Related
We're considering using UUID values as primary keys for our MySQL database. The data being inserted is generated from dozens, hundreds, or even thousands of remote computers and being inserted at a rate of 100-40,000 inserts per second, and we'll never do any updates.
The database itself will typically get to around 50M records before we start to cull data, so not a massive database, but not tiny either. We're also planing to run on InnoDB, though we are open to changing that if there is a better engine for what we're doing.
We were ready to go with Java's Type 4 UUID, but in testing have been seeing some strange behavior. For one, we're storing as varchar(36) and I now realize we'd be better off using binary(16) - though how much better off I'm not sure.
The bigger question is: how badly does this random data screw up the index when we have 50M records? Would we be better off if we used, for example, a type-1 UUID where the leftmost bits were timestamped? Or maybe we should ditch UUIDs entirely and consider auto_increment primary keys?
I'm looking for general thoughts/tips on the performance of different types of UUIDs when they are stored as an index/primary key in MySQL. Thanks!
At my job, we use UUID as PKs. What I can tell you from experience is DO NOT USE THEM as PKs (SQL Server by the way).
It's one of those things that when you have less than 1000 records it;s ok, but when you have millions, it's the worst thing you can do. Why? Because UUID are not sequential, so everytime a new record is inserted MSSQL needs to go look at the correct page to insert the record in, and then insert the record. The really ugly consequence with this is that the pages end up all in different sizes and they end up fragmented, so now we have to do de-fragmentation periodic.
When you use an autoincrement, MSSQL will always go to the last page, and you end up with equally sized pages (in theory) so the performance to select those records is much better (also because the INSERTs will not block the table/page for so long).
However, the big advantage of using UUID as PKs is that if we have clusters of DBs, there will not be conflicts when merging.
I would recommend the following model:
PK INT Identity
Additional column automatically generated as UUID.
This way, the merge process is possible (UUID would be your REAL key, while the PK would just be something temporary that gives you good performance).
NOTE: That the best solution is to use NEWSEQUENTIALID (like I was saying in the comments), but for legacy app with not much time to refactor (and even worse, not controlling all inserts), it is not possible to do.
But indeed as of 2017, I'd say the best solution here is NEWSEQUENTIALID or doing Guid.Comb with NHibernate.
A UUID is a Universally Unique ID. It's the universally part that you should be considering here.
Do you really need the IDs to be universally unique? If so, then UUIDs may be your only choice.
I would strongly suggest that if you do use UUIDs, you store them as a number and not as a string. If you have 50M+ records, then the saving in storage space will improve your performance (although I couldn't say by how much).
If your IDs do not need to be universally unique, then I don't think that you can do much better then just using auto_increment, which guarantees that IDs will be unique within a table (since the value will increment each time)
Something to take into consideration is that Autoincrements are generated one at a time and cannot be solved using a parallel solution. The fight for using UUIDs eventually comes down to what you want to achieve versus what you potentially sacrifice.
On performance, briefly:
A UUID like the one above is 36
characters long, including dashes. If
you store this VARCHAR(36), you're
going to decrease compare performance
dramatically. This is your primary
key, you don't want it to be slow.
At its bit level, a UUID is 128 bits,
which means it will fit into 16 bytes,
note this is not very human readable,
but it will keep storage low, and is
only 4 times larger than a 32-bit int,
or 2 times larger than a 64-bit int.
I will use a VARBINARY(16)
Theoretically, this can work without a
lot of overhead.
I recommend reading the following two posts:
Brian "Krow" Aker's Idle Thoughts - Myths, GUID vs Autoincrement
To UUID or not to UUID ?
I reckon between the two, they answer your question.
I tend to avoid UUID simply because it is a pain to store and a pain to use as a primary key but there are advantages. The main one is they are UNIQUE.
I usually solve the problem and avoid UUID by using dual key fields.
COLLECTOR = UNIQUE ASSIGNED TO A MACHINE
ID = RECORD COLLECTED BY THE COLLECTOR (auto_inc field)
This offers me two things. Speed of auto-inc fields and uniqueness of data being stored in a central location after it is collected and grouped together. I also know while browsing the data where it was collected which is often quite important for my needs.
