New performance counter in SSY-V PSP3

One could think PSP3 is only a bug fix release and only adding some minor and not really important new enhancements to SANsymphony-V 10 but that's not correct. For all of you who permanetly get asked if your environment can serve more systems and you don't know where to look at to answer this question, DataCore added a new counter that can help you.

Right at the beginning, if you think this new counter will always tell you immediately if your storage is overloaded or not or asking for a feature that will tell you how much VMs or systems exactly you can add to your storage before it begins suffering performance you will get disappointed.  Such an information is still hard work to get especially in a "free-to-scale-and-configure" solution like SSY-V where there are nearly unlimited combinations of hardware components.

BUT, the new counter can give you hints, especially if you think your physical disks are the performance bottleneck.

 

To cut a long story short, the new counter is called "% idle time" and is a subcounter of the physical pool disk object.

IdleTime01

In this example I choosed one of the pool member disks and added the counter "% Idle Time". What does this counter tell you? Well quite easy to guess, it will tell you the percentage of a given time interval when the disk was unaccessed and was ready to serve new requests. Obviously, the higher the better....

Normally DataCore won't give you recommendations for strict vaules. The answer "it depends..." is the default answer here but during one of my last projects where a customer suffered strange performance problems I had the chance to get in touch with Alex Best from DataCore. Alex is one of the "technical superheroes" from DataCore in the DACH region and has plenty of experience in performance related problems. So I told him about the customer problems and he asked me if we already have PSP3 installed. We had so he asked me to add the counter to the live performance charts for the disks building my pool. I added all (or most of them because I can only add 50 counters but have more than 100 disks in both pools of the server group) and looked at the values. Nearly all ranged between 0 and 15%, averaging at ~6-7%. I thought "perfect, 6-7% free time, should be enough" I never guessed here is a big misunderstanding. And that's a big problems with DataCores performance counters. Interpreting them is quite hard and there is no guide or best-practice or whatever may help you. You even won't get definite statements if you ask the support guys. 
Luckily this time Alex made a recommendation: DataCore recommends to have % Idle Time for all disks to be AT LEAST 25%. Values below are a big sign of overload on the disks as they don't have enough spare time to serve new requests. Queues get filled up, IOs are throttled and overall performance can suffer, especially if the disks affected by low idle times are SAS and SSDs which should serve most of your requests.

Hope this helps you in future performance troubleshooting sessions.

Leave your comments

Post comment as a guest

0
Your comments are subjected to administrator's moderation.

People in this conversation

  • Guest - attken

    Thanks for your share , this is very useful information for those admin guys who using DataCore product.

Powered by Komento
joomla templatesfree joomla templatestemplate joomla
2017  v-strange.de   globbers joomla template