Here is the scenario:
SQL Server 2000 (8.0.2055)
Table currently has 478 million rows of data. The Primary Key column is an INT with IDENTITY. There is an Unique Constraint imposed on two other columns with a Non-Clustered Index. This is a vendor application and we are only responsible for maintaining the DB.
Now the vendor has recommended doing the following "to improve performance"
Drop the PK and Clustered Index
Drop the non-clustered index on the two columns with the UNIQUE CONSTRAINT
Recreate the PK, with a NON-CLUSTERED index
Create a CLUSTERED index on the two columns with the UNIQUE CONSTRAINT
I am not convinced that this is the right thing to do. I have a number of concerns.
By dropping the PK and indexes, you will be creating a heap with 478 million rows of data. Then creating a CLUSTERED INDEX on two columns would be a really mammoth task. Would creating another table with the same structure and new indexing scheme and then copying the data over, dropping the old table and renaming the new one be a better approach?
I am also not sure how the stored procs will react. Will they continue using the cached execution plan, considering that they are not being explicitly recompiled.
I am simply not able to understand what kind of "performance improvement" this change will provide. I think that this will actually have the reverse effect.
All thoughts welcome.
Thanks in advance,
Raj
All stored procs will recompile automatically. This will happen anyway when stats change and after index maintenance anyway.
At some point, you have to reorganise 478 million rows (drop/create indexes) or move (new table). Neither way is better then the other, unfortunately. I feel your pain though: we have similarly large tables with pending new columns and an index changes.
Saying that, you should do it step 2-1-4-3 to avoid unnecessary non-clustered index maintenance when you drop/create the clustered index.
And drop the unique constraint. The clustered index could be unique and clustered. A unique constrint is just another index that would be unnecessary.
As for the performance benefit, perhaps ask the vendor why.
The one thing I would have a serious look at is the question what type those other two columns are - how big are they, compared to the INT IDENTITY (4 byte) ??
The reason I ask: the clustering key will be added to all non-clustered indices on the table, too - and if you have close to 500 million rows, it will make a huge difference whether the clustering key is a single 4-byte INT, or e.g. two 16-byte GUID's.
This is not only on disk, mind you - the pages are loaded into SQL Server's RAM in their entirety - so by potentially bloating up your clustering key, you'd incur performance penalties due to the larger number of pages on disk (and in RAM) that your non-clustered indices would need.
The only compelling reason I could see to actually go through with those changes would be if by clustering the table using those two other columns, you'd gain something in terms of query performance, e.g. if some of the most frequent queries would be faster, due to the fact that the table is now clustered by these two columns. That's really hard to know unless you know what the access and query patterns really are....
Would creating another table with the same structure and new indexing scheme and then copying the data over, dropping the old table and renaming the new one be a better approach?
I believe that this is what SQL Enterprise Manager will do behind the scenes anyways if you use the visual tools to do this. If you make a schema change such as add a column in the middle of a table, or change primary keys, there is a little button that will allow you to "Script Changes". If you view this script, you can see the steps that Enterprise Manager will take in order to do what you requested.
Related
What happens if a clustered index is not unique? Can it lead to bad performance because inserted rows flow to an "overflow" page of some sorts?
Is it "made" unique and if so how? What is the best way to make it unique?
I am asking because I am currently using a clustered index to divide my table in logical parts, but the performance is so-so, and recently I got the advice to make my clustered indexes unique. I'd like a second opinion on that.
They don't have to be unique but it certainly is encouraged.
I haven't encountered a scenario yet where I wanted to create a CI on a non-unique column.
What happens if you create a CI on a non-unique column
If the clustered index is not a unique
index, SQL Server makes any duplicate
keys unique by adding an internally
generated value called a uniqueifier
Does this lead to bad performance?
Adding a uniqueifier certainly adds some overhead in calculating and in storing it.
If this overhead will be noticable depends on several factors.
How much data the table contains.
What is the rate of inserts.
How often is the CI used in a select (when no covering indexes exist, pretty much always).
Edit
as been pointed out by Remus in comments, there do exist use cases where creating a non-unique CI would be a reasonable choice. Me not having encountered one off those scenarios merely shows my own lack of exposure or competence (pick your choice).
I like to check out what The Queen of Indexing, Kimberly Tripp, has to say on the topic:
I'm going to start with my recommendation for the Clustering Key - for a couple of reasons. First, it's an easy decision to make and second, making this decision early helps to proactively prevent some types of fragmentation. If you can prevent certain types of base-table fragmentation then you can minimize some maintenance activities (some of which, in SQL Server 2000 AND less of which, in SQL Server 2005) require that your table be offline. OK, I'll get to the rebuild stuff later.....
