Sign in to follow this  
Brian

In The Lab Series Discussion

Recommended Posts

Share this post


Link to post
Share on other sites

Were there any other additional details on how airflow was being managed? Just curious, didn't see it the other articles (I may have missed it), I don't see any obvious air vents or false floors...

Share this post


Link to post
Share on other sites

Back when we weren't running as much gear at one time, we were using the TrippLite SRCOOL12K for primary cooling, where now it is more of a backup of things get too high and the air movement from the fan provides most of the cooling.

http://www.storagereview.com/tripp_lite_smartrack_srcool12k_review

We generally didn't talk much about cooling besides vendors coming into the lab. The lab in general isn't in an optimal spot when it comes to cooling, although when it first started to be the area that all the gear was located it was fine... think of it like the cursed telco closet that finally got too big. Its located in a basement, so the floor is poured concrete. Right now the main airflow during the day is provided by the doors being open and an inline fan positioned behind the racks (not shown in the photos). At night when the doors are closed we have some big passive vents drawing airflow from the hallway into the main room.

Share this post


Link to post
Share on other sites

This came in via email from Andrew:

Just curious - when you run your fio tests, what is the --size typically set at?

Share this post


Link to post
Share on other sites

When testing NAS we use 25G per share (4-8 shares per device). For SAN testing we provision 4-8 25G LUNs and don't use the --size attribute which defaults to using the entire space. For raw storage devices (SSD, PCIe SSDs) we don't set that attribute and again use the full storage device.

Share this post


Link to post
Share on other sites

Thanks for the speedy response. Just curious, if you set the size (in your case 25G) on NAS/SAN you must take the time limit off? How did you end up deciding on 25G and not a range of sizes?

I have to admit that I haven't just tested a single hard drive so I have always done my tests based on size, when I have tried running fio on just a plain filesystem it still asks for a size, is there a flag I am possibly missing?

Again, many thanks and I am enjoying the website!

Share this post


Link to post
Share on other sites

When we run our tests, we use the time attribute to force it to run a set length. If you don't it just runs until it fills up that small slice, instead of continuously going into a steady-state condition.

On the 25G aspect, its a number picked out of a hat mixed with being larger than the cache size of most devices. We wanted a size small enough to see some caching improvement from hot-spot caching algorithms but not too small to completely live inside DRAM cache from a controller.

On our FIO page we have some sample scripts (for a linux environment) here:

http://www.storagereview.com/fio_flexible_i_o_tester_synthetic_benchmark

Share this post


Link to post
Share on other sites

New in the lab about installing the Mellanox ConnectX-4 rNDC Upgrade into our Dell PowerEdge Servers.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

Sign in to follow this