A quick note! My team here at EMC has a brand new page dedicated to Microsoft SQL Server: http://www.emc.com/campaign/global/sql-server/index.htm There is an excellent best practices guide to SQL Server running on EMC Symmetrix Storage Systems: http://www.emc.com/collateral/software/solution-overview/h2203-ms-sql-svr-symm-ldv.pdf We work extremely hard to provide guides that help you implement excellent systems. Be…
The Microsoft Exchange Product Group has been hard at work on Exchange 2010 Service Pack 1 for many months and has just shipped Service Pack 1 today. Definately a congrats to those folks who’ve been working like crazy to get this out there!
See the official announcement on the Exchange Team blog at: http://msexchangeteam.com/archive/2010/08/25/455861.aspx. The download is also available from the announcement. This is about a ~500MB d/l so this is a big service pack!
First of all, there is many new features in SP1, this is clearly not just bug fixes. I think most notable are the DAG improvements (block mode for log shipping), the ability to seperate a personal archive database to a separate MBX database, and numerous Exchange Online integration improvements and many others. To see a full list of everything that’s new, see:
What’s New in Exchange 2010 Service Pack 1: http://technet.microsoft.com/en-us/library/ff459257(EXCHG.141).aspx
Release Notes for SP1: http://technet.microsoft.com/en-us/library/ff728620(EXCHG.141).aspx
Just wanted to give you all a quick update on some recent work the solution teams did with regards to SharePoint storage tiering.
SharePoint, as you know (or not?) stores all the content it manages in SQL Server tables. while there is some merit in doing so, it’s mostly disadvantageous when considering larger deployments of SharePoint as a true framework for content collaboration. Those unstructured binary objects once stored in SQL are called BLOBs (Binary Large Objects). As the size of the content databases keeps growing, the main contributor is the BLOB data, which grows significantly faster than any associated metadata; a BLOB would usually comprise ~95% of the content database size.
In order to support the “ECM for the masses” message, Microsoft introduced a couple of APIs to accomplish that externalization task. The first is EBS (External BLOB Storage) that is available since MOSS 2007 SP1 and recently RBS (Remote BLOB Storage) which is available in several flavors for SharePoint 2010.
In this post I’m going to highlight the recent integration work we have accomplished with a MetaLogix product called StoragePoint.
The solutions team in Santa Clara worked on a neat Cloud storage solution for SharePoint BLOBs based on EMC Atmos.
The 3TB SharePoint farm content was externalized to EMC Atmos which dramatically decreased the size of the content databases and demonstrated MetaLogix’s StoragePoint capability to effectively manage the externalization of SharePoint BLOBs to EMC Atmos through it’s Atmos connector. oh and BTW, that farm was 100% Hyper-V.
Atmos on-premise testing shows that the performance is nearly identical to the traditional setup of SharePoint with SQL as indicated in the following table. Results indicate that relocating BLOBs to an external BLOB store (EBS) shows no impact to the overall user experience across the three user profiles simulated:
The Unified Midrange Storage Group (UMSG) in RTP,NC conducted similar tests on EMC Celerra NS-120 in a VMware vSphere virtualized farm.
Various BLOB store flavors were tested in that case, involving EBS BLOB store provisined by FC drives with and w/o Deduplication as well as SATA drives with and w/o Deduplication, all through a CIFS share.
The following figure shows the disk layout of the storage design:
Some highlights from that test:
- SQL disk usage reports after BLOB externalization showed an 88% reduction in size of the content databases.
- After deduplication was completed, 18% of the file system space was saved.
- While externalizing BLOBs presents overall performance improvement in retrieving objects it may add some latency to search and modification activities which in most deployments would represent a smaller precentage than browsing content (Another factor is the size of the BLOBs externalized, the larger the object the more efficient EBS/RBS is).
- Once content is externalized, SharePoint indexing gets a boost. full crawl activity finished in less than 1/3 of the original configuration. I believe it has to do with the nature of indexing which is sequential.
While these two solutions are based on SharePoint 2007, we plan to re-validate it soon on SharePoint 2010, but don’t expect any magic there. I suspect results would be similar.
I believe that BLOB extenalization is the catalyst to SharePoint adoption in larger organizations, leveraging SharePoint ECM capabilities in the Multi-TB club. EMC has a wide range of offerings in that aspect and these two solutions demonstrate only part of it.
I try to keep up with as many people who are interesting and important work in the field of Microsoft’s enterprise application products such as Exchange, SQL Server, and SharePoint and now and then it surprises me when someone really “gets it” in terms of their audience.
Denny Cherry is one of those guys. He presents topics not for the storage geeks among us, but for SQL geeks who could benefit from understanding more about how a SAN is set up and configured. A couple of days ago, he wrote a post summarized a recent presentation he gave to a group of SQL pro’s – a dry run for his upcoming SQL Pass presentation. What caught my eye is that he was generous enough to share his presentation materials online, for all to share and digest – great stuff on storage and virtualization for the SQL DBA.
A very smart dude – and I’ve happy to say I’ve had the pleasure of meeting him at EMC World in the bloggers lounge. I think I also walked past him a few times on Bourbon Street at TechEd, but … I can’t be sure
Anyway…. as an EMC employee, I was very happy to see the references to EMC, but (there’s always a but isn’t there) I did want to correct two minor corrections I would make on the materials:
1. Slide 22 indicates Exchange belongs on FC disks. I would just mention that at EMC, we’ve seen a lot of people put Exchange 2010 on SATA. RAID-protected SATA nonetheless, but SATA combined with Virtual Provisioning (EMC’s term for Thin Provisioning) works very well for most situations. Caveats apply when using SATA and replicating that data, of course – but tread carefully, and it can be done. Thin Provisioning is great for Exchange 2010 because Exchange teams want to give their users enormous mailboxes (up to 25GB in some cases) and they don’t want to buy and allocate all of that space upfront. SATA with Virtual Provisioning is a great way to cut the cost of an Exchange infrastructure that used to demand those Tier 1 FC disks.
2. Slide 43 indicates EMC can only do EMC to EMC array replication. One nice surprise comes from EMC’s RecoverPoint product. It’s a journaling appliance that implements a write-splitter that sits on the host, SAN switch, or in the storage array itself. This replication appliance that splits the writes and keeps a copy in a local or remote journal and uses policy- driven bandwidth reduction and data compression technologies to shrink bandwidth significantly – sometimes by a factor of 5-10x. This is usually enough to justify the purchase of the product, due to the cost savings in bandwidth. Oh… so back to the main point… we have a lot of customers that use EMC storage at their primary site and another vendor’s storage in a secondary location. It’s heterogeneous. You don’t even need an EMC array and it works with FC or iSCSI protocols.
I’m hoping Denny doesn’t take this the wrong way – I learned a lot about what SQL DBA’s need to know after reading this… and I bet you can too. Remember to check out his PPT consider following him on Twitter and check out his blog here.