I have seen many cases while dealing with other data sets for clients where they have decided to use UUID but then still have a field for where the data was collected which really is a waste of effort. Simply using two (or more if needed) fields as your key really helps.
I have just seen too many performance hits using UUID. They feel like a cheat...
Instead of centrally generating unique keys for each insertion, how about allocating blocks of keys to individual servers? When they run out of keys, they can request a new block. Then you solve the problem of overhead by connecting for each insert.
Keyserver maintains next available id
Server 1 requests id block.
Keyserver returns (1,1000)
Server 1 can insert a 1000 records until it needs to request a new block
Server 2 requests index block.
Keyserver returns (1001,2000)
etc...
You could come up with a more sophisticated version where a server could request the number of needed keys, or return unused blocks to the keyserver, which would then of course need to maintain a map of used/unused blocks.
I realize this question is rather old but I did hit upon it in my research. Since than a number of things happened (SSD are ubiquitous InnoDB got updates etc).
In my research I found this rather interesting post on performance:
claiming that due to the randomness of a GUID/UUID index trees can get rather unbalanced. in the MariaDB KB I found another post suggested a solution.
But since than the new UUID_TO_BIN takes care of this. This function is only available in MySQL (tested version 8.0.18) and not in MariaDB (version 10.4.10)
TL;DR: Store UUID as converted/optimized BINARY(16) values.
I would assign each server a numeric ID in a transactional manner.
Then, each record inserted will just autoincrement its own counter.
Combination of ServerID and RecordID will be unique.
ServerID field can be indexed and future select performance
based on ServerID (if needed) may be much better.
The short answer is that many databases have performance problems (in particular with high INSERT volumes) due to a conflict between their indexing method and UUIDs' deliberate entropy in the high-order bits. There are several common hacks:
choose a different index type (e.g. nonclustered on MSSQL) that doesn't mind it
munge the data to move the entropy to lower-order bits (e.g. reordering bytes of V1 UUIDs on MySQL)
make the UUID a secondary key with an auto-increment int primary key
... but these are all hacks--and probably fragile ones at that.
The best answer, but unfortunately the slowest one, is to demand your vendor improve their product so it can deal with UUIDs as primary keys just like any other type. They shouldn't be forcing you to roll your own half-baked hack to make up for their failure to solve what has become a common use case and will only continue to grow.
What about some hand crafted UID? Give each of the thousands of servers an ID and make primary key a combo key of autoincrement,MachineID ???
Since the primary key is generated decentralised, you don't have the option of using an auto_increment anyway.
If you don't have to hide the identity of the remote machines, use Type 1 UUIDs instead of UUIDs. They are easier to generate and can at least not hurt the performance of the database.
The same goes for varchar (char, really) vs. binary: it can only help matters. Is it really important, how much performance is improved?
The main case where UUIDs cause miserable performance is ...
When the INDEX is too big to be cached in the buffer_pool, each lookup tends to be a disk hit. For HDD, this can slow down the access by 10x or worse. (No, that is not a typo for "10%".) With SSDs, the slowdown is less, but still significant.
This applies to any "hash" (MD5, SHA256, etc), with one exception: A type-1 UUID with its bits rearranged.
Background and manual optimization: UUIDs
MySQL 8.0: see UUID_TO_BIN() and BIN_TO_UUID()
MariaDB 10.7 carries this further with its UUID datatype.
I have encountered the fact that some people, after performing deletion of rows from a table, also reset the AUTO_INCREMENT for the primary key column of that table to re-number all the values as if they started from 1 again (or whatever the initial starting point).
My question is, is there a specific reason for doing this, other than just preference? As in, is there any detrimental impact on the database or future queries if you do not reset the auto-increment and just leave it as-is? If there is, could somebody provide an example where it would be necessary to reset AUTO_INCREMENT?
Thanks!
I don't think it is ever necessary to reset auto_increment, unless you are running out of values.
One case where auto-increment is often reset is when all the rows are deleted. If you use truncate table, then the auto-increment value is reset automatically. This does not always happen with delete without a where clause, so for consistency, you might want to reset it.
Another case is when a large insert fails, particularly if it fails repeatedly. You might not want the really large gaps.
When moving tables around you might want to keep the original id values. So, essentially, you ignore the auto-increment on inserts. Afterwards, though, you might want to set the automatic value to be consistent with other systems.
