Subject: Re: Postgres performance comparing GCP and AWS
To: pgsql-performance@lists.postgresql.org
References: 
 <CAMzrzcaeYtkJN+tQWpKhpdaAeD-EjxJ3Pfcpfy-QsOKnrdy17w@mail.gmail.com>
From: Gunther Schadow <raj@gusw.net>
Message-ID: <3dcfe362-b9fd-5168-863d-ef5651b74a71@gusw.net>
Date: Wed, 24 Feb 2021 08:10:20 -0500
User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:78.0) Gecko/20100101
 Thunderbird/78.7.1
MIME-Version: 1.0
In-Reply-To: 
 <CAMzrzcaeYtkJN+tQWpKhpdaAeD-EjxJ3Pfcpfy-QsOKnrdy17w@mail.gmail.com>
Content-Type: multipart/alternative;
 boundary="------------4D75A144FFF51D2F20F412A7"
Content-Language: en-US
Precedence: bulk

This is a multi-part message in MIME format.
--------------4D75A144FFF51D2F20F412A7
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Transfer-Encoding: 8bit

Hi Maurici,

in my experience the key factor about speed in big queries is sequential 
scan. There is a huge variance in how the system is tuned. In some cases 
I cannot read more than 10 MB/s, in others I get to expect 20-40 MB/s. 
But then, when things are tuned well and the parallel workers set in, I 
see the throughput spike to 100-200 MB/s.

You may have to enable the parallel workers in your postgresql.conf

So, to me, this is what you want to check first. While the query runs, 
have both iostat and top running, with top -j or -c or -a or whatever it 
is on that particular OS to see the detail info about the process. 
Perhaps even -H to see threads.

Then you should see good flow with high read speed and reasonable CPU 
load %. If you get low read speed and low CPU that is a sign of IO 
blockage somewhere. If you get high CPU and low IO, that's a planning 
mistake (the nested loop trap). You don't have that here apparently. But 
index scans I have seen with much worse IO throughput than seq table 
scans. Not sure.

Also, on AWS you need to be sure you have enough IOPS provisioned on 
your EBS (I use gp3 now where you can have up to 10k IOPS) and also 
check bus throughput of the EC2 instance. Needless to say you don't want 
a t* instance where you have a limited burst CPU capacity only.