Let's start with the key things that I look for in a clustering key:
* Unique
* Narrow
* Static
Why Unique?
A clustering key should be unique because a clustering key (when one exists) is used as the lookup key from all non-clustered indexes. Take for example an index in the back of a book - if you need to find the data that an index entry points to - that entry (the index entry) must be unique otherwise, which index entry would be the one you're looking for? So, when you create the clustered index - it must be unique. But, SQL Server doesn't require that your clustering key is created on a unique column. You can create it on any column(s) you'd like. Internally, if the clustering key is not unique then SQL Server will “uniquify” it by adding a 4-byte integer to the data. So if the clustered index is created on something which is not unique then not only is there additional overhead at index creation, there's wasted disk space, additional costs on INSERTs and UPDATEs, and in SQL Server 2000, there's an added cost on a clustereD index rebuild (which because of the poor choice for the clustering key is now more likely).
Source: Ever-increasing clustering key debate - again!
Do clustered indexes have to be unique?
They don't, and there are times where it's better if they're not.
Consider a table with a semi-random, unique EmployeeId, and a DepartmentId for each employee: if your select statement is
SELECT * FROM EmployeeTable WHERE DepartmentId=%DepartmentValue%
then it's best for performance if the DepartmentId is the clustered index even though (or even especially because) it's not the unique index (best for performance because it ensures that all the records within a given DepartmentId are clustered).
Do you have any references?
There's Clustered Index Design Guidelines for example, which says,
With few exceptions, every table
should have a clustered index defined
on the column, or columns, that offer
the following:
Can be used for frequently used queries.
Provide a high degree of uniqueness.
Can be used in range queries.
My understanding of "high degree of uniqueness" for example is that it isn't good to choose "Country" as the clusted index if most of your queries want to select the records within a given town.
If you are tuning an old DB this is a Godsend. I am working on Perf issues on a 20-year-old DB. It has nonclustered PKs with 3 - 8 columns. Instead of using all 8 columns to be unique I can pick one column with broad distribution, and it applies a Uniqueifier. It is an Int but by using a column like Project ID it can handle 2147483647 unique projectIDs which is enough for most use-cases. If it is not enough add a second or third column to the cluster.
This works without any coding modification in the App layer. 20 years in production and management doesn't have to order a major rewrite.
My question is about limitation of clustered index on a table.
By theory, in a single table we can have only one cluster index. But what if I have datetime columns in a table say "From date" and "To date"? These columns will often required in WHERE clause to populate reports in my application. And if I also require a cluster index on primary key in the same table, then still how to get advantage of cluster index on other columns? In this case my queries will still run slower with larger records.
In practice, you also can have just a single clustered index on a table - since the table's data is physically ordered by that clustered index.
If you need two datetime columns frequently in WHERE clauses, the best choice would be to have a non-clustered index on those two columns, and possibly include additional columns that you frequently retrieve with those queries, in order to make it a covering index.
There's really not much difference between a good, covering non-clustered index, and the clustered index, in terms of query performance.
However, you don't want to bloat your clustered index, since those columns will also be added to all non-clustered indices on the same table - keep it small, preferably an INT, ever-increasing, stable (not changing) and you should be just fine.
Another option is indexed (or materialized) views: you could create multiple views on the table, each with a different clustered index. That might be useful in a reporting scenario, but indexed views have lots of restrictions and will affect the performance of queries that modify the table data. Books Online has all the information you need to create and test them.
I suspect your real requirement is indeed to implement a reporting solution, and if so then it might be best to do it properly: create a separate database with a schema optimized for reporting (Google "star schema") and load data regularly from the main database into the reporting one. But that's a whole new area of development to investigate, and I wouldn't rush into it.
If you need the performance of a cluster index table for multiple indexes of the same table, the only route I see is holding a copy of the table for each cluster index.
The clustered index effects the physical storage of the data in a table, so by definition there can only be one. You can widen your clustered index to include the other columns, but this can have its own disadvantages.
The performance advantage from a clustered index is that the records are stored in a manner that reflects the index (which is why random inserts and updates to the clustered index very quickly fragment the table), and therefore query performance based on this index can be as good as performance as reads from your storage device, you can't get this on other indexes.
I suggest that you choose the index from which you would derive the most benefit from clustering and make that your clustered index. Make the rest of the indexes non-clustered. You may want to run some tests to find out what benefits will be derived from making different indexes clustered vs. non-clustered.
Share and enjoy.
We have a legacy database which is a sql server db (2005, and 2008).
