CQL Execution [returns instantly, assuming uses clustering key index]:
cqlsh:stats> select count(*) from events where month='2015-04' and day = '2015-04-02';
count
-------
5447
Presto Execution [takes around 8secs]:
presto:default> select count(*) as c from cassandra.stats.events where month = '2015-04' and day = timestamp '2015-04-02';
c
------
5447
(1 row)
Query 20150228_171912_00102_cxzfb, FINISHED, 1 node
Splits: 2 total, 2 done (100.00%)
0:08 [147K rows, 144KB] [17.6K rows/s, 17.2KB/s]
Why should presto get to process 147K rows when cassandra itself responds with just 5447 rows for the same query [I tried select * too]?
Why presto is not able to use the clustering key optimization?
I tried all possible values like timestamp, date, different formats of dates. Not able to see any effect on number of rows being fetched.
CF Reference:
CREATE TABLE events (
month text,
day timestamp,
test_data text,
some_random_column text,
event_time timestamp,
PRIMARY KEY (month, day, event_time)
) WITH comment='Test Data'
AND read_repair_chance = 1.0;
Added event_timestamp too as a constraint in response to Dain's answer
presto:default> select count(*) from cassandra.stats.events where month = '2015-04' and day = timestamp '2015-04-02 00:00:00+0000' and event_time = timestamp '2015-04-02 00:00:34+0000';
_col0
-------
1
(1 row)
Query 20150301_071417_00009_cxzfb, FINISHED, 1 node
Splits: 2 total, 2 done (100.00%)
0:07 [147K rows, 144KB] [21.3K rows/s, 20.8KB/s]
The Presto engine will pushdown simple WHERE clauses like this to a connector (you can see this in the Hive connector), so the question is, why does the Cassandra connector not take advantage of this. To see why, we'll have to look at the code.
The pushdown system first interacts with connectors in the ConnectorSplitManager.getPartitions(ConnectorTableHandle, TupleDomain) method, so looking at the CassandraSplitManager, I see it is delegating the logic to getPartitionKeysSet. This method looks for a range constraint (e.g., x=33 or x BETWEEN 1 AND 10) for every column in the primary key, so in your case, you would need to add a constraint on event_time.
I don't know why the code insists on having a constraint on every column in the primary key, but I'd guess that it is a bug. It should be easy to tweak this code to remove that constraint.
Related
Assume data with pk (text), start (int), end (int), extra_data(text).
Query is: given a pk (e.g. 'pk1') and a range (e.g [1000, 2000]), find all rows for 'pk1' which intersect that range. This (sql) logically translates to WHERE pk=pk1 AND end>=1000 AND start<=2000 (intersection condition)
Notice this is NOT the same as the more conventional query of:
all rows for pk1 where start>1000 and start<2000
If I define a table with end as part of the clustering key:
CREATE TABLE test1 (
pk text,
start bigint,
end bigint,
extra_data text,
PRIMARY KEY ((pk), start, end)
)...
Then Cassandra does not allow the query:
select * from test1 where pk='pk1' and start < 2000 and end > 1000;
with "Clustering column "end" cannot be restricted (preceding column "start" is restricted by a non-EQ relation)"
Why does Cassandra not allow further filtering to limit ranged rows (forces to do this filter with results application-side).
A second try would be to remove 'end' from clustering columns:
CREATE TABLE test1 (
pk text,
start bigint,
end bigint,
extra_data text,
PRIMARY KEY ((pk), start)
)...
Then Cassandra warns the query:
select * from test1 where pk='pk1' and start < 2000 and end > 1000;
with "Cannot execute this query as it might involve data filtering and thus may have unpredictable performance. If you want to execute this query despite the performance unpredictability, use ALLOW FILTERING"
Here I would like to understand if I can safely add the ALLOW FILTERING and be assured Cassandra will perform the scan only of 'pk1'.
Using cqlsh 5.0.1 | Cassandra 3.11.3
Actually, I think you made the fatal mistake of designing your table first and then trying to adapt the application query to fit the table design.
In Cassandra data modelling, the primary principle is to always start by listing all your application queries THEN design a table for each of those application queries -- not the other way around.
Let's say I have an IoT use case where I have sensors collecting temperature readings once a day. If my application needs to retrieve the readings from the last 7 days from a sensor, the app query is:
Get the temperature for the last 7 days for sensor X
Assuming today is October 25, a more SQL-like representation of this app query is:
SELECT temperature FROM table
WHERE sensor = X
AND reading_date >= 2022-10-18
AND reading_date < 2022-10-25
This means that we need to design the table such that:
it is partitioned by sensor, and
the data is clustered by date.