regards,
-Gunther

On 2/23/2021 1:12 PM, Maurici Meneghetti wrote:
> Hi everyone,
>
> I have 2 postgres instances created from the same dump (backup), one 
> on a GCP VM and the other on AWS RDS. The first instance takes 18 
> minutes and the second one takes less than 20s to run this simples query:
> SELECT "Id", "DateTime", "SignalRegisterId", "Raw" FROM 
> "SignalRecordsBlobs" WHERE "SignalSettingId" = 103 AND "DateTime" 
> BETWEEN '2019-11-28T14:00:12.540200000' AND 
> '2020-07-23T21:12:32.249000000';
> I’ve run this query a few times to make sure both should be reading 
> data from cache.
> I expect my postgres on GPC to be at least similar to the one managed 
> by AWS RDS so that I can work on improvements parallelly and compare.
>
> *DETAILS:
> Query explain for Postgres on GCP VM:
> *Bitmap Heap Scan on SignalRecordsBlobs SignalRecordsBlobs 
>  (cost=18.80..2480.65 rows=799 width=70) (actual time=216.766..776.032 
> rows=5122 loops=1)
>     Filter: (("DateTime" >= \'2019-11-28 14:00:12.5402\'::timestamp 
> without time zone) AND ("DateTime" <= \'2020-07-23 
> 21:12:32.249\'::timestamp without time zone))
>     Heap Blocks: exact=5223
>     Buffers: shared hit=423 read=4821
>   ->  Bitmap Index Scan on IDX_SignalRecordsBlobs_SignalSettingId 
>  (cost=0.00..18.61 rows=824 width=0) (actual time=109.000..109.001 
> rows=5228 loops=1)
>           Index Cond: ("SignalSettingId" = 103)
>           Buffers: shared hit=3 read=18
> Planning time: 456.315 ms
> Execution time: 776.976 ms
>
> *Query explain for Postgres on AWS RDS:
> *Bitmap Heap Scan on SignalRecordsBlobs SignalRecordsBlobs 
>  (cost=190.02..13204.28 rows=6213 width=69) (actual time=2.215..14.505 
> rows=5122 loops=1)
>     Filter: (("DateTime" >= \'2019-11-28 14:00:12.5402\'::timestamp 
> without time zone) AND ("DateTime" <= \'2020-07-23 
> 21:12:32.249\'::timestamp without time zone))
>     Heap Blocks: exact=5209
>     Buffers: shared hit=3290 read=1948
>   ->  Bitmap Index Scan on IDX_SignalRecordsBlobs_SignalSettingId 
>  (cost=0.00..188.46 rows=6405 width=0) (actual time=1.159..1.159 
> rows=5228 loops=1)
>           Index Cond: ("SignalSettingId" = 103)
>           Buffers: shared hit=3 read=26
> Planning time: 0.407 ms
> Execution time: 14.87 ms
>
> *PostgreSQL version number running:
> • VM on GCP*: PostgreSQL 11.10 (Debian 11.10-0+deb10u1) on 
> x86_64-pc-linux-gnu, compiled by gcc (Debian 8.3.0-6) 8.3.0, 64-bit
> *• Managed by RDS on AWS:* PostgreSQL 11.10 on x86_64-pc-linux-gnu, 
> compiled by gcc (GCC) 4.8.5 20150623 (Red Hat 4.8.5-11), 64-bit
>
> *How PostgreSQL was installed:
> • VM on GCP*: Already installed when created VM running Debian on 
> Google Console.
> *• Managed by RDS on AWS:* RDS managed the installation.
>
> *Changes made to the settings in the postgresql.conf file:
> *Here are some postgres parameters that might be useful:
> *Instance on VM on GCP (2 vCPUs, 2 GB memory, 800 GB disk):*
> • effective_cache_size: 1496MB
> • maintenance_work_mem: 255462kB (close to 249MB)
> • max_wal_size: 1GB
> • min_wal_size: 512MB
> • shared_buffers: 510920kB (close to 499MB)
> • max_locks_per_transaction 1000
> • wal_buffers: 15320kB (close to 15MB)
> • work_mem: 2554kB
> • effective_io_concurrency: 200
> • dynamic_shared_memory_type: posix
> On this instance we installed a postgres extension called timescaledb 
> to gain performance on other tables. Some of these parameters were set 
> using recommendations from that extension.
>
> *Instance managed by RDS (2 vCPUs, 2 GiB RAM, 250GB disk, 750 de IOPS):*
> • effective_cache_size: 1887792kB (close to 1844MB)
> • maintenance_work_mem: 64MB
> • max_wal_size: 2GB
> • min_wal_size: 192MB
> • shared_buffers: 943896kB (close to 922MB)
> • max_locks_per_transaction 64
>
> *Operating system and version by runing "uname -a":
> • VM on GCP:* Linux {{{my instance name}}} 4.19.0-14-cloud-amd64 #1 
> SMP Debian 4.19.171-2 (2021-01-30) x86_64 GNU/Linux
> *• Managed by AWS RDS:* Aparently Red Hay as shown using SELECT version();
>
> *Program used to connect to PostgreSQL:* Python psycopg2.connect() to 
> create the connection and pandas read_sql_query() to query using that 
> connection.
>
> Thanks in advance

--------------4D75A144FFF51D2F20F412A7
Content-Type: text/html; charset=utf-8
Content-Transfer-Encoding: 8bit

