|
Home > Archive > SQL Anywhere database replication > April 2005 > SQL Remote vs. MobiLink performance
You are viewing an archived Text-only version of the thread.
To view this thread in it's original format and/or if you want to reply to
this thread please [click here]
| Author |
SQL Remote vs. MobiLink performance
|
|
| Carl Kayser 2005-04-12, 11:24 am |
| We currently use ASE consolidated - ASA consolidated - ASA Remotes
SQL-Remote replication. We are fortunate in that we do not have multiple
copies of our data partitioned across our users. We also use "natural
keys", which are multi-column and include the remote id, and do not use PK
pools, autoincrement, etc.
We made a fairly simple database (10 tables) and benchmarked the two
products. For the remotes DBRemote was definitely faster than MobiLink with
inserts and deletes. (We did not benchmark deletes since we anticipate
multiple copies in the future, hence logical deletes fro MobiLink).
So, does anyone have real-life comparisons of the two products with respect
to the durations of the remote sessions? (It's rather difficult making a
sell to our customer base if we go to a product that is perceived as being
slower. Reducing latency is not as important to them.)
| |
| Reg Domaratzki \(iAnywhere Solutions\) 2005-04-13, 9:23 am |
| We haven't done any performance testing that puts MobiLink against SQL
Remote for ASE, but the scalability of MobiLink is heads and shoulders above
that of SQL Remote for ASE. Make sure you've read over
http://www.sybase.com/detail?id=1009664.
> "For the remotes DBRemote was definitely faster than MobiLink"
So you're timing the time it takes dbremote to apply data on the remote
versus dbmlsync correct? The only reason you'd see an increase in
performance on the remote would come from the fact that MobiLink/dbmlsync
does not replicate operations, but synchronizes changes. If you modify the
same row ten times on the consolidated, with SQL Remote, ten updates are
sent, but with MobiLink, only a single update is sent. The same is true
when sending data from the remote to the consolidated as well. Depending on
how your data is manipulated by your application, this could be a huge
savings.
I believe the difference you'll see in performance is all at the
consolidated database. Inserting data from the remote into the consolidated
from the remotes will likely be roughly equivalent, but the time to generate
the download stream will be significantly faster. While SQL Remote for ASE
needs to scan the ASE transaction log, populate the stable queue and then
scan the stable queue to generate messages for everybody, MobiLink is
essentially executing a select statement on the consolidated database. This
will not only be faster, but also removes the need to keep a redundant copy
of the data in the ASE server in the stable queue.
Also remember that dbmlsync and dbremote on the remotes are doing vastly
different things. dbmlsync connects to MobiLink and gets a snapshot of data
from the consolidated when the synchronization occurs. If a log offset
problem is detected, dbmlsync and MobiLink attempt to resolve it while
connected, instead of giving up and sending a message back to consolidated
asking for a resend. I also can't stress how much more scalable MobiLink is
over SQL Remote for ASE.
PS : Have you read my whitepaper (well, white book really) on migrating from
SQL Remote for ASE to MobiLink?
http://www.ianywhere.com/whitepapers/migrate_to_ml.html
--
Reg Domaratzki, Sybase iAnywhere Solutions
Sybase Certified Professional - Sybase ASA Developer Version 8
Please reply only to the newsgroup
iAnywhere Developer Community : http://www.ianywhere.com/developer
iAnywhere Documentation : http://www.ianywhere.com/developer/product_manuals
ASA Patches and EBFs : http://downloads.sybase.com/swx/sdmain.stm
-> Choose SQL Anywhere Studio
-> Set "Platform Preview" and "Time Frame" to ALL
"Carl Kayser" <kayser_c@bls.gov> wrote in message
news:425bf334$1@foru
ms-1-dub...
> We currently use ASE consolidated - ASA consolidated - ASA Remotes
> SQL-Remote replication. We are fortunate in that we do not have multiple
> copies of our data partitioned across our users. We also use "natural
> keys", which are multi-column and include the remote id, and do not use PK
> pools, autoincrement, etc.
>
> We made a fairly simple database (10 tables) and benchmarked the two
> products. For the remotes DBRemote was definitely faster than MobiLink
with
> inserts and deletes. (We did not benchmark deletes since we anticipate
> multiple copies in the future, hence logical deletes fro MobiLink).
