DTS Package taking hours to complete

Question

Post reply

DTS Package taking hours to complete

Tinus

SSC Rookie

Points: 39
More actions
September 11, 2006 at 11:45 pm

#76577

I have a dts package that runs in 4 steps.
The 1st step is to copy data from multiple columns into a single table. This gets done in 15 minutes. The next step is to update all the missing data with an update script. This is where the dts package slows down to an eternity.
It use to run anywhere from 30-45 minutes, now if I am lucky it finishes after 65 hours. We moved from an Intel server to an IBM 345 server. I would like to know where I can start to look for the reason why it takes so long.
Below is a copy of the update script:
-- Update Blood Group Information of each Unit in zcusIss
update zcusIss
set
  Blood_Group = Case
    when left(sbhx.blood_type,3) = 'APO' then 'Group A'
    when left(sbhx.blood_type,3) = 'ANE' then 'Group A'
    when left(sbhx.blood_type,3) = 'A1N' then 'Group A'
    when left(sbhx.blood_type,3) = 'A1P' then 'Group A'
    when left(sbhx.blood_type,3) = 'A2N' then 'Group A'
    when left(sbhx.blood_type,3) = 'A2P' then 'Group A'
    when left(sbhx.blood_type,3) = 'APO' then 'Group A'
    when left(sbhx.blood_type,3) = 'AWP' then 'Group A'
    when left(sbhx.blood_type,3) = 'AWN' then 'Group A'
    when left(sbhx.blood_type,3) = 'BNE' then 'Group B'
    when left(sbhx.blood_type,3) = 'BPO' then 'Group B'
    when left(sbhx.blood_type,3) = 'OPO' then 'Group O'
    when left(sbhx.blood_type,3) = 'ONE' then 'Group O'
    when left(sbhx.blood_type,3) = 'A1B' then 'Group AB'
    when left(sbhx.blood_type,3) = 'A2B' then 'Group AB'
    when left(sbhx.blood_type,3) = 'ABN' then 'Group AB'
    when left(sbhx.blood_type,3) = 'ABP' then 'Group AB'
    when left(sbhx.blood_type,3) = 'AWB' then 'Group AB'
    when left(sbhx.blood_type,3) = 'INC' then 'Other'
    when left(sbhx.blood_type,3) = 'NAT' then 'Other'
    when left(sbhx.blood_type,3) = 'UNK' then 'Other'
    else 'Unk Group'
   end,
   Blood_Type = coalesce(
      case
        when sbhx.blood_type <> '' then sbhx.blood_type
        else 'No Group'
   end, 'No Group')
from
  zcusIss     (nolock)
  left join sabts_lab_b_hx_main sbhx (nolock)
   on hx_urn = sbhx.urn
where
  Blood_Group is null
The joins used in both tables are indexed.
Any help in this regard would be greatly appreaciated.

Viewing 14 posts - 1 through 13 (of 13 total)

You must be logged in to reply to this topic. Login to reply

Sreejith Sreedharan SSCertifiable Points: 5403 More actions · Answer 1

Have u looked at query execution plan? That should point you in the right direction.Do you have Index on bloodType?

Randy Doub SSCertifiable Points: 5322 More actions · Answer 2

Put an index on bloodtype. Even if its only temporary to get your update done. You might then try updating everything to GroupA, then test and update the other groups. You'd have to play around with your updates. Maybe after indexed, seperate updates for each group would be quicker than a Case statement.

Tinus SSC Rookie Points: 39 More actions · Answer 3

I have an index for the bloodgroup setup. As for the query execution plan I have run it but it does not make any sense to me. It gives a breakdown with cost allocated to each.

Michael Tocik SSCertifiable Points: 7051 More actions · Answer 4

Hi

Do you have any foreign key relationships on the table. If so try remove the relationship for the update

Mike

Tinus SSC Rookie Points: 39 More actions · Answer 5

Hi there,

Below is the table structure and indexes used on table:

