Server Sizing... Need Examples

[#1]

Hello,

My team is relatively new to QlikView, and we have a relatively small user base using a very light QV applicatoin at the moment, but we're just around the corner from our first official launch and we still feel very "in the grey" when we think about if our server is strong enough for the task it is about to undertake.

A brief background first... we're a clothing retailer, and our first application is for our inventory management team and it contains the following dimensions/facts;

1. Date
2. Product
3. Location
4. Sales
5. Inventory
6. Purchase Order
7. Transfers

Our users maintain the need to look at data at the lowest level of granularity, being SKU/LOC/DAY in our case, however they rarely ask for this level to be their default view, therefore we're nearly unversially aggregating up from the lowest levels to much higher levels of granularity.

Now, from what we know about how QlikView uses hardware, the granular data is loaded into RAM, and then CPU is used to peform the on-the-fly calculations and then consequently store the results in cache.

Below are the specifications of our server;

1. Manufacturer: IBM
2. Model: 786310X
3. Processor: Intel Xeon CPU E5-2630 0 @ 2.30GHz (x2)
4. RAM: 128 GB
5. 64-bit Architecture

Aside from the large amount of RAM, this feels like a very normal server to us... we know the hard disk requirements for QVW files, and we have done the math around QVW size and number of users to get an estimate as to how much RAM this will use, and we appear to have PLENTY!  Yet... our application is SLOW!  Every selection causes the screen to dim, and you actually see the in process circle clocking before the screen refreshes.

We have yet to get a very black and white answer as to how big is big enough, and it is appearing to us that this is a subjective question, so I am not really looking for anyone to provide their best guess as to what we need for our specific installation, but what I would like to do is gather data to see where our proposed server build ranks amongst the entire QlikView server population.

If you wouldn't mind sharing the above specifications for your QlikView standalone installation, I would greatly appreciate it!

Thanks,

Chris

[#2]

Hi Chris,

I would look at the CPU, maybe to see if it's possible to get a few more core with NUMA disabled.

Then, next is your architecture.

Or, on the front screen itself. Make sure on load, you have limited data on display, otherwise, it gives a feeling to user that on the load, they are waiting for the data, hence, the "slow" effect.

You might also want to preload the app, so that when the first user open the dashboard for the day, the user doesnt wait for the time.