All of the primary keys in the tables are UniqueIdentifiers.
The tables currently have no clustered index created on them and we are running into performance issues on tables with only 750k records. This is the first database i've worked on with unique identifiers as the sole primary key and I've never seen sql server be this slow with returning data.
I don't want to create a clustered index on the uniqueidentifier as they are not sequential and will therefore slow the apps down when it comes to inserting data.
We cannot remove the uniqueidentifier as that is used for remote site record identity management purposes.
I had thought about adding a big integer identity column to the tables and creating the clustered index on this column and including the unique identifier column.
i.e.
int identity - First column to maintain insert speeds
unique identifier - To ensure the application keeps working as expected.
The goal is to improve the identity query and joined table query performance.
Q1: Will this improve the query performance of the db or will it slow it down?
Q2: Is there an alternative to this that I haven't listed?
Thanks
Pete
Edit: The performance issues are on retrieving data quickly through select statements, especially if a few of the more "transactional / changing" tables are joined together.
Edit 2: The joins between tables are generally all between the primary key and foreign keys, for tables that have foreign keys they are included in the non-clustered index to provide a more covering index.
The tables all have no other values which would provide a good clustered index.
I'm leaning more towards adding an additional identity column on each of the high load tables and then including the current Guid PK column within the clustered index to provide the best query performance.
Edit 3:
I would estimate that 80% of the queries are performed on primary and foreign keys alone through the data access mechanism. Generally our data model has lazy loaded objects which perform the query when accessed, these queries use the objects id and the PK column. We have a large amount of user driven data exclusion / inclusion queries which use the foreign key columns as a filter based on the criteria of for type X exclude the following id's. The remaining 20% is where clauses on Enum (int) or date range columns, very few text based queries are performed in the system.
Where possible I have already added covering indexes to cover the heaviest queries, but as yet i'm still dissapointed by the performance. As bluefooted says the data is being stored as a heap.
If you don't have a clustered index on the table, it is being stored as a heap rather than a b-tree. Heap data access is absolutely atrocious in SQL Server so you definitely need to add a clustered index.
I agree with your analysis that the GUID column is a poor choice for clustering, especially since you don't have the ability to use NEWSEQUENTIALID(). You could create a new artificial integer key if you like, but if there is another column or combination of columns that would make sense as a clustered index, that is fine as well.
Do you have a field that is used frequently for range scans? Which columns are used for joins? Is there a combination of columns that also uniquely identifies the row aside from the GUID? Posting a sample of the data model would help us to suggest a good candidate for clustering.
I'm not sure where your GUIDs come from, but if they're being generated during the insert, using the NEWSEQUENTIALID() in SQL Server instead of NEWID() will help you avoid fragmentation issues during insert.
Regarding the choice of a clustered index, as Kimberly L. Tripp states here: "the most important factors in choosing a clustered index are that it's unique, narrow and static (ever-increasing has other benefits to minimizing splits)." A GUID falls short on the narrow requirement when compared to an INT or even BIGINT.
Kimberly also has an excellent article on GUIDs as PRIMARY KEYs and/or the clustering key.
It's not 100% clear to me: is your number 1 access pattern to query the tables by the GUID or by other columns? And when joining to other tables, what columns (and data types) are most often used?
I can't really give you any solid recommendations until I understand more about how these GUIDs are used. I realize you said they're primary keys, but that doesn't guarantee they are used as the primary conditions on queries or in joins.
UPDATE
Now that I know a little more, I have a crazy suggestion. Do cluster those tables on the GUIDs, but set the fill factor to 60%. This will ameliorate the page split problem and give you better performance querying on those puppies.
As for using Guid.NewGuid(), it seems that you can do sequentialGUIDs in C# after all. I found the following code here on SO:
[DllImport("rpcrt4.dll", SetLastError = true)]
static extern int UuidCreateSequential(out Guid guid);
public static Guid SequentialGuid()
{
const int RPC_S_OK = 0;
Guid g;
if (UuidCreateSequential(out g) != RPC_S_OK)
return Guid.NewGuid();
else
return g;
}
newsequentialID() is actually just a wrapper for UuidCreateSequential. I'm sure if you can't use this directly on the client you can figure out a way to make a quick round-trip to the server to get a new sequential id from there, perhaps even with a "dispenser" table and a stored procedure to do the job.
You don't indicate what your performance issues are. If the worst performing action is an INSERT, then maybe your solution is right. If it's something else, then I'd look at how the clustered index can help that.
You might look at existing indexes on the table and the queries that use them. You may be able to select an index that, while degrades INSERTs slightly, provides a greater benefit to the current performance-problem areas.
