I'm building a Web Application that is connected to a MySQL database.
I've got two huge tables containing each about 40 millions rows at the moment, and they are receiving new rows everyday (which adds ~ 500 000-1000 000 rows everyday).
The process to add new rows runs during the night, while no one can use the application, and the new rows' content depends on the result of some basic SELECT queries on the current database.
In order to get the result of those SELECT statement fast enough, I'm using simple indexes (one column per index) on each column that appears at least once in a WHERE clause.
The thing is, during the day, some totally different queries are run against those tables, including some "range WHERE clause" (SELECT * FROM t1 WHERE a = a1 AND b = b1 AND (date BETWEEN d1 AND d2)).
I found on stack this very helpful mini-cookbook that advises you on which INDEXes you should use depending on how the database is queried: http://mysql.rjweb.org/doc.php/index_cookbook_mysql
They advice to use compound index: in my example query above it would give INDEX(a, b, date).
It indeed increased the speed of the queries run during the day (from 1 minute to 8 seconds so I was truly happy).
However, with those compound indexes, the required time to add new rows during the night totally explode (it would take more than one day to add the daily content).
Here is my question: would that be ok to drop all the indexes every night, add the new content, and set back up the daily indexes?
Or would that be dangerous since indexes are not meant to be rebuilt every day, especially on such big tables?
I know such an operation would take approximately two hours in total (drop and recreate INDEXes).
I am aware of the existence of ALTER TABLE table_name DISABLE KEYS; but I'm using InnoDB and I believe it is not made to work on InnoDB table.
I believe you have answered your own question: You need the indexes during the day, but not at night. Given what you describe, you should drop the indexes for the bulk inserts at night and re-create them afterwards. Dropping indexes for data loads is not unheard of, and seems appropriate in your case.
I would ask about how you are inserting new data. One method is to insert the values one row at a time. Another is to put the values into a temporary table (with no index) and do a bulk insert:
insert into bigtable( . . .)
select . . .
from smalltable;
These have different performance characteristics. You might find that using a single insert (if you are not already doing so) is fast enough for your purposes.
A digression... PARTITIONing by date should be very useful for you since you are deleting things over a year ago. I would recommend PARTITION BY RANGE(TO_DAYS(...)) and breaking it into 14 or 54 partitions (months or weeks, plus some overhead). This will eliminate the time it takes to delete the old rows, since DROP PARTITION is almost instantaneous.
More details are in my partition blog. Your situation sounds like both Use case #1 and Use case #3.
But back to your clever idea of dropping and rebuilding indexes. To others, I point out the caveat that you have the luxury of not otherwise touching the table for long enough to do the rebuild.
With PARTITIONing, all the rows being inserted will go into the 'latest' partition, correct? This partition is a lot smaller than the entire table, so there is a better chance that the indexes will fit in RAM, thereby be 10 times as fast to update (without rebuilding the indexes). If you provide SHOW CREATE TABLE, SHOW TABLE STATUS, innodb_buffer_pool_size, and RAM size, I can help you do the arithmetic to see if your 'last' partition will fit in RAM.
A note about index updates in InnoDB -- they are 'delayed' by sitting in the "Change buffer", which is a portion of the buffer_pool. See innodb_change_buffer_size_max, available since 5.6. Are you using that version, or newer? (If not, you ought to upgrade, for many reasons.)
The default for that setting is 25, meaning that 25% of the buffer_pool is set aside for pending updates to indexes, as caused by INSERT, etc. That acts like a "cache", such that multiple updates to the same index block are held there until they get bumped out. A higher setting should make index updates hit the disk less often, hence finish faster.
Where I am heading with this... By increasing this setting, you would make the inserts (direct, not rebuild) more efficient. I'm thinking that this might speed it up:
Just before the nightly INSERTs:
innodb_change_buffer_size_max = 70
innodb_old_blocks_pct = 10
Soon after the nightly INSERTs:
innodb_change_buffer_size_max = 25
innodb_old_blocks_pct = 37
(I am not sure about that other setting, but it seems reasonable to push it out of the way.)
Meanwhile, what is the setting of innodb_buffer_pool_size? Typically, it should be 70% of available RAM.
In a similar application, I had big, hourly, dumps to load into a table, and a 90-day retention. I stretched my Partition rules by having 90 daily partitions and 24 hourly partitions. Every night, I spent a lot of time (but less than an hour) doing REORGANIZE PARTITION to turn the 24 hourly partitions into a new daily (and dropping the 90-day-old partition). During each hour, the load had the added advantage that nothing else was touching the 1-hour partition -- I could do normalization, summarization, and loading all in 7 minutes. The entire 90 days fit in 400GB. (Side note: a large number of partitions is a performance killer until 8.0; so don't even consider daily partitions for you 1-year retention.)
