Connect. Communicate. Collaborate. Securely.

Home » Kerio User Forums » Kerio Connect » Storage I/O performance - options and measuring tools
  •  
greg_m

Messages: 65
Karma: 0
Send a private message to this user
I've read several threads discussing storage options for Kerio and the optimization of disk I/O, and have learned much from these posts. However, I have a specific setup that I wish to configure in the best way possible to maximize I/O and thus I have a few questions. First, here's my setup.

Around 100 users, mainly Mac Exchange type accounts, with 6 or 7 KOFF users and a dozen or so iPhones. Our mailstore is currently around 560GB. Server will be an Apple Xserve 2.26 4C Xeon with 24GB RAM. This Xserve has an Apple RAID card fitted and three SATA drives, however measurements so far would indicate the disk I/O using the built-in drives is poor. Plus, I've less redundancy with the card. So I plan to create an iSCSI target on a Nexsan SATABoy, which currently holds the mailstore. I have tested with a single ethernet path, but I would prefer multipath setup.

My questions are:

1) What, in your experience, is the best OS X tool for measuring disk I/O, and what are the most important results? I've been using Xbench, but don't understand fully what it is telling me and how that is relevant to Kerio. For instance there are figures for differing block sizes and sequential vs random patterns.

2) What factors, that I can control, will affect I/O in my environment? I can't employ anything other than RAID 1 or 5. I do have options for the block size when creating an array on the SATABoy (16, 32, 64 and 128k). If I can get multipathing working, then this should give me the best iSCSI performance.

Any help or advice much appreciated.
  •  
GlennK

Messages: 252
Karma: 3
Send a private message to this user
Greg,

Try AJA System Test, freely available for download. Do a search. It does a good job of showing disk speed tests.

As far as the important results, its a bit of gray area except that of course you want as quick as possible. I have decided to go with a 32k block in a RAID 1 or 10 system.
  •  
scottwilkins

Messages: 654
Karma: 7
Send a private message to this user
Make sure the SATA disks you use have NCQ ability and as large of a cache as you can get. And don't go RAID 5.
  •  
greg_m

Messages: 65
Karma: 0
Send a private message to this user
Thanks to you both for your replies. I don't have much option with the storage as I mentioned, we have a Nexsan SATABoy and I can't go changing this. So RAID 5 it is. The only thing I can chamge is the block size, so I will experiment with this. In my mind, as a mail server deals with lots of very small files randomly read/written, I should use the smallest block size. Am I correct?

I've downloaded AJATU. It's designed for testing video capture and storage systems, which presumably handle large sequential read/write tasks. Does this make it the right tool to test a mail server? In any case, I still don't understand what it is telling me fully anyway, but that is down to my lack of understanding I'm sure.
  •  
GlennK

Messages: 252
Karma: 3
Send a private message to this user
Greg, no you don't want to have the smallest block size possible. Tiny block sizes mean larger files will be broken up to hundreds of pieces. You don't want that. To truly understand the best block size you need to analyze how the file will be spread across the drives. It's a bit complicated, but for instance, in a RAID0 (striped), best case scenario would be one file split in 2, for example, a 64k file on 32k blocks. That would be exact split between the two drives. Take a look at what your average file will be and how it is split across the drives, but don't discount the largest and smallest. I believe RAID 5 stripes across all drives, so you might consider smaller blocks. So that being said, with multiple drives 16k may be your best option, which is where you were headed anyway. Smile
  •  
GlennK

Messages: 252
Karma: 3
Send a private message to this user
Regarding the test tool, use it and test with the smallest sizes.
Previous Topic: Kerio Connector
Next Topic: Galaxy Tab 10.1 GT-P7500
Goto Forum:
  


Disclaimer:
Kerio discussion forums are intended for open communication between forum members and may contain information and material posted by members which may be useful in learning about Kerio products. The discussion forums are not intended to provide technical support for any specific product. Any information implied or expressed in the discussion forums is that of the posting member. Kerio is in no way responsible for the information posted in the forums, or its accuracy. Kerio employees may participate in the discussions, but their postings do not represent an offical position of the company on any issues raised or discussed. Kerio reserves the right to monitor and maintain the forums to promote free and accurate exchange of information.

Current Time: Wed Nov 22 00:11:25 CET 2017

Total time taken to generate the page: 0.00382 seconds
.:: Contact :: Home ::.
Powered by: FUDforum 3.0.4.