Re: FSM, now without WAL-logging - Mailing list pgsql-hackers

Heikki Linnakangas napsal(a):
> Zdenek Kotala wrote:

> 
> Attached is a new version, now with WAL-logging of the FSM truncation. I 
> decided to go with the separate WAL record for that, rather than 
> piggybacking on the smgrtruncate's WAL record. It seems much better from 
> a modularity point of view this way. I've also worked on the comments, 
> renamed many of the internal functions, in a more coherent scheme, and I 
> also started using the "struct FSMAddress" you suggested a while ago.
> 
> But I don't think I've changed anything that could explain that crash. 
> Let me know if it still doesn't work.

This version works on my "old" repo.

I performed performance test (iGEN) on SUN x4600 with 60 concurrent 
users and see result:

Original:
---------
MQThL (Maximum Qualified Throughput LIGHT): 1209.60 tpm
MQThM (Maximum Qualified Throughput MEDIUM): 2576.72 tpm
MQThH (Maximum Qualified Throughput HEAVY): 2191.20 tpm


TRANSACTION MIX

Total number of transactions = 181232
TYPE            TX. COUNT       MIX
----            ---------       ---
Light:          30240           16.69%
Medium:         64418           35.54%
DSS:            19865           10.96%
Heavy:          54780           30.23%
Connection:     11929           6.58%


RESPONSE TIMES          AVG.            MAX.            90TH

Light                   0.304           6.405           0.400
Medium                  0.317           6.533           0.400
DSS                     0.266           6.343           0.020
Heavy                   0.361           6.737           3.000
Connections             0.264           5.983           0.400
Number of users = 60
Sum of Avg. RT * TPS for all Tx. Types = 32.770142


FSM with WAL
------------
MQThL (Maximum Qualified Throughput LIGHT): 1199.36 tpm
MQThM (Maximum Qualified Throughput MEDIUM): 2569.12 tpm
MQThH (Maximum Qualified Throughput HEAVY): 2171.64 tpm


TRANSACTION MIX

Total number of transactions = 180625
TYPE            TX. COUNT       MIX
----            ---------       ---
Light:          29984           16.60%
Medium:         64228           35.56%
DSS:            20181           11.17%
Heavy:          54291           30.06%
Connection:     11941           6.61%


RESPONSE TIMES          AVG.            MAX.            90TH

Light                   0.309           6.560           0.400
Medium                  0.323           6.529           0.400
DSS                     0.268           6.327           0.020
Heavy                   0.360           6.675           3.000
Connections             0.274           6.359           0.400
Number of users = 60
Sum of Avg. RT * TPS for all Tx. Types = 32.845712


FSM no WAL last version
-----------------------
MQThL (Maximum Qualified Throughput LIGHT): 1207.92 tpm
MQThM (Maximum Qualified Throughput MEDIUM): 2611.84 tpm
MQThH (Maximum Qualified Throughput HEAVY): 2177.68 tpm


TRANSACTION MIX

Total number of transactions = 182222
TYPE            TX. COUNT       MIX
----            ---------       ---
Light:          30198           16.57%
Medium:         65296           35.83%
DSS:            20118           11.04%
Heavy:          54442           29.88%
Connection:     12168           6.68%


RESPONSE TIMES          AVG.            MAX.            90TH

Light                   0.301           6.106           0.400
Medium                  0.315           6.130           0.400
DSS                     0.261           5.977           0.020
Heavy                   0.361           6.220           3.000
Connections             0.260           6.044           0.400
Number of users = 60
Sum of Avg. RT * TPS for all Tx. Types = 32.696832


-----------------------------------------

I don't see any big difference. Throughput is similar. Only response 
time seems to be better with your last FSM version.

I personally happy with performance.
        Zdenek