23.07.2014 Views

Lustre 1.6 Operations Manual

Lustre 1.6 Operations Manual

Lustre 1.6 Operations Manual

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

21.5.5 Further Tuning Tips<br />

Here are some tips we have drawn from testing at a large installation:<br />

■<br />

■<br />

■<br />

Use the full device instead of a partition (sda vs sda1). When using the full device,<br />

<strong>Lustre</strong> writes nicely-aligned 1 MB chunks to disk. Partitioning the disk can<br />

destroy this alignment and will noticeably impact performance.<br />

Separate the EXT3 OST into two LUNs, a small LUN for the EXT3 journal and a<br />

big one for the "data".<br />

Since <strong>Lustre</strong> 1.0.4, we supply EXT3 mkfs options when we create the OST like<br />

-j, -J and so on in the following manner (where /dev/sdj has been formatted<br />

before as a journal). The journal size should not be larger than 1 GB (262144 4 KB<br />

blocks) as it can consume up to this amount of RAM on the OSS node per OST.<br />

# mke2fs -O journal_dev -b 4096 /dev/sdj [optional size]<br />

Tip – A very important tip—on the S2A DDN 8500 storage array, you need to create<br />

one OST per TIER, especially in write through (see output below). This is of concern<br />

if you have 16 tiers. Create 16 OSTs consisting of one tier each, instead of eight made<br />

of two tiers each.<br />

- Performance is significantly better on the S2A DDN 9500 and 9550 storage arrays<br />

with two tiers per LUN.<br />

- Do NOT partition the DDN LUNs, as this causes all I/O to the LUNs to be<br />

misaligned by 512 bytes. The DDN RAID stripes and cachelines are aligned on 1 MB<br />

boundaries. Having the partition table on the LUN causes all 1 MB writes to do a<br />

read-modify-write on an extra chunk, and ALL 1 MB reads to, instead, read 2 MB<br />

from disk into the cache, causing a noticeable performance loss.<br />

- You are not obliged to lock in cache the small LUNs.<br />

- Configure the MDT on a separate volume that is configured as RAID 1+0. This<br />

reduces the MDT I/O and doubles the seek speed.<br />

Chapter 21 <strong>Lustre</strong> Tuning 21-9

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!