<html>
  <head>
    <meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
  </head>
  <body>
    <p>Hi Maurici, <br>
    </p>
    <p>in my experience the key factor about speed in big queries is
      sequential scan. There is a huge variance in how the system is
      tuned. In some cases I cannot read more than 10 MB/s, in others I
      get to expect 20-40 MB/s. But then, when things are tuned well and
      the parallel workers set in, I see the throughput spike to 100-200
      MB/s. <br>
    </p>
    <p>You may have to enable the parallel workers in your
      postgresql.conf<br>
    </p>
    <p>So, to me, this is what you want to check first. While the query
      runs, have both iostat and top running, with top -j or -c or -a or
      whatever it is on that particular OS to see the detail info about
      the process. Perhaps even -H to see threads. <br>
    </p>
    <p>Then you should see good flow with high read speed and reasonable
      CPU load %. If you get low read speed and low CPU that is a sign
      of IO blockage somewhere. If you get high CPU and low IO, that's a
      planning mistake (the nested loop trap). You don't have that here
      apparently. But index scans I have seen with much worse IO
      throughput than seq table scans. Not sure.</p>
    <p>Also, on AWS you need to be sure you have enough IOPS provisioned
      on your EBS (I use gp3 now where you can have up to 10k IOPS) and 
      also check bus throughput of the EC2 instance. Needless to say you
      don't want a t* instance where you have a limited burst CPU
      capacity only.</p>
    <p>regards,<br>
      -Gunther<br>
    </p>
    <div class="moz-cite-prefix">On 2/23/2021 1:12 PM, Maurici
      Meneghetti wrote:<br>
    </div>
    <blockquote type="cite"
cite="mid:CAMzrzcaeYtkJN+tQWpKhpdaAeD-EjxJ3Pfcpfy-QsOKnrdy17w@mail.gmail.com">
      <meta http-equiv="content-type" content="text/html; charset=UTF-8">
      <div dir="ltr">
        <div>Hi everyone,</div>
        <div><br>
        </div>
        I have 2 postgres instances created from the same dump (backup),
        one on a GCP VM and the other on AWS RDS. The first instance
        takes 18 minutes and the second one takes less than 20s to run
        this simples query:<br>
        SELECT "Id", "DateTime", "SignalRegisterId", "Raw" FROM
        "SignalRecordsBlobs" WHERE "SignalSettingId" = 103 AND
        "DateTime" BETWEEN '2019-11-28T14:00:12.540200000' AND
        '2020-07-23T21:12:32.249000000';<br>
        I’ve run this query a few times to make sure both should be
        reading data from cache.<br>
        I expect my postgres on GPC to be at least similar to the one
        managed by AWS RDS so that I can work on improvements parallelly
        and compare.<br>
        <br>
        <b>DETAILS:<br>
          Query explain for Postgres on GCP VM:<br>
        </b>Bitmap Heap Scan on SignalRecordsBlobs SignalRecordsBlobs
         (cost=18.80..2480.65 rows=799 width=70) (actual
        time=216.766..776.032 rows=5122 loops=1)<br>
            Filter: (("DateTime" &gt;= \'2019-11-28
        14:00:12.5402\'::timestamp without time zone) AND ("DateTime"
        &lt;= \'2020-07-23 21:12:32.249\'::timestamp without time zone))<br>
            Heap Blocks: exact=5223<br>
            Buffers: shared hit=423 read=4821<br>
          -&gt;  Bitmap Index Scan on
        IDX_SignalRecordsBlobs_SignalSettingId  (cost=0.00..18.61
        rows=824 width=0) (actual time=109.000..109.001 rows=5228
        loops=1)<br>
                  Index Cond: ("SignalSettingId" = 103)<br>
                  Buffers: shared hit=3 read=18<br>