In many places it's recommended that clustered indexes are better utilized when used to select range of rows using BETWEEN statement. When I select joining by foreign key field in such a way that this clustered index is used, I guess, that clusterization should help too because range of rows is being selected even though they all have same clustered key value and BETWEEN is not used.
Considering that I care only about that one select with join and nothing else, am I wrong with my guess ?
Discussing this type of issue in the absolute isn't very useful.
It is always a case-by-case situation !
Essentially, access by way of a clustered index saves one indirection, period.
Assuming the key used in the JOIN, is that of the clustered index, in a single read [whether from an index seek or from a scan or partial scan, doesn't matter], you get the whole row (record).
One problem with clustered indexes, is that you only get one per table. Therefore you need to use it wisely. Indeed in some cases, it is even wiser not to use any clustered index at all because of INSERT overhead and fragmentation (depending on the key and the order of new keys etc.)
Sometimes one gets the equivalent benefits of a clustered index, with a covering index, i.e. a index with the desired key(s) sequence, followed by the column values we are interested in. Just like a clustered index, a covering index doesn't require the indirection to the underlying table. Indeed the covering index may be slightly more efficient than the clustered index, because it is smaller.
However, and also, just like clustered indexes, and aside from the storage overhead, there is a performance cost associated with any extra index, during INSERT (and DELETE or UPDATE) queries.
And, yes, as indicated in other answers, the "foreign-key-ness" of the key used for the clustered index, has absolutely no bearing on the the performance of the index. FKs are constraints aimed at easing the maintenance of the integrity of the database but the underlying fields (columns) are otherwise just like any other field in the table.
To make wise decisions about index structure, one needs
to understands the way the various index types (and the heap) work
(and, BTW, this varies somewhat between SQL implementations)
to have a good image of the statistical profile of the database(s) at hand:
which are the big tables, which are the relations, what's the average/maximum cardinality of relation, what's the typical growth rate of the database etc.
to have good insight regarding the way the database(s) is (are) going to be be used/queried
Then and only then, can one can make educated guesses about the interest [or lack thereof] to have a given clustered index.
I would ask something else: would it be wise to put my clustered index on a foreign key column just to speed a single JOIN up? It probably helps, but..... at a price!
A clustered index makes a table faster, for every operation. YES! It does. See Kim Tripp's excellent The Clustered Index Debate continues for background info. She also mentions her main criteria for a clustered index:
narrow
static (never changes)
unique
if ever possible: ever increasing
INT IDENTITY fulfills this perfectly - GUID's do not. See GUID's as Primary Key for extensive background info.
Why narrow? Because the clustering key is added to each and every index page of each and every non-clustered index on the same table (in order to be able to actually look up the data row, if needed). You don't want to have VARCHAR(200) in your clustering key....
Why unique?? See above - the clustering key is the item and mechanism that SQL Server uses to uniquely find a data row. It has to be unique. If you pick a non-unique clustering key, SQL Server itself will add a 4-byte uniqueifier to your keys. Be careful of that!
So those are my criteria - put your clustering key on a narrow, stable, unique, hopefully ever-increasing column. If your foreign key column matches those - perfect!
However, I would not under any circumstances put my clustering key on a wide or even compound foreign key. Remember: the value(s) of the clustering key are being added to each and every non-clustered index entry on that table! If you have 10 non-clustered indices, 100'000 rows in your table - that's one million entries. It makes a huge difference whether that's a 4-byte integer, or a 200-byte VARCHAR - HUGE. And not just on disk - in server memory as well. Think very very carefully about what to make your clustered index!
SQL Server might need to add a uniquifier - making things even worse. If the values will ever change, SQL Server would have to do a lot of bookkeeping and updating all over the place.
So in short:
putting an index on your foreign keys is definitely a great idea - do it all the time!
I would be very very careful about making that a clustered index. First of all, you only get one clustered index, so which FK relationship are you going to pick? And don't put the clustering key on a wide and constantly changing column
An index on the FK column will help the JOIN because the index itself is ordered: clustered just means that the data on disk (leaf) is ordered rather then the B-tree.
If you change it to a covering index, then clustered vs non-clustered is irrelevant. What's important is to have a useful index.
It depends on the database implementation.
For SQL Server, a clustered index is a data structure where the data is stored as pages and there are B-Trees and are stored as a separate data structure. The reason you get fast performance, is that you can get to the start of the chain quickly and ranges are an easy linked list to follow.
Non-Clustered indexes is a data structure that contains pointers to the actual records and as such different concerns.
Refer to the documentation regarding Clustered Index Structures.
