2009-06-18 09:22 AM
I know of many customers using de-dupe for LUNs in VMware environment for VM OS images - it seems to make sense, as savings are substantial & performance overhead in most cases remains unnoticeable.
So here is the question:
How about key MS apps - Exchange, SQL, SharePoint - is there already some existing experience to share showing that de-dupe can be beneficial in this environments?
2009-06-18 10:08 AM
Running dedupe on an Exchange server does not typically yield significant savings, so is probably not worth it. However, as the emails are moved to an email archive the typical savings are 30%, which is significant.
I do not have any specific experiences with SQL data, but running deduplication on live sharepoints data usually provides around 30% savings also.
In general, these savings are on the conservative side of the ranges I have seen. The best approach for determining the amount of expected savings with your data set is to contact your local NetApp account team to discuss using the Space Savings Estimation Tool (SSET). SSET is a tool that can be run against a data set, whether it is stored on NetApp storage or non-NetApp storage. It will provide an estimation for savings, with about 5% margin of error.
You can refer to the the deduplication Deployment and Implementation Guide for best practices, and it also includes a table of typical savings for some specific data types.
2009-06-18 02:08 PM
Many thanks for that.
To be honest though I was more curious about a view on a balancing act between saving space & facing potential performance problems (assuming there is any). It's something like with snapshots - people are using them in production environment, because their benefits outweighs by far negligible performance degradation.
It would be really handy to provide some real-life examples where customers are using de-dupe for these types of data & see this as something worth recommending to others.
(I am aware there is neighbouring post about performance in particular)
BTW - I never thought about SSET being able to asses non-file-sharing data.
2009-06-24 05:08 PM
Radek - I'll bite...
I do have a number of customers doing de-dupe on Exchange and SQL Server, but I don't know that I really recommend it. The savings I have seen are minimal (10%-30%) but because the Exchange and SQL are typically LUNS (usually iSCSI in my case), the OS doesn't see any savings and you only save some space in the volume. I guess if you need the space it is a great idea (more room for snapshots!) but at the end of the day you don't see the large savings gain like you do in a VMWare envionment.
As an aside, We have even set up a few VM's of Exchange for smaller customers running Snap Manager for Exchange inside the VM's by software iSCSI mapping the Exchange LUNs. In that situaion we will typically dedupe the os of the vm's but give the customer the choice if they want to dedupe the Exchange data luns.
2009-06-26 02:52 PM
We're not seeing huge space savings for Exchange....no real performance issues either though.
I think the main reason for Exchange is Exchange's built-in single instancing inside each storage group. But.....
We are seeing decent space savings in CIFS environments though. For SharePoint it's driven by whether people are using file versioning (if they are, then dedup is the best thing since sliced bread).
2009-06-26 03:34 PM
It is probably slightly off topic, but out ot my couriosity: why do you reckon Microsoft [edited as this stuff may be covered by NDA]?
Doesn't make much sense to me (or rather sounds like a shoot in own foot )
Message was edited by: Radek Kubka