The Summary tables made so that 50-minute queries (in the prototype) shrank to only 2 seconds. Perhaps you need a summary table with PRIMARY KEY (a, b, date)? That will let you get rid of such an index on the 'Fact' table. Oops, that eliminates the entire premise of your original question ! See the links at the bottom of my blogs; look for "Summary Tables". A general rule: Don't have any indexes (other than the PRIMARY KEY) on the Fact table; use Summary tables for things that need messier indexes.
Related
I'm using MySQL InnoDB, one of the most important tables has over 700 million records with totally 23 actively used indexes.
I am trying to delete the records in a batch of 2000 based on the record date (and order by on primary key column). Each date has around 6 million records, I delete it date by date with limit 2000. Each batch takes around 25 seconds to complete. Since it is a Production database, I want this delete operation to complete faster. Is there a better way to do this?
There are many solutions. See http://mysql.rjweb.org/doc.php/deletebig .
Which takes 25 seconds? A batch of 2000 rows? Or several of those? Are they in a single, big, transaction? Lots of little transactions would be slower, but less invasive. And there would be no ACID problem since re-deleting the same 'date' should be idempotent.
If the table is "locked" at some level for 25 seconds, then I understand your concern. If it is a bunch of sub-second deletes, then does it really matter that it takes a long time?
Furthermore, instead of deleting once a day, you could delete once an hour. This might decrease the tasks to 2 seconds instead of 25.
25 indexes is terribly large. Please provide SHOW CREATE TABLE. It is all too common that there are "redundant" indexes that could (should) be dropped. The main example is INDEX(a,b) takes care of INDEX(a) (but not b), so if you have both, drop the latter.
It may be impractical to make the change now, but PARTITION BY RANGE(TO_DAYS(...)) lets you DROP PARTITION, which is virtually instantaneous. (Adding partitioning to a 700M row table would take a long time.)
I have a table that contains 1.5 million rows, has 39 columns, contains sales data of around 2 years, and grows every day.
I had no problems with it until we moved it to a new server, we probably have less memory now.
Queries are currently taking a very long time. Someone suggested partitioning the large table that is causing most of the performance issues but I have a few questions.
Is it wise to partition the table I described and is it
likely to improve its performance?
If I do partition it, will
I have to make changes to my current INSERT or SELECT statements or
will they continue working the same way?
Does the partition
take a long time to perform? I worry that with the slow performance,
something would happen midway through and I would lose the data.
Should I be partioning it to years or months? (we usually
look at the numbers within the month, but sometimes we take weeks or
years). And should I also partition the columns? (We have some
columns that we rarely or never use, but we might want to use them
later)
(I agree with Bill's answer; I will approach the Question in a different way.)
When is it time to partion my tables?
Probably never.
is it likely to improve its performance?
It is more likely to decrease performance a little.
I have a table that contains 1.5 million rows
Not big enough to bother with partitioning.
Queries are currently taking a very long time
Usually that is due to the lack of a good index, probably a 'composite' one. Secondly is the formulation of the query. Please show us a slow query, together with SHOW CREATE TABLE.
data of around 2 years, and grows every day
Will you eventually purge "old" data? If so, the PARTITION BY RANGE(TO_DAYS(..)) is an excellent idea. However, it only helps during the purge. This is because DROP PARTITION is a lot faster than DELETE....
we probably have less memory now.
If you are mostly looking at "recent" data, then the size of memory (cf innodb_buffer_pool_size) may not matter. This is due to caching. However, it sounds like you are doing table scans, perhaps unnecessarily.
will I have to make changes to my current INSERT or SELECT
No. But you probably need to change what column(s) are in the PRIMARY KEY and secondary key(s).
Does the partition take a long time to perform?
Slow - yes, because it will copy the entire table over. Note: that means extra disk space, and the partitioned table will take more disk.
something would happen midway through and I would lose the data.
Do not worry. The new table is created, then a very quick RENAME TABLE swaps it into place.
Should I be partioning it to years or months?
Rule of thumb: aim for about 50 partitions. With "2 years and growing", a likely choice is "monthly".
we usually look at the numbers within the month, but sometimes we take weeks or years
Smells like a typical "Data Warehouse" dataset? Build and incrementally augment a "Summary table" with daily stats. With that table, you can quickly get weekly/monthly/yearly stats -- possibly 10 times as fast. Ditto for any date range. This also significantly helps with "low memory".
And should I also partition the columns? (We have some columns that we rarely or never use, but we might want to use them later)
You should 'never' use SELECT *; instead, specify the columns you actually need. "Vertical partitioning" is the term for your suggestion. It is sometimes practical. But we need to see SHOW CREATE TABLE with realistic column names to discuss further.