In general, though, resetting the auto-increment is not recommended.
Unfortunately, I've seen this behavior. And from what I observed, it's not due to a technical reason - it's closer to OCD.
Some people really don't like gaps in the ID column - they like the idea of it smoothly increasing by 1 for each record. The idea that some manual data manipulation they're doing screwing that up isn't pleasant - so they go through some hoops to make sure they don't cause gaps in the numbers.
But, yeah, this is a terrible practice. It's just asking for data integrity problems.
Resetting auto-inc is an uncommon operation. Under normal day to day work, just let it keep incrementing.
I've done reset of auto-inc in MySQL instances used for automated testing. A given set of tables is loaded with data over and over, and deletes its test data afterwards. Resetting the auto-inc may be the best way to make tests repeatable, if they're looking for specific values in the results.
Another scenario is when creating archive tables. Suppose you have a huge table, and you want to empty out the data efficiently (not using DELETE), but you want to archive the data, and you want new data to use id values higher than your old data.
CREATE TABLE mytable_new LIKE mytable;
SELECT AUTO_INCREMENT FROM INFORMATION_SCHEMA.TABLES
WHERE TABLE_NAME='mytable';
ALTER TABLE mytable_new AUTO_INCREMENT = /* value + 10000 */;
RENAME TABLE mytable TO mytable_archive, mytable_new TO mytable;
The above series of statements allow you to shuffle a new empty table into place atomically, so your app can continue writing to the table by the name it's used to. The auto-inc value you reset in the new table should be a value higher than the max id value in the old table, plus some comfortable gap to avoid overlap during the moments between the statements.
Reseting the auto increment usually helps in terms of organization, you can see no gap between id 6 and 60 if the rows between have been deleted.
However, you should be carefull about working with resetting auto-increments, because most likely, your code will depend on specific id's to fetch certain information.
In my opinion, just truncate the whole thing after your tests and seed the database with the correct information. If it's production, let it run wild and free, it could cause more harm and no beneficial output
As per comment on abr's answer, assuming that auto-increment ids are contiguous (or even sequential) is not just a bad idea, it is a dangerous one.
There may be good reason for deliberately creating gaps in the allocated ids if you intend to patch the data at a later point (e.g. if you have restored from an old backup and expect to recover some of the missing data but need to restore a service asap) or when you migrate from a single active server to multiple master nodes. But in these scenarios you are setting the counter to higher value than currently used - not resetting it back to the start.
If there is a risk that you are going to wrap around the numbers, then you've probably picked the wrong data type for your auto-increment attribute - changing the data type is the right way to fix the problem, not deleting data and resetting the counter to 0.
Ok I'm new with MySQL so excuse any stupid questions and such but I'm having a problem with the simple database table I created.
See:
I have it set up so each new entry has a automatically incrementing id however if I were to delete the third row, there is a gap. I have heard that there is a way to cascade the other numbers down when a previous entry is deleted. I have looked but couldn't find anything that would help me change it in phpMyAdmin without recreating the table.
How would I do this?
This question pops up every once in a while, and there's this confusion that auto_increment gives you sequential numbers that you can use, well, for sequences. You can't.
The sole purpose of primary key is to uniquely identify a row. And incrementing integer fits that role perfectly, it's easy to implement in C/C++ MySQL code and it works extremely fast and well.
But that's all it does. You don't get nice, sequential numbering feature out of it. You can't use it for what you want because there are, as you called it - gaps.
And no, you don't make MySQL fill the gaps. It's bad and it's dangerous and it creates problems that you didn't even think of.
Bottom line is that you never rely on auto_increment to reuse "wasted" numbers.
Here's why:
InnoDB, default MySQL engine, uses the primary key internally to physically organize records on the hard drive. It relies on the feature that every next id is greater than one before. I won't get into details, but the key idea is that the index and data are written on the same page. That makes InnoDB extremely fast when doing primary key lookups (SELECT col FROM table WHERE id = 1000000 types of queries).
Now, what happens when you "reuse" the keys that had gaps - imagine this scenario: you have 1 million records. There were no number losses.
You delete the record 500 000.
Afterwards, you add new record, and using your logic - you need to "reuse" number 500 000. So you do it.
