Actually testing memory isn't that hard.
I have always told clients to go to the vendor's labs to perform testing so
you don't have to acquire the extra memory yourself.
However, in most cases for companies new equipment is arriving every couple
of months and you normally can scavenge memory to make a full server to run
some tests. You aso can look at your spare parts pool to run a couple of
days of testing.
Lots of ways to skin this cat.
I normally use some open source benchmarking SW that represents standard
apps and create some stub apps in the languages the client is using. It is
fairly straight forward and you can complete testing on 2 days.
Matthew Parker
Chief Technologist
Dimensional DBA
425-891-7934 (cell)
D&B 047931344
CAGE 7J5S7
Dimensional.dba@xxxxxxxxxxx
<http://www.linkedin.com/pub/matthew-parker/6/51b/944/> View Matthew
Parker's profile on LinkedIn
www.dimensionaldba.com <http://www.dimensionaldba.com/>
From: Tim Gorman [mailto:tim.evdbt@xxxxxxxxx] ;
Sent: Friday, March 31, 2017 1:09 PM
To: Dimensional DBA; Amir.Hameed@xxxxxxxxx; 'ORACLE-L'
Subject: Re: Memory speed question
"Your mileage will vary" and "it depends" are always 100% accurate, but that
does not make them useful.
Its generally correct to advise testing, but benchmarking memory access is a
big task, and anyone holding a title like "architect" should expect to
substantiate.
On 3/31/17 13:52, Dimensional DBA wrote:
The information about going from 1600MHZ to 1066MHZ is in the Intel and HW
documentation relative to Intel architecture.
I have benchmarked systems with a variety of different applications java,
python, perl, databases etc.
The "your milage will vary" is of course very accurate. The performance
differential to the app in some cases was very low and in other cases was up
to 40% differential.
Depends on your app and what the app is doing.
Matthew Parker
Chief Technologist
Dimensional DBA
425-891-7934 (cell)
D&B 047931344
CAGE 7J5S7
Dimensional.dba@xxxxxxxxxxx
<http://www.linkedin.com/pub/matthew-parker/6/51b/944/> View Matthew
Parker's profile on LinkedIn
www.dimensionaldba.com <http://www.dimensionaldba.com/>
From: Tim Gorman [mailto:tim.evdbt@xxxxxxxxx] ;
Sent: Friday, March 31, 2017 12:33 PM
To: Amir.Hameed@xxxxxxxxx; Dimensional DBA; 'ORACLE-L'
Subject: Re: Memory speed question
Amir,
I'd recommend pushing hard for concrete information: a KB or support
documentation from Cisco. Lacking substantiation, it's just an opinion.
Don't allow hearsay, opinions, and myths to propagate without challenge. Be
respectful, be polite, but be firm.
Hope this helps...
-Tim
On 3/31/17 13:08, Hameed, Amir wrote:
For that we will need to buy memory and then add it to the host which is
what we are trying to avoid. We have also asked our data center to check
with Cisco and the response we got was that "mileage will vary".
From: Dimensional DBA [mailto:dimensional.dba@xxxxxxxxxxx] ;
Sent: Friday, March 31, 2017 3:01 PM
To: Hameed, Amir <mailto:Amir.Hameed@xxxxxxxxx> <Amir.Hameed@xxxxxxxxx>;
'ORACLE-L' <mailto:oracle-l@xxxxxxxxxxxxx> <oracle-l@xxxxxxxxxxxxx>
Subject: RE: Memory speed question
It is simplest to run a benchmark and you will see that your app will notice
a difference if you have a low latency app.
It also limits the overall amount of work that can be accomplished on that
blade.
Matthew Parker
Chief Technologist
Dimensional DBA
425-891-7934 (cell)
D&B 047931344
CAGE 7J5S7
Dimensional.dba@xxxxxxxxxxx
<http://www.linkedin.com/pub/matthew-parker/6/51b/944/> View Matthew
Parker's profile on LinkedIn
www.dimensionaldba.com <http://www.dimensionaldba.com/>
From: oracle-l-bounce@xxxxxxxxxxxxx [mailto:oracle-l-bounce@xxxxxxxxxxxxx]
On Behalf Of Hameed, Amir
Sent: Friday, March 31, 2017 11:55 AM
To: 'ORACLE-L'
Subject: Memory speed question
We have a four-node ESX farm/cluster where each node is loaded with 512G of
physical memory. These ESX nodes are Cisco UCS blades. We are looking into
adding more memory to each of these ESX hosts to increase the footprint to
750G/node. One of the data center architects had informed us last year that
per Cisco, the sweet-spot of memory footprint on these blades is 512GB,
after that the memory access speed would drop from 1600MHZ to 1066MHZ (a
drop of ~ 30%). Because all of our VMs host application tiers and they run a
lot of JVM processes, we are concerned that a drop of 30% will negatively
impact our applications' performance. However, our data center has been
telling us that the drop will not be noticed by the application which I find
hard to believe. We also do not want to add another ESX host because of
licensing cost.
I was wondering if anyone is running their ESX hosts with memory over
512G/host and if they observed any issue with memory speed.
Thank you,
Amir