Name	Value
QUOTED_IDENTIFIER	ON
ANSI_NULLS	ON
ANSI_PADDING	OFF

Columns

Name	Description	Data Type	Max Length	Nullable	Default	IsGUID
unit_number		varchar	20
Unit_Urn		varchar	20
Donor_Urn		varchar	20
hx_urn		varchar	20
root_unit		varchar	20
patient		varchar	50
Collection_Date		datetime	8
Product_Group		varchar	27
Product_Type		varchar	22
Product_Category		varchar	25
Product		varchar	25
Status		varchar	13
Issued_to_Spec		varchar	20
Iss_Time_Frame		varchar	7
Issue_Hour		varchar	7
Issue_Time		varchar	14
Blood_Group		varchar	15
Blood_Type		varchar	15
IPGroup		varchar	11
IP		varchar	15
Issue_User		varchar	20
Issue_Type		varchar	9
Specimen_Priority		varchar	25
Collection_Zone		varchar	21
Collection_Zone_Branch		varchar	17
Collection_Zone_Site		varchar	20
Collection_Province		varchar	14
Collection_Branch		varchar	17
Collection_Clinic		varchar	20
Issue_Date		datetime	8
Issue_Province		varchar	16
Issue_Branch		varchar	17
Issue_Site		varchar	20
Issue_Zone		varchar	21
Issue_Zone_Branch		varchar	19
Issue_Zone_Site		varchar	20
Processing_Province		varchar	17
Processing_Branch		varchar	21
Processing_Site		varchar	20
Hospital_Province		varchar	25
Hospital_Branch		varchar	25
Hospital		varchar	15
Ward		varchar	15
Classification		varchar	10
Insurance_Group		varchar	10
Medical_Aid		varchar	15
ICD10		varchar	15
Unit_Count		int	4

Total: 48 column(s)

Indexes

Index	Primary	Unique
blood_group_idx
collection_date_idx
donor_urn_idx
hx_urn_idx
issue_date_idx
issued_to_spec_idx
patient_idx
root_unit_idx
unit_urn_idx

krs SSC Journeyman Points: 91 More actions · Answer 6

You could try updating from a table var, rather than the lengthy CASE statement:

-- CREATE TABLE VAR

DECLARE @Blood_Groups TABLE (Blood_Type varchar(3), Blood_Group varchar(8) )

-- Insert updatable values

INSERT INTO @Blood_Groups (Blood_Type, Blood_Group) VALUES ('APO', 'Group A')

INSERT INTO @Blood_Groups (Blood_Type, Blood_Group) VALUES ('ANE', 'Group A')

INSERT INTO @Blood_Groups (Blood_Type, Blood_Group) VALUES ('A1N', 'Group A')

INSERT INTO @Blood_Groups (Blood_Type, Blood_Group) VALUES ('A1P', 'Group A')

INSERT INTO @Blood_Groups (Blood_Type, Blood_Group) VALUES ('A2N', 'Group A')

INSERT INTO @Blood_Groups (Blood_Type, Blood_Group) VALUES ('A2P', 'Group A')

INSERT INTO @Blood_Groups (Blood_Type, Blood_Group) VALUES ('APO', 'Group A')

INSERT INTO @Blood_Groups (Blood_Type, Blood_Group) VALUES ('AWP', 'Group A')

INSERT INTO @Blood_Groups (Blood_Type, Blood_Group) VALUES ('AWN', 'Group A')

INSERT INTO @Blood_Groups (Blood_Type, Blood_Group) VALUES ('BNE', 'Group B')

INSERT INTO @Blood_Groups (Blood_Type, Blood_Group) VALUES ('BPO', 'Group B')

INSERT INTO @Blood_Groups (Blood_Type, Blood_Group) VALUES ('OPO', 'Group O')

INSERT INTO @Blood_Groups (Blood_Type, Blood_Group) VALUES ('ONE', 'Group O')

INSERT INTO @Blood_Groups (Blood_Type, Blood_Group) VALUES ('A1B', 'Group AB')

INSERT INTO @Blood_Groups (Blood_Type, Blood_Group) VALUES ('A2B', 'Group AB')

INSERT INTO @Blood_Groups (Blood_Type, Blood_Group) VALUES ('ABN', 'Group AB')

INSERT INTO @Blood_Groups (Blood_Type, Blood_Group) VALUES ('ABP', 'Group AB')

INSERT INTO @Blood_Groups (Blood_Type, Blood_Group) VALUES ('AWB', 'Group AB')

INSERT INTO @Blood_Groups (Blood_Type, Blood_Group) VALUES ('INC', 'Other')

INSERT INTO @Blood_Groups (Blood_Type, Blood_Group) VALUES ('NAT', 'Other')