But, InnoDB expects every next record to be larger. So to conform to your needs, it has to rebalance what it has written. And it has to start from record 500 000. Now you have 500 000 records that are being reorganized, and that means you have 500 000 checks and writes going on. This kills your performance, COMPLETELY. Let's say you have a mechanical hard drive. It's capable of about 200-300 input output operations per second (IOPS). If every reorganized record requires 1 I/O, to reorganize 500 000 of them would take 20-30 minutes. Now you have inserts that take 30 minutes to complete.
The other problem, much more severe than the performance problem is concurrency and isolation problem.
What people don't understand is that MySQL (and other relational databases) solve the problem of concurrency, or if you will - problem of simultaneous access or "What happens when two users write at the same time" problem.
MySQL takes care of that, and even more. And the feature where it "wastes" numbers is what actually makes it happen. Every transaction relies on the primary key at one point. Even before you commit the transaction, an auto_increment was already assigned to it. So even when transaction fails or isn't commited, an auto_increment gets "wasted". It's actually DESIRABLE behavior.
I definitely didn't list all the disadvantages, just two that I could think of (I also didn't describe them in detail due to lack of time but principles do apply).
Conclusion is - do not "reuse" wasted auto_increment numbers, do not listen to people who tell you how to do it, do not assume that your project won't have problems I listed above. If you do, prepare to encounter problems you could never think of.
We're considering using UUID values as primary keys for our MySQL database. The data being inserted is generated from dozens, hundreds, or even thousands of remote computers and being inserted at a rate of 100-40,000 inserts per second, and we'll never do any updates.
The database itself will typically get to around 50M records before we start to cull data, so not a massive database, but not tiny either. We're also planing to run on InnoDB, though we are open to changing that if there is a better engine for what we're doing.
We were ready to go with Java's Type 4 UUID, but in testing have been seeing some strange behavior. For one, we're storing as varchar(36) and I now realize we'd be better off using binary(16) - though how much better off I'm not sure.
The bigger question is: how badly does this random data screw up the index when we have 50M records? Would we be better off if we used, for example, a type-1 UUID where the leftmost bits were timestamped? Or maybe we should ditch UUIDs entirely and consider auto_increment primary keys?
I'm looking for general thoughts/tips on the performance of different types of UUIDs when they are stored as an index/primary key in MySQL. Thanks!
At my job, we use UUID as PKs. What I can tell you from experience is DO NOT USE THEM as PKs (SQL Server by the way).
It's one of those things that when you have less than 1000 records it;s ok, but when you have millions, it's the worst thing you can do. Why? Because UUID are not sequential, so everytime a new record is inserted MSSQL needs to go look at the correct page to insert the record in, and then insert the record. The really ugly consequence with this is that the pages end up all in different sizes and they end up fragmented, so now we have to do de-fragmentation periodic.
When you use an autoincrement, MSSQL will always go to the last page, and you end up with equally sized pages (in theory) so the performance to select those records is much better (also because the INSERTs will not block the table/page for so long).
However, the big advantage of using UUID as PKs is that if we have clusters of DBs, there will not be conflicts when merging.
I would recommend the following model:
PK INT Identity
Additional column automatically generated as UUID.
This way, the merge process is possible (UUID would be your REAL key, while the PK would just be something temporary that gives you good performance).
NOTE: That the best solution is to use NEWSEQUENTIALID (like I was saying in the comments), but for legacy app with not much time to refactor (and even worse, not controlling all inserts), it is not possible to do.
But indeed as of 2017, I'd say the best solution here is NEWSEQUENTIALID or doing Guid.Comb with NHibernate.
A UUID is a Universally Unique ID. It's the universally part that you should be considering here.
Do you really need the IDs to be universally unique? If so, then UUIDs may be your only choice.
I would strongly suggest that if you do use UUIDs, you store them as a number and not as a string. If you have 50M+ records, then the saving in storage space will improve your performance (although I couldn't say by how much).
If your IDs do not need to be universally unique, then I don't think that you can do much better then just using auto_increment, which guarantees that IDs will be unique within a table (since the value will increment each time)
Something to take into consideration is that Autoincrements are generated one at a time and cannot be solved using a parallel solution. The fight for using UUIDs eventually comes down to what you want to achieve versus what you potentially sacrifice.
