If you have updated to PSP3, please note the following;
Physical disks added to NMV pools must be initialized in GPT layout and be unpartitioned, this information has been updated in the help file. Existing allocated disks are not converted and remain MBR.
Because GPT disks are now supported for NMV Pools, larger RAW disks can be utilized (greater than 2TB).
If you have an MBR RAW disk that you are trying to add to the NMV pool, it will simply not be listed in "Available Disks" under NMV pool. Convert it to GPT and it will become available to add.
Disks to be used for "Partition born" virtual volumes must still be in MBR format, if you convert them to GPT and create a partition, they will simply not be available to use.
BT.
Wednesday, December 23, 2009
Friday, December 18, 2009
PSP3 for SANmelody 3.0 release
Some nice new features / functionality included in this release
* Added support for 1 petabyte volumes.
Also