INSERT INTO @Blood_Groups (Blood_Type, Blood_Group) VALUES ('UNK', 'Other')

-- Update Blood Group Information of each Unit in zcusIss

-- Left join to table var to retrieve blood group names

update zcusIss

set

Blood_Group = IsNull(BG.Blood_Group,'Unk Group'),

Blood_Type = coalesce(

case

when sbhx.blood_type <> '' then sbhx.blood_type

else 'No Group'

end, 'No Group')

from

zcusIss (nolock)

left join sabts_lab_b_hx_main sbhx (nolock)

on hx_urn = sbhx.urn

LEFT JOIN @Blood_Groups BG

ON sbhx.Blood_Type = BG.Blood_Type

where

zcusIss.Blood_Group is null

Kent Kester SSC Enthusiast Points: 100 More actions · Answer 7

Ah, darn it. Somehow lost my post.

krs has a good idea. Although I wonder about using the left([fieldname], 3) repeatedly in a case statement. That's assuming that blood_type varies beyond those 3 characters.

You don't show an index on blood_type. What if you created one and, rather than using the case statement, rewrote this as 5 different update statements? Doing that you could add lines to your where clause to find the blood_type. That would actually use an index rather than brute-forcing your way through the table.

"and sbhx.blood_type like 'APO%' and sbhx.blood_type like 'ANE%'" etc.

Another thought is to run this as an INSERT INTO a temp table and then UPDATE back into zcusIss based on the primary key. This would prevent you from SELECTing and UPDATEing one table at the same time.

My biggest concern, though, is that a change in hardware caused this problem. If you still have the old hardware around you might want to compare your old configuartion to hte new one to see what's different that might have caused the problem. Any time a change in hardware causes something to take 130x longer you can bet there's something else going on as well.

cliffb SSCarpal Tunnel Points: 4598 More actions · Answer 8

What about using your LEFT in the case only once. You are evaluating the same value to be some other value so there is no need to put the full evaluation in the WHEN.

Blood_Group = Case left(sbhx.blood_type,3)

when 'APO' then 'Group A' <-- first

when 'ANE' then 'Group A'

when 'A1N' then 'Group A'

when 'A1P' then 'Group A'

when 'A2N' then 'Group A'

when 'A2P' then 'Group A'

when 'APO' then 'Group A' <-- second.

when 'AWP' then 'Group A'

when 'AWN' then 'Group A'

when 'BNE' then 'Group B'

when 'BPO' then 'Group B'

when 'OPO' then 'Group O'

when 'ONE' then 'Group O'

when 'A1B' then 'Group AB'

when 'A2B' then 'Group AB'

when 'ABN' then 'Group AB'

when 'ABP' then 'Group AB'

when 'AWB' then 'Group AB'

when 'INC' then 'Other'

when 'NAT' then 'Other'

when 'UNK' then 'Other'

else 'Unk Group'

end

This is of course, if you do not use KRS's solution which is better than the above suggestion.

Dennis Link SSC-Addicted Points: 420 More actions · Answer 9

Dennis Link

SSC-Addicted

Points: 420

September 13, 2006 at 9:39 am

#660133

Can you post the query plan?

Tinus SSC Rookie Points: 39 More actions · Answer 10

How would I go about posting the query plan?

The old hardware was a 4 dual xeon processor with 8gb RAM. The new hardware is a IBM 345 2 dual processors and 4gb RAM. I ran the same query on the old server and it completes in 30 - 1.5 Hours. The reason for moving it to the IBM was that we suffered a db crash and had to recover the db from the mdf files. With the data recovered we are facing a new problem where the dts jobs takes forever to complete.

The table I am creating is in turn used as the fact table for an Analsys Server Cube. The dts package consists out of 5 steps. The first is to populate the majority of data from table using pk's. This takes 15 minutes. Then the other jobs uses custom indexes created to speed up the update, these are blood group, patient, collection date and everything below hospital province. Once the skeleton is created I use the indexes created in this table (zcusIss) to join to other master data tables.

The full blood group will look something like OPOSL, APOSL, AwBPOS, A2BPOS etc. With the left I am trying to group each blood group together for the cube to use as a dimension.

Please let me know how to go about posting the query plan.

Thanks for all the suggestions so far.

Tinus SSC Rookie Points: 39 More actions · Answer 11