More on partitioning: http://mysql.rjweb.org/doc.php/partitionmaint
More on Summary tables: http://mysql.rjweb.org/doc.php/summarytables
In most circumstances, you're better off using indexes instead of partitioning as your main method of query optimization.
The first thing you should learn about partitioning in MySQL is this rule:
All columns used in the partitioning expression for a partitioned table must be part of every unique key that the table may have.
Read more about this rule here: Partitioning Keys, Primary Keys, and Unique Keys.
This rule makes many tables ineligible for partitioning, because you might want to partition by a column that is not part of the primary or unique key in that table.
The second thing to know is that partitioning only helps queries using conditions that unambiguously let the optimizer infer which partitions hold the data you're interested in. This is called Partition Pruning. If you run a query that could find data in any or all partitions, MySQL must search all the partitions, and you gain no performance benefit compared to have a regular non-partitioned table.
For example, if you partition by date, but then you run a query for data related to a specific user account, it would have to search all your partitions.
In fact, it might even be a little bit slower to use partitioned tables in such a query, because MySQL has to search each partition serially.
You asked how long it would take to partition the table. Converting to a partitioned table requires an ALTER TABLE to restructure the data, so it takes about the same time as any other alteration that copies the data to a new tablespace. This is proportional to the size of the table, but varies a lot depending on your server's performance. You'll just have to test it out, there's no way we can estimate how long it will take on your server.
Data is increasing in one table everyday, it might lower the performance . I was thinking if I can create a trigger which move table A into A1 and create a new table A every a period of time, so that insert or update could be faster in table A. Is this the right way to save performance ? If not, what should I do ?
(for example, insert or update 1000 rows per second in table A, how is the performance after 3 years ?)
We are designing softwares for a factory. There are product lines which pcb boards are made on. We need to insert almost 60 pcb records per second for years. (1000 rows seem to be exaggerated)
First, you are talking about several terabytes for a single table. Is your disk that big? Yes, MySQL can handle that big a table.
Will it slow down? It depends on
The indexes. If you have 'random' indexes, the INSERTs will slow down to about 1 insert per disk hit. On a spinning HDD, that is only about 100 per second. SSD might be able to handle 1000/sec. Please provide SHOW CREATE TABLE.
Does the table have an AUTO_INCREMENT? If so, it needs to be BIGINT, not INT. But, if possible, get rid of it all together (to save space). Again, let's see the SHOW.
"Point" queries (load one row via an index) are mostly unaffected by the size of the table. They will be about twice as slow in a trillion-row table as in a million-row table. A point query will take milliseconds or tens of milliseconds; no big deal.
A table scan will take hours or days; hopefully you are not doing that.
A billion-row scan of part of the table will take days or weeks unless you are using the PRIMARY KEY or have a "covering" index. Let's see the queries and the SHOW.
The best technique is not to store the data. Summarize it as it arrives, save the summaries, then toss the raw data. (OK, you might store the raw in a csv file just in case you need to build a new summary table or fix a bug in an existing one.)
Having a few summary tables instead of the raw data would shrink the data to under 1TB and allow the relevant queries to run 10 times as fast. (OK, point queries would be only slightly faster.)
PARTITIONing (or otherwise splitting up the table)? It depends. Let's see the queries and the SHOW. In many situations, PARTITIONing does not speed up anything.
Will you be deleting or modifying existing rows? I hope not. That adds more dimensions of problems. If, on the other hand, you need to purge 'old' data, then that is an excellent use for PARTITIONing. For 3 years' worth of data, I would PARTITION BY RANGE(TO_DAYS(..)) and have monthly partitions. Then a monthly DROP PARTITION would be very fast.
Very Huge data may decrease the performance of server, So there is a way to handle this :
1) you have to create another table to store archive data ( old data ) using Archive storage mechanism . ( https://dev.mysql.com/doc/refman/8.0/en/archive-storage-engine.html )
2) create MySQL job/scheduler to move older records to archive table. schedule in timeslot
when server is maximum idle.
3) after moving older records to archive table, re-index the original table.
this will serve the purpose of performance.
It is unlikely that 1000 row tables perform sufficiently poorly that doing a table copy every once in a while is an overall net gain. And anyway, what would the new table have that the old one did not which would improve performance?
The key to having tables perform efficiently is intelligent table design and management of indexes. That is how zillion row tables are effective in geospatial work, library catalogs, astronomy, and how internet search engines find useful data, etc.
Each index defined does cause more mysql impact especially at row insert time. Assuming there are more reads than inserts, this is an advantage because most queries are rapidly completed thanks to a suitable index.