The table schema would look like:
CREATE TABLE readings_by_sensor (
sensor text,
reading_date date,
temp float,
PRIMARY KEY (sensor, reading_date)
)
We can then perform a range query on the date:
SELECT temperature FROM readings_by_sensor
WHERE sensor = ?
AND reading_date >= 2022-10-18
AND reading_date < 2022-10-25
You don't need two separate columns to represent the start and end range because. Cheers!
I have a data modeling question. In my application I'm reading data from a few different sensors and storing it in Cassandra. The sensors generate new values at very different rates: Some every other second, some every other month.
Furthermore, the assumption is that a value stays valid until the next one is encountered. Example: Sensor 1 sent a value of 500 at 10s after EPOCH and a value of 1000 at 20s after EPOCH. The valid value for 15s after EPOCH would need to be 500.
Since some rates are going to be high and I don't want unbounded partitions, I want to apply bucketing. I'm thinking about modeling my data like this:
CREATE TABLE sensor_data (
sensor_id text,
some_timing_bucket date,
measured_at time,
value double
PRIMARY KEY ((sensor_id, some_timing_bucket), measured_at)
) WITH CLUSTERING ORDER BY (measured_at DESC);
The usual queries the application would need to serve are "give me the data of the last 5/15 minutes/1 day", so I would choose the some_timing_bucket accordingly. Maybe even have multiple tables with different bucket sizes.
What I cannot wrap my head around is this: Consider I choose one day as bucketing interval. Now I want to retrieve the current value of a sensor that hasn't updated in ten days. There will be no partition for today, so on my application layer I would need to send nine queries that yield nothing until I have gone far enough back in time to encounter the value that is currently valid. That doesn't sound very efficient and I'd appreciate any input on how to model this.
Side note: This would not be an issue if all data for the same sensor was in the same partition: Just ask for all the points with a timestamp less than the beginning of the ranged query and limit the results to one. But that's not feasible because of the unbounded partition.
There is a much simpler way to model your data by using one-day buckets. Something like:
CREATE TABLE sensor_data_by_day (
sensor_id text,
year int,
month int,
day int,
measured_at timestamp,
value double,
PRIMARY KEY ((sensor_id, year, month, day), measured_at)
) WITH CLUSTERING ORDER BY (measured_at DESC)
If a sensor measures a data point every second, then there are 86,400 maximum possible values for a single day (60 secs x 60 mins * 24 hrs). 86K rows per partition is still manageable.
If today is 17 August 2022 and you wanted to retrieve the data for the previous day, the query would be:
SELECT value FROM sensor_data_by_day
WHERE sensor_id = ?
AND year = 2022
AND month = 8
AND day = 16
Assuming it is currently 08:30:00 GMT on the 17th of August (1660725000000 ms since epoch), to retrieve the data for the last 15 minutes (900 secs ago or 1660724100000 ms):
SELECT value FROM
WHERE sensor_id = ?
AND year = 2022
AND month = 8
AND day = 17
AND measured_at > 1660724100000
I think you'll find that it is easier to work with timestamps because it provides a bit more flexibility when it comes to doing range queries. Cheers!
you can do this with a simpler table like this:
CREATE TABLE sensor_data (
sensor_id text,
day_number_from_1970 int,
measured_at timestamp,
value double,
PRIMARY KEY ((sensor_id, day_number_from_1970), measured_at)
) WITH CLUSTERING ORDER BY (measured_at DESC)
and you can query data like that:
SELECT value
FROM sensor_data
WHERE sensor_id = some_sensor_id
AND day_number_from_1970 = day_number
AND measured_at > start_time
AND measured_at < end_time
with a single int column, you should less data on disk and get results well
I am looking for a good way to store time specific data in cassandra.
Each entry can look like (start_time, value). Later, I would like to retrieve the current value.
Logic of retrieving current value is like following.
Find all rows with start_time<=current_time.
Then find the value with maximum start_time from the rows obtained in the first step.
PS:- Edited the question to make it more clear
The exact requirements are not possible. But we can get close to it with one more column.
First, to be able to use <= operator, your start_time column need to be the clustering key of your table.
Then, you need a different partition key. You could choose a fixed value but it could bring problems when the partition will have too many rows. Then you should better use something like the year or the month of the start_time.