        Planning time: 456.315 ms<br>
        Execution time: 776.976 ms<br>
        <br>
        <b>Query explain for Postgres on AWS RDS:<br>
        </b>Bitmap Heap Scan on SignalRecordsBlobs SignalRecordsBlobs
         (cost=190.02..13204.28 rows=6213 width=69) (actual
        time=2.215..14.505 rows=5122 loops=1)<br>
            Filter: (("DateTime" &gt;= \'2019-11-28
        14:00:12.5402\'::timestamp without time zone) AND ("DateTime"
        &lt;= \'2020-07-23 21:12:32.249\'::timestamp without time zone))<br>
            Heap Blocks: exact=5209<br>
            Buffers: shared hit=3290 read=1948<br>
          -&gt;  Bitmap Index Scan on
        IDX_SignalRecordsBlobs_SignalSettingId  (cost=0.00..188.46
        rows=6405 width=0) (actual time=1.159..1.159 rows=5228 loops=1)<br>
                  Index Cond: ("SignalSettingId" = 103)<br>
                  Buffers: shared hit=3 read=26<br>
        Planning time: 0.407 ms<br>
        Execution time: 14.87 ms<br>
        <br>
        <b>PostgreSQL version number running:<br>
          • VM on GCP</b>: PostgreSQL 11.10 (Debian 11.10-0+deb10u1) on
        x86_64-pc-linux-gnu, compiled by gcc (Debian 8.3.0-6) 8.3.0,
        64-bit<br>
        <b>• Managed by RDS on AWS:</b> PostgreSQL 11.10 on
        x86_64-pc-linux-gnu, compiled by gcc (GCC) 4.8.5 20150623 (Red
        Hat 4.8.5-11), 64-bit<br>
        <br>
        <b>How PostgreSQL was installed:<br>
          • VM on GCP</b>: Already installed when created VM running
        Debian on Google Console. <br>
        <b>• Managed by RDS on AWS:</b> RDS managed the installation.<br>
        <br>
        <b>Changes made to the settings in the postgresql.conf file:<br>
        </b>Here are some postgres parameters that might be useful:<br>
        <b>Instance on VM on GCP (2 vCPUs, 2 GB memory, 800 GB disk):</b><br>
        • effective_cache_size: 1496MB<br>
        • maintenance_work_mem: 255462kB (close to 249MB)<br>
        • max_wal_size: 1GB<br>
        • min_wal_size: 512MB<br>
        • shared_buffers: 510920kB (close to 499MB)<br>
        • max_locks_per_transaction 1000<br>
        • wal_buffers: 15320kB (close to 15MB)<br>
        • work_mem: 2554kB<br>
        • effective_io_concurrency: 200<br>
        • dynamic_shared_memory_type: posix<br>
        On this instance we installed a postgres extension called
        timescaledb to gain performance on other tables. Some of these
        parameters were set using recommendations from that extension.<br>
        <br>
        <b>Instance managed by RDS (2 vCPUs, 2 GiB RAM, 250GB disk, 750
          de IOPS):</b><br>
        • effective_cache_size: 1887792kB (close to 1844MB)<br>
        • maintenance_work_mem: 64MB<br>
        • max_wal_size: 2GB<br>
        • min_wal_size: 192MB<br>
        • shared_buffers: 943896kB (close to 922MB)<br>
        • max_locks_per_transaction 64<br>
        <br>
        <b>Operating system and version by runing "uname -a":<br>
          • VM on GCP:</b> Linux {{{my instance name}}}
        4.19.0-14-cloud-amd64 #1 SMP Debian 4.19.171-2 (2021-01-30)
        x86_64 GNU/Linux<br>
        <b>• Managed by AWS RDS:</b> Aparently Red Hay as shown using
        SELECT version();<br>
        <br>
        <b>Program used to connect to PostgreSQL:</b> Python
        psycopg2.connect() to create the connection and pandas
        read_sql_query() to query using that connection.
        <div><br>
        </div>
        <div>
          <div>Thanks in advance</div>
        </div>
      </div>
    </blockquote>
  </body>
</html>

--------------4D75A144FFF51D2F20F412A7--