Indexes are best defined with a thorough understanding of the queries made against the table—both in quality and quantity. And, if there is any tendency for the nature of the queries to trend over months or years, then the indexes would need additions, modifications, or—yes—even deletions.
It seems to me there is something inherently wrong with the way you are using MySQL to begin with.
A database system is supposed to manage data that is required by your application in order for it to work. If you think flushing the table every so often is something acceptable, then that doesn't seem to be the case.
Perhaps you are better off just using log files. Split them by date, delete old ones if and when you decide they are no longer relevant or need the disk space. It's even safer to do that way from a recovery perspective.
If you need a better suggestion, then improve your question to include exactly what you are trying to accomplish so we can help you with it.
I have a database with a single table that keeps track of user state. When I'm done handling the row, its no longer necessary to keep it in the database and can be deleted.
Now lets say I wanted to keep track of the row instead of deleting it (for historical purposes, analytics, etc). Would it be better to:
Leave the data in the same table and mark the row as 'used' (with an extra column or something like that)
Delete the row from the table and insert it into a separate table that is created only for historical purposes
For choice #1, I wonder if leaving the unnecessary rows in the database will start to affect query performance. (All of my queries are on indexed columns, so maybe this doesn't matter?)
For choice #2, I wonder if the constant deleting of rows will end up causing problems such as fragmentation?
Query performance will be better in the long run:
What is happening with forever inserts:
The table grows, indexes grow, index performance (lookup) is decreases with the size of the table, especially insert performance is hurt.
What is happening with delete:
Table pages get fragmented, so the deleted space is not re-used 100% as expected, more near 50% in MySQL. So the table still grows to about twice the size you might expect for your amount of data. The index gets fragmented and becomes lob sided: It contains your new data but also the structure for your old data. It depends on the structure of your data on how bad this gets. This situation however stabilizes at a certain performance. This performance point has 2 benefits:
1) The table is more limited in size, so potential full table scans are faster
2) Your performance is predictable.
Due to the fragmentation however this performance point is not equal to about twice your data amount, it tends to be a bit worse (benchmark it to see yourself). The benefit of the delete scenario is however since you have a smaller data set, that you might be able to rebuild your index once every reasonable period, thus improving your performance.
Alternatives
There are two alternatives you can look at to improve performance:
Switch to MariaDB: This gains about 8% performance on large datasets (my observation, dataset just about 200GB compressed data)
Look at partitioning: If you have a handy partitioning parameter, you can create a series of "small tables" for you and prevent logic for delete, rebuild and historic data management. This might give you the best performance profile.
If most of the table is flagged as deleted, you will be stumbling over them as you look for the non-deleted records. Adding is_deleted to many of the indexes is likely to help.
If you are deleting records purely on age, then PARTITION BY RANGE(TO_DAYS(...)) is an excellent way to build the table. The DROP TABLE is instantaneous and the ALTER TABLE ... REORGANIZE ... to create a new week (or month or ...) partition is also instantaneous. See my blog for details.
If you "move" records to another table, then the table will not shrink very fast due to fragmentation. If you have enough disk space, this is not a bug deal. If some queries need to see both current and archived records, use UNION ALL; it is pretty easy and efficient.
I have a table that has millions of records, and they utilize EFF_FROM and EFF_TO date fields to version the records.
99% of the time, when this table is queried by an application, it is only concerned with records that have an EFF_TO of 2099-12-31, or records that are active and not historical.
I copied just the active records to a test version of the table and the application's SELECT query went from 60 seconds to 3 seconds.
I don't necessarily want to partition every EFF_TO date. I don't want to add that overhead especially to processes that populate the table. I only want the optimization for querying records with 2099-12-31, and I want the performance to be instant.
Is there a straightforward way to do this? Or do I have to resort to creating an active table and a historical table?
Partition like function for a single set of data?
This is something of any oxymoron, however you are asking about partitioning into two sets of data, one where EFF_TO is in the future and one where it is in the past.
have an EFF_TO of 2099-12-31
Design fault - these should be null.
If they were null the the partitioning would be simple. As it stands you will have to drop and recreate the partitions - which is rather an expensive operation (have a look at tools for doing online schema updates).
You could minimize the impact by creating multiple partitions defining the period around NOW then adding an extra one onto the end of and removing one from the beginning at regular intervals.
application's SELECT query went from 60 seconds to 3 seconds.
There are lots of other reasons why the performance improved than just the size of the table
if it's doing a full table table scan, this is a design fault in the application.
You're indexes may not be as up to date as they should be
the logical structure of the indexes may be unbalanced and need optimized
the physical structure of the table and indexes many be fragmented and need optimized