CREATE TABLE time_specific_table (
year bigint,
start_time timestamp,
value text,
PRIMARY KEY((year), start_time)
) WITH CLUSTERING ORDER BY (start_time DESC);
The problem is that when you will query the table, you will need to know the value of the partition key :
Find all rows with start_time<=current_time
SELECT * FROM time_specific_table
WHERE year = :year AND start_time <= :time;
select the value with maximum start_time
SELECT * FROM time_specific_table
WHERE year = :year LIMIT 1;
Create two separate table like below :
CREATE TABLE data (
start_time timestamp,
value int,
PRIMARY KEY(start_time, value)
);
CREATE TABLE current_value (
partition int PRIMARY KEY,
value int
);
Now you have to insert data into both table, to insert data into second table use a static value like 1
INSERT INTO current_value(partition, value) VALUES(1, 10);
Now In current value table your data will be upsert and You will get latest value whenever you select.
I have the requirement to forward data at certain intervals from my system to an external system. To do this, I already stored all rows in a table. Already forwarded data should not be exported again.
The idea is to memorize the last export time on client side and export the following records the next time. Old rows are deleted after a successful export.
CREATE TABLE export(
id int,
import_date_time timestamp,
data text,
PRIMARY KEY (id, import_date_time)
) WITH CLUSTERING ORDER BY (import_date_time DESC)
insert into export(id, import_date_time, data) values (1, toUnixTimestamp(now()), 'content')
select * from export where id = 1 and import_date_time > '2017-03-30 16:22:37'
delete from export where id = 1 and import_date_time <= '2017-03-30 16:22:37'
Has anyone already implemented similar or do you have a different
solution?
If possible, I do not need an id for the request because I want to
export all data
If you used fixed partition key value (id = 1), then all the insert, select and delete will happen on a same node (If RF=1) over and over. And also for every delete cassandra create a tombstone entry, when you execute select query cassandra needs to merge each entry. So your select query performance will degrade.
So instead of having fixed value, use dynamic value like the below one :
CREATE TABLE export(
hour int,
day int,
month int,
year int,
import_date_time timestamp,
data text,
PRIMARY KEY ((hour, day, month, year), import_date_time)
) WITH CLUSTERING ORDER BY (import_date_time DESC);
Here you can insert the value of hour, day, month, year extracted from import_date_time
You need to take care of two case When selecting data :
Previous export time and current export time both at same hour.
Both time are not inside same hour.
For case one you need only one query and for case two you have to execute two query.
Example Query :
SELECT * FROM export WHERE hour = 16 AND day = 30 AND month = 3 AND year = 2017 AND import_date_time > '2017-03-30 16:22:37';
I have a column family with primary key definition like this:
...
PRIMARY KEY ((website_id, item_id), user_id, date)
which will be queried using queries such as:
SELECT * FROM myCF
WHERE website_id = 30 AND item_id = 10
AND user_id = 0 AND date > 'some_date' ;
However, I'd like to keep my column family ordered by date only, such as SELECT date FROM myCF ; would return the most recent inserted date.
Due to the order of clustering columns, what I get is an order per user_id then per date.
If I change the primary key definition to:
PRIMARY KEY ((website_id, item_id), date, user_id)
I can no longer run the same query, as date must be restricted is user_id is.
I thought there might be some way to say:
...
PRIMARY KEY ((website_id, shop_id), store_id, date)
) WITH CLUSTERING ORDER BY (store_id RANDOMPLEASE, date DESC) ;
But it doesn't seem to exist. Worst, maybe this is completely stupid and I don't get why.
Is there any ways of achieving this? Am I missing something?
Many thanks!
Your query example restricts user_id so that should work with the second table format. But if you are actually trying to run queries like
SELECT * FROM myCF
WHERE website_id = 30 AND item_id = 10
AND date > 'some_date'
Then you need an additional table which is created to handle those queries, it would only order on Date and not on user id
Create Table LookupByDate ... PRIMARY KEY ((website_id, item_id), date)
In addition to your primary query, if all you try to get is "return the most recent inserted date", you may not need an additional table. You can use "static column" to store the last update time per partition. CASSANDRA-6561
It probably won't help your particular case (since I imagine your list of all users is unmanagably large), but if the condition on the first clustering column is matching one of a relatively small set of values then you can use IN.
SELECT * FROM myCF
WHERE website_id = 30 AND item_id = 10
AND user_id IN ? AND date > 'some_date'
Don't use IN on the partition key because this will create an inefficient query that hits multiple nodes putting stress on the coordinator node. Instead, execute multiple asynchronous queries in parallel. But IN on a clustering column is absolutely fine.