On performance, briefly:
A UUID like the one above is 36
characters long, including dashes. If
you store this VARCHAR(36), you're
going to decrease compare performance
dramatically. This is your primary
key, you don't want it to be slow.
At its bit level, a UUID is 128 bits,
which means it will fit into 16 bytes,
note this is not very human readable,
but it will keep storage low, and is
only 4 times larger than a 32-bit int,
or 2 times larger than a 64-bit int.
I will use a VARBINARY(16)
Theoretically, this can work without a
lot of overhead.
I recommend reading the following two posts:
Brian "Krow" Aker's Idle Thoughts - Myths, GUID vs Autoincrement
To UUID or not to UUID ?
I reckon between the two, they answer your question.
I tend to avoid UUID simply because it is a pain to store and a pain to use as a primary key but there are advantages. The main one is they are UNIQUE.
I usually solve the problem and avoid UUID by using dual key fields.
COLLECTOR = UNIQUE ASSIGNED TO A MACHINE
ID = RECORD COLLECTED BY THE COLLECTOR (auto_inc field)
This offers me two things. Speed of auto-inc fields and uniqueness of data being stored in a central location after it is collected and grouped together. I also know while browsing the data where it was collected which is often quite important for my needs.
I have seen many cases while dealing with other data sets for clients where they have decided to use UUID but then still have a field for where the data was collected which really is a waste of effort. Simply using two (or more if needed) fields as your key really helps.
I have just seen too many performance hits using UUID. They feel like a cheat...
Instead of centrally generating unique keys for each insertion, how about allocating blocks of keys to individual servers? When they run out of keys, they can request a new block. Then you solve the problem of overhead by connecting for each insert.
Keyserver maintains next available id
Server 1 requests id block.
Keyserver returns (1,1000)
Server 1 can insert a 1000 records until it needs to request a new block
Server 2 requests index block.
Keyserver returns (1001,2000)
etc...
You could come up with a more sophisticated version where a server could request the number of needed keys, or return unused blocks to the keyserver, which would then of course need to maintain a map of used/unused blocks.
I realize this question is rather old but I did hit upon it in my research. Since than a number of things happened (SSD are ubiquitous InnoDB got updates etc).
In my research I found this rather interesting post on performance:
claiming that due to the randomness of a GUID/UUID index trees can get rather unbalanced. in the MariaDB KB I found another post suggested a solution.
But since than the new UUID_TO_BIN takes care of this. This function is only available in MySQL (tested version 8.0.18) and not in MariaDB (version 10.4.10)
TL;DR: Store UUID as converted/optimized BINARY(16) values.
I would assign each server a numeric ID in a transactional manner.
Then, each record inserted will just autoincrement its own counter.
Combination of ServerID and RecordID will be unique.
ServerID field can be indexed and future select performance
based on ServerID (if needed) may be much better.
The short answer is that many databases have performance problems (in particular with high INSERT volumes) due to a conflict between their indexing method and UUIDs' deliberate entropy in the high-order bits. There are several common hacks:
choose a different index type (e.g. nonclustered on MSSQL) that doesn't mind it
munge the data to move the entropy to lower-order bits (e.g. reordering bytes of V1 UUIDs on MySQL)
make the UUID a secondary key with an auto-increment int primary key
... but these are all hacks--and probably fragile ones at that.
The best answer, but unfortunately the slowest one, is to demand your vendor improve their product so it can deal with UUIDs as primary keys just like any other type. They shouldn't be forcing you to roll your own half-baked hack to make up for their failure to solve what has become a common use case and will only continue to grow.
What about some hand crafted UID? Give each of the thousands of servers an ID and make primary key a combo key of autoincrement,MachineID ???
Since the primary key is generated decentralised, you don't have the option of using an auto_increment anyway.
If you don't have to hide the identity of the remote machines, use Type 1 UUIDs instead of UUIDs. They are easier to generate and can at least not hurt the performance of the database.
The same goes for varchar (char, really) vs. binary: it can only help matters. Is it really important, how much performance is improved?
The main case where UUIDs cause miserable performance is ...
When the INDEX is too big to be cached in the buffer_pool, each lookup tends to be a disk hit. For HDD, this can slow down the access by 10x or worse. (No, that is not a typo for "10%".) With SSDs, the slowdown is less, but still significant.
