze. You can do this one of two ways:

Using Management Studio – if you need to optimize the performance of a single query, you can use Management Studio to provide an input to DTA. Simply type the query in Management Studio, highlight it and then right click on it to choose Analyze in Database Engine Tuning Advisor. This will launch DTA with your query as the input.
Using Profiler – if you want to want to determine the optimum index set for the entire workload of your server, you should collect a profiler trace using the built-in TUNING template.

To unleash the true effectiveness of DTA, you should always use a representative profiler trace. Make sure that you subject your server to all the queries that will typically be run against the data, while you are collecting the trace. This could lead to a huge trace file, but that is normal. If you simply collect a profiler trace over a 5-10 minute period, you can be pretty sure it won't be truly representative of all the queries run against your database.

The TUNING template captures only minimal events, so there should not be any significant performance impact on your server.

As part of the code download for this article, I've provided a profiler trace (Workload2Analyze.trc) that was captured with the TUNING template while running numerous queries ( Queries4Workload.sql) against the AdventureWorks database. I would strongly recommend that you use this trace, or the one from your server, to get a hands-on perspective of DTA.

Consuming the workload using DTA
Having collected a truly representative profiler trace, you can use it as an input to DTA, which will then generate recommendations. You can perform one of the following two types of analysis.

SCENARIO I: Keep my existing PDS and tell me what else I am missing
This type of analysis is common and is useful if you have previously established the set of indexes that you deem to be most useful for your given workload, and are seeking further recommendations. To conduct this analysis:

Launch a new session in DTA.
Choose the profiler trace as the input to this session.
In the Select databases and tables to tune section select your target database (AdventureWorks, if you are using my script).
In the Database for workload analysis dropdown, you can either use the same database or Tempdb (or any other database). I generally use Tempdb for his purpose.
At the Tuning Options tab, select the following options:
Physical Design Structures to use in database -> Indexes and Indexed views
Physical Design Structures to keep in database -> Keep all existing PDS
Uncheck the checkbox for limit tuning time.
Hit START ANALYSIS and DTA will start consuming your workload.
Once DTA is finished consuming the workload, it will list all its recommendations, under the Recommendations tab. We will talk more about reading and implementing DTA's recommendations in the section Reading output from DTA.

SCENARIO II: Ignore my existing PDS and tell me what query optimizer needs
In scenario I, DTA makes recommendations for any missing indexes. However, this doesn't necessarily mean your existing indexes are optimal for the query optimizer. You may also consider conducting an analysis whereby DTA ignores all existing physical design structures and recommends what it deems the best possible set of PDS for the given workload. In this way, you can validate your assumptions about what indexes are required.

To conduct this analysis, follow steps 1 to 6 as above, except that at step 5b, choose "Do not keep any existing PDS".

Contrary to how this might sound, DTA will not actually drop or delete any existing PDSs. This is the biggest advantage of using DTA, as it means you can use the tool to perform what-if analysis without actually introducing any changes to the underlying schema.

A note on the DTA Tuning Log
As it consumes the workload, DTA will store in the DTA Tuning log (found under the Progress tab, in the bottom section) details of any errors it encounters.

Typically, you might see a message of the form "XX% of the consumed workload has syntax errors". The tuning log will provide further details of these errors, many of which can be safely ignored. For example:

Statement doesn't reference any table – statements such as SET or DECLARE in your workload will cause these errors, and they can generally be ignored
Statement references only small tables – DTA will not tune a query if it references a small table (10 data pages or less)
Incorrect syntax or object related errors – if these errors indicate queries statments containing keywords such as BEGIN, TRY, INSERTED then you can ignore them.
To learn more about how to analyze the tuning log, refer to About the Tuning Log and Determining Whether Events Can Be Tuned on BOL

Reading output from DTA
After consuming the workload, DTA presents, under the Recommendations tab, a set of recommendations for tuning your PDS. I tend to focus on the following sections:


Recommendation – this is the action that you need to take. Possible values include Create or Drop.
Target of Recommendation – this is the proposed name of the PDS to be created. The naming convention is typical of DTA and generally starts with _dta*. However, I recommend that you change this name based on the naming convention in your database.
Definition – this is the list of columns that this new PDS will include. If you click on the hyperlink, it will open up a new window with the T-SQL script to implement this recommendation.
Estimated Improvements – this is the estimated percentage improvement that you can expect in your workload performance, if you implement all the recommendations made by DTA.
Space used by recommendation (MB) – under the Tuning Summary section of the Reports tab, you can find out the extra space in MB that you would need, if you decide to implement these recommendations.
Inbuilt analysis reports
There are 15 inbuilt reports under the Reports tab. However, the following three reports are the most important.