>
> So, does anyone have real-life comparisons of the two products with
respect
> to the durations of the remote sessions? (It's rather difficult making a
> sell to our customer base if we go to a product that is perceived as being
> slower. Reducing latency is not as important to them.)
>
>
| |
| Carl Kayser 2005-04-13, 11:23 am |
|
"Reg Domaratzki (iAnywhere Solutions)" < Spam_bad_rdomarat@ia
nywhere.com>
wrote in message news:425d286b$1@foru
ms-2-dub...
> We haven't done any performance testing that puts MobiLink against SQL
> Remote for ASE, but the scalability of MobiLink is heads and shoulders
above
> that of SQL Remote for ASE. Make sure you've read over
> http://www.sybase.com/detail?id=1009664.
I had skimmed over this previously. We have 400-450 remotes and the
scale is less than for the benchmarks described (1,000 remotes).
It also appeared as if there were many factors that could affect
performance: slow/fast clients, number of worker threads, number of
concurrent sessions, etc. Fine tuning can be quite exhaustive.
>
>
> So you're timing the time it takes dbremote to apply data on the remote
> versus dbmlsync correct?
Yes. (We also noted that MobiLink extractions were faster and consolidated
updates were much slower.)
The only reason you'd see an increase in
> performance on the remote would come from the fact that MobiLink/dbmlsync
> does not replicate operations, but synchronizes changes. If you modify
the
> same row ten times on the consolidated, with SQL Remote, ten updates are
> sent, but with MobiLink, only a single update is sent. The same is true
> when sending data from the remote to the consolidated as well. Depending
on
> how your data is manipulated by your application, this could be a huge
> savings.
>
> I believe the difference you'll see in performance is all at the
> consolidated database. Inserting data from the remote into the
consolidated
> from the remotes will likely be roughly equivalent, but the time to
generate
> the download stream will be significantly faster. While SQL Remote for
ASE
> needs to scan the ASE transaction log, populate the stable queue and then
> scan the stable queue to generate messages for everybody, MobiLink is
> essentially executing a select statement on the consolidated database.
This
> will not only be faster, but also removes the need to keep a redundant
copy
> of the data in the ASE server in the stable queue.
>
> Also remember that dbmlsync and dbremote on the remotes are doing vastly
> different things. dbmlsync connects to MobiLink and gets a snapshot of
data
> from the consolidated when the synchronization occurs. If a log offset
> problem is detected, dbmlsync and MobiLink attempt to resolve it while
> connected, instead of giving up and sending a message back to consolidated
> asking for a resend. I also can't stress how much more scalable MobiLink
is
> over SQL Remote for ASE.
OK, but I'm not sure that we are in the "scalability ballpark" where
benefits will be reaped.
We do see that our current DBRemotes for our batch consolidated downloads
can be pretty long
(especially when we have to add in the log dumps for each or our SSRemote &
DBRemote cycles). So better performance and recoverability are very nice
for us admins. However, the comparatively longer sessions on the remotes
are
what our customers will see.
>
> PS : Have you read my whitepaper (well, white book really) on migrating
from
> SQL Remote for ASE to MobiLink?
> http://www.ianywhere.com/whitepapers/migrate_to_ml.html
No. We had heard that it covered the basics but that there would likely be
more work than is described.
>
> --
> Reg Domaratzki, Sybase iAnywhere Solutions
> Sybase Certified Professional - Sybase ASA Developer Version 8
> Please reply only to the newsgroup
>
> iAnywhere Developer Community : http://www.ianywhere.com/developer
> iAnywhere Documentation :
http://www.ianywhere.com/developer/product_manuals
> ASA Patches and EBFs : http://downloads.sybase.com/swx/sdmain.stm
> -> Choose SQL Anywhere Studio
> -> Set "Platform Preview" and "Time Frame" to ALL
>
> "Carl Kayser" <kayser_c@bls.gov> wrote in message
> news:425bf334$1@foru
ms-1-dub...
multiple[color=darkr
ed]
PK[color=darkred]
> with
> respect
a[color=darkred]
being[color=darkred]
>
>
|
|
|
|
|