This applies to any "hash" (MD5, SHA256, etc), with one exception: A type-1 UUID with its bits rearranged.
Background and manual optimization: UUIDs
MySQL 8.0: see UUID_TO_BIN() and BIN_TO_UUID()
MariaDB 10.7 carries this further with its UUID datatype.
I have a 100 million rows, and it's getting too big.
I see a lot of gaps. (since I delete, add, delete, add.)
I want to fill these gaps with auto-increment.
If I do reset it..is there any harM?
If I do this, will it fill the gaps?:
mysql> ALTER TABLE tbl AUTO_INCREMENT = 1;
Potentially very dangerous, because you can get a number again that is already in use.
What you propose is resetting the sequence to 1 again. It will just produce 1,2,3,4,5,6,7,.. and so on, regardless of these numbers being in a gap or not.
Update: According to Martin's answer, because of the dangers involved, MySQL will not even let you do that. It will reset the counter to at least the current value + 1.
Think again what real problem the existence of gaps causes. Usually it is only an aesthetic issue.
If the number gets too big, switch to a larger data type (bigint should be plenty).
FWIW... According to the MySQL docs applying
ALTER TABLE tbl AUTO_INCREMENT = 1
where tbl contains existing data should have no effect:
To change the value of the
AUTO_INCREMENT counter to be used for
new rows, do this:
ALTER TABLE t2 AUTO_INCREMENT = value;
You cannot reset the counter to a
value less than or equal to any that
have already been used. For MyISAM, if
the value is less than or equal to the
maximum value currently in the
AUTO_INCREMENT column, the value is
reset to the current maximum plus one.
For InnoDB, if the value is less than
the current maximum value in the
column, no error occurs and the
current sequence value is not changed.
I ran a small test that confirmed this for a MyISAM table.
So the answers to you questions are: no harm, and no it won't fill the gaps. As other responders have said: a change of data type looks like the least painful choice.
Chances are you wouldn't gain anything from doing this, and you could easily screw up your application by overwriting rows, since you're going to reset the count for the IDs. (In other words, the next time you insert a row, it'll overwrite the row with ID 1, and then 2, etc.) What will you gain from filling the gaps? If the number gets too big, just change it to a larger number (such as BIGINT).
Edit: I stand corrected. It won't do anything at all, which supports my point that you should just change the type of the column to a larger integer type. The maximum possible value for a BIGINT is 2^64, which is over 18 quintillion. If you only have 100 million rows at the moment, that should be plenty for the foreseeable future.
I agree with musicfreak... The maximum for an integer (int(10)) is 4,294,967,295 (unsigned ofcoarse). If you need to go even higher, switching to BIGINT brings you up to 18,446,744,073,709,551,615.
Since you can't change the next auto-increment value, you have other options. The datatype switch could be done, but it seems a little unsettling to me since you don't actually have that many rows. You'd have to make sure your code can handle IDs that large, which may or may not be tough for you.
Are you able to do much downtime? If you are, there are two options I can think of:
Dump/reload the data. You can do this so it won't keep the ID numbers. For example you could use a SELECT ... INTO to copy the data, sans-IDs, to a new table with identical DDL. Then you drop the old table and rename the new table to the old name. Depending on how much data there is, this could take a noticeable about of time (and temporary disk space).
You could make a little program to issue UPDATE statements to change the IDs. If you let that run slowly, it would "defragment" your IDs over time. Then you could temporarily stop the inserts (just a minute or two), update the last IDs, then restart it. After updating the last IDs you can change the AUTO_INCREMENT value to be the next number and your hole will be gone. This shouldn't cause any real downtime (at least on InnoDB), but it could take quite a while depending on how aggressive your program is.
Of course, both of these ignore referential integrity. I'm assuming that's not a problem (log statements that aren't used as foreign keys, or some such).
Does it really matter if there are gaps?
If you really want to go back and fill them, you can always turn off auto increment, and manually scan for the next available id every time you want to insert a row -- remembering to lock the table to avoid race conditions, of course. But it's a lot of work to do for not much gain.
Do you really need a surrogate key anyway? Depending on the data (you haven't mentioned a schema) you can probably find a natural key.