An index will not help in relation to a Foreign Key relationship, but it will help due to the concept of "covered" index. If your WHERE clause contains a constraint based upon the index. it will be able to generate the returned data set faster. That is where the performance comes from.
The performance gains usually come if you are selecting data sequentially within the cluster. Also, it depends entirely on the size of the table (data) and the conditions in your between statement.
OK. I've read things here and there about SQL Server heaps, but nothing too definitive to really guide me. I am going to try to measure performance, but was hoping for some guidance on what I should be looking into. This is SQL Server 2008 Enterprise. Here are the tables:
Jobs
JobID (PK, GUID, externally generated)
StartDate (datetime2)
AccountId
Several more accounting fields, mainly decimals and bigints
JobSteps
JobStepID (PK, GUID, externally generated)
JobID FK
StartDate
Several more accounting fields, mainly decimals and bigints
Usage: Lots of inserts (hundreds/sec), usually 1 JobStep per Job. Estimate perhaps 100-200M rows per month. No updates at all, and the only deletes are from archiving data older than 3 months.
Do ~10 queries/sec against the data. Some join JobSteps to Jobs, some just look at Jobs. Almost all queries will range on StartDate, most of them include AccountId and some of the other accounting fields (we have indexes on them). Queries are pretty simple - the largest part of the execution plans is the join for JobSteps.
The priority is the insert performance. Some lag (5 minutes or so) is tolerable for data to appear in the queries, so replicating to other servers and running queries off them is certainly allowable.
Lookup based on the GUIDs is very rare, apart from joining JobSteps to Jobs.
Current Setup: No clustered index. The only one that seems like a candidate is StartDate. But, it doesn't increase perfectly. Jobs can be inserted anywhere in a 3 hour window after their StartDate. That could mean a million rows are inserted in an order that is not final.
Data size for a 1 Job + 1 JobStepId, with my current indexes, is about 500 bytes.
Questions:
Is this a good use of a heap?
What's the effect of clustering on StartDate, when it's pretty much non-sequential for ~2 hours/1 million rows? My guess is the constant re-ordering would kill insert perf.
Should I just add bigint PKs just to have smaller, always increasing keys? (I'd still need the guids for lookups.)
I read GUIDs as PRIMARY KEYs and/or the clustering key, and it seemed to suggest that even inventing a key will save considerable space on other indexes. Also some resources suggest that heaps have some sort of perf issues in general, but I'm not sure if that still applies in SQL 2008.
And again, yes, I'm going to try to perf test and measure. I'm just trying to get some guidance or links to other articles so I can make a more informed decision on what paths to consider.
Yes, heaps have issues. Your data will logically fragment all over the show and can not be defragmented simply.
Imagine throwing all your telephone directory into a bucket and then trying to find "bob smith". Or using a conventional telephone directory with a clustered index on lastname, firstname.
The overhead of maintaining the index is trivial.
StartDate, unless unique, is not a good choice. A clustered index requires internal uniqueness for the non-clustered indexes. If not declared unique, SQL Server will add a 4 byte "uniquifier".
Yes, I'd use int or bigint to make it easier. As for GUIDs: see the questions at the right hand side of the screen.
Edit:
Note, PK and clustered index are 2 separate issues even if SQL Server be default will make the PK clustered.
Heap fragmentation isn't necessarily the end of the world. It sounds like you'll rarely be scanning the data, so that's not the end of the world.
Your non-clustered indexes are the things that will impact your performance. Each one will need to store the address of the row in the underlynig table (either a heap or a clustered index). Ideally, your queries never have to use the underlying table itself, because it stores all the information needed in the ideal way (including all columns, so that it's a covering index).
And yes, Kimberly Tripp's stuff is the best around for indexes.
Rob
As your own research has shown, and as all the other answerers have mentioned, using a GUID as the clustered index on a table is a bad idea.
However, having a heap also isn't really a good choice, since heaps have other issues, mostly to do with fragmentation and other things that just don't work well with a heap.
My best practice advice would always be this:
do use a primary, clustered key on any data table (unless it's a temporary table, or a table used for bulk-loading)
try to make sure the clustered key is a INT IDENTITY or BIGINT IDENTITY
I would argue that the benefits you get by adding a INT/BIGINT - even just for the sake of having a good clustered index - far outweigh the drawbacks this has (as Kim Tripp also argues in her blog post you cited).
Marc
As a GUId is your primary and foreign key your database will still need to check the contraints on every insert you will probably need to index this. Indexing a GUId is not advisable due to it's randomness. Therefore I'd say absolutely you should go down the bigint (probably identity) route for your primary key and use it as a clustered index.