NOTE:
For information on these and the other 12 reports, visit the article Choosing a Database Engine Tuning Advisor Report on MSDN.

Index Usage Report (current)
Start with this report to see how your existing indexes are being used by the queries running against your server. Each index that has been used by a query is listed here. Each referenced index has a Percent Usage value which indicates the percentage of statements in your workload that referenced this index.

If an index is not listed here, it means that it has not been used by any query in your workload.

TIP: identifying indexes that are not used
If you are certain that all the queries that run against your server have been captured by your profiler trace, then you can use this report to identify indexes that are not required and possibly delete them.

Index Usage Report (recommended)
Next, look at this report to identify how index usage will change if the recommended indexes are implemented. If you compare these two reports, you will see that the index usage of some of the current indexes has fallen while some new indexes have been included with a higher usage percentage, indicating a different execution plan for your workload and improved performance.

Statement Cost Report
This report lists individual statements in your workload and the estimated performance improvement for each one of them. Using this report, you can identify your poorly performing queries and see the sort of improvement you can expect if you implement the recommendations made by DTA.

You will find that some statements don't have any improvements (Percent improvement = 0). This is because either the statement was not tuned for some reason or it already has all the indexes that it needs to perform optimally.

Implementing DTA's recommendations
By now, we have collected a workload using Profiler, consumed it using DTA and got a set of recommendations to improve performance. You then have the choice to either:

Save recommendations – you can save the recommendations in an SQL script by navigating ACTIONS | SAVE RECOMMENDATIONS. You can then manually run the script in Management Studio to create all the recommended PDS.
Apply recommendations using DTA – if you are happy with the set of recommendations then simply navigate ACTIONS | APPLY RECOMMENDATIONS. You can also schedule a later time to apply these recommendations (during off-peak hours, for example)
More often than not I use the Save Recommendations option. It means that I can easily compare the resulting PDS set for several scenarios. The recommendation reports are also useful for future reference. As the workload on the server changes, you need to re-evaluate the index usage and make corrections (drop redundant indexes or create new ones). You can compare the index usage on a periodic basis (may be quarterly or so) and see how the usage is changing.

Performing what-if analysis using DTA
This is a very cool feature of DTA. Suppose you don't want to apply all the recommendations that DTA provided. However, since the Estimated Improvement value can only be achieved if you apply all of these recommendations together, you are not really sure what kind of an impact it will have if you only choose to apply a sub-set of these recommendations.

Deselect the recommendations that you don't want to apply. Now, go to ACTIONS | EVALUATE RECOMMENDATIONS.

This will launch another session with the same options as the earlier one. However, now when you click on START ANALYSIS, DTA will provide data on estimated performance improvements, based on just this sub-set of the recommendations.

Again, the key thing to remember is that DTA performs this "what-if" analysis without actually implementing anything in the database.

TIP: checking for redundant objects left behind by DTA
In theory, DTA should not leave any extra "hypotetical" PDS (indexes etc.) in your database, after the analysis is complete. However, you can easily check for this by running the following query:

select name from sys.indexes where is_hypothetical = 1Using STATISTICS XMLIf you just wish to determine the optimum index set for a particular query, then the STATISTICS XML option provides a quick alternative to DTA.When a query is executed with STATISTICS XML enabled, SQL Server not only returns the query result-set, but also executes T-SQL statements that retrieve the execution plan for the query, in the form of an XML file (Showplan.xml). For example, try executing the following SQL statement:SET STATISTICS XML ONGOUse AdventureworksSELECT CustomerID, PurchaseOrderNumber, TotalDueFROM Sales.SalesOrderHeaderWHERE ShipMethodID > 2AND TotalDue > 200.00AND TerritoryID = 3;GOSET STATISTICS XML OFFGOBelow the query results, you should see a link to an XML file. The most important elements to look for in the XML file are: - the value of the StatementText will be the query that you just ran. You can confirm that this matches the query that you ran in Management Studio. - details of any missing indexes that the optimizer believes would improve the performance of this query are logged here For the above query, the relevant section in your XML file might look something like this: It indicates that we have one missing index on the Sales.SalesOrderHeader table. Look for the value of Impact in the element, which estimates the percentage increase in response time you can expect to see if you implement the set of indexes highlighted in this XML.Within the element: lists index columns that are used for equality in the query ( TerritortyID, in our case). Of all the conditions specified in the query, the query optimizer first looks for an index on the equality column in order to evaluate the rows that meet the criteria lists index columns that are used to evaluate inequality conditions in the query ( ShipMethodID and TotalDue, in this case) lists index columns that are included to cover the query ( PurchaseOrderNumber and CustomerID). You can specify included columns for non-clustered indexes to avoid a bookmark lookup in your execution plan NOTE:To find out more about included columns on non-clustered indexes please refer to the Index with Included Columns article on MSDN.To create this missing index, you should use the following template:CREATE NONCLUSTERED INDEX <>ON <> (equality_columns, followed by inequality_columns)INCLUDE (included_columns);So to create the missing index in this case, the command is as follows:CREATE NONCLUSTERED INDEX TestIndex ON Sales.SalesOrderHeader (TerritoryID, ShipMethodID, TotalDue) INCLUDE (PurchaseOrderNumber, CustomerID);Using index-related Dynamic Management Views (DMVs)Dynamic Management Views are a new feature of SQL 2005. They maintain server-wide state and configuration information for your SQL Server installation. There are more than 50 DMVs in SQL 2005 and the information stored in them can be used to monitor the health of a server instance, diagnose problems, and tune performance. There are several sets of DMVs, each of which store particular server information. One such set of DMVs store information about indexes and their usage. If the query optimizer finds that the indexes required for implementing a particular execution plan are not present, it will log the information about these missing indexes in the index-related DMVs. This information is persisted in the DMVs, until the next restart of SQL Server or when the metadata associated with the objects gets dropped.You can query these DMVs to find the missing indexes or store the information for periodic analysis. Let's take a look at how you can use the information in the DMVs and how you can interpret the output to create the requisite indexes. Before we begin, go ahead and restart your SQL Server on your test machine so that the information is cleared in each DMV. Next, in a new query window, execute the same query as in the previous section, but with the STATISTICS XML option turned off:SELECT CustomerID, PurchaseOrderNumber, TotalDueFROM Sales.SalesOrderHeaderWHERE ShipMethodID > 2AND TotalDue > 200.00AND TerritoryID = 3;The query optimizer will kick in to optimize the query and find the best execution plan. We already know that we have a missing index, so the query optimizer will settle for a sub-optimal plan. However, it will make a note about this in the relevant index related DMV.To retrieve the missing index details, simply query the relevant DMV, as follows:select * from sys.dm_db_missing_index_detailsThis query returns the id of the object on which you should create the index, which happens to be Sales.SalesOrderHeader along with the equality_columns, inequality_columns and included_columns, which are as described previously. As you will see, the DMV recommends creating the same "missing" index as the STATISTICS XML option and the syntax for creating this index is also the same as described above.Every time you restart SQL Server, these DMVs are flushed out. So you might want to copy the values in these DMVs into some other tables in your database to conduct periodic analysis.Using the right tool for the jobThe STATISTICS XML option and the index-related DMVs are both great tools for assessing the impact of a particular query on your indexing requirements. The biggest limitation in each case is that it cannot be used to fine tune the entire workload on your server. If you are running mission critical applications which involve a high number of queries, this approach will be very time and labor intensive. Furthermore, when using the DMVs, you can only access raw information about columns on which indexes might be missing. This approach also has several limitations and is less accurate for queries involving inequality predicates. For more details on the limitations, refer to the Limitations for Using the Missing Indexes Feature article on BOL.Database Engine Tuning Advisor is a more sophisticated and advanced tool, and provides more options than any of the other methods describes in this article. While using DTA, you can use a representative workload and, as a result, the recommendations made by DTA are more accurate. DTA can also be used against SQL 2000, while the other features cannot. Plus, you can perform what-if analysis using DTA without having to introducing any changes in your database. As a best practice recommendation, I would strongly advocate that you use DMVs or STATISTICS XML option only as a means to quickly assess the impact of a query on indexing requirements. You should always validate your findings, and the estimated improvements, using Database Engine Tuning Advisor, with a representative workload.ConclusionIn this first installment, I essentially provided an overview of how to use the different tools and features in SQL 2005 to determine the optimal set of indexes for the workload on your server.My personal preference is to use Database Engine Tuning Advisor. To get the most effective recommendations, you should capture a representative workload of your server using the TUNING template and use it as an input to the DTA.In the second part of this article, I will delve deeper into how to measure and monitor the usage and effectiveness of your existing indexes using DTA and DMVs. I'll also cover some more advanced "what-if" analysis using DTA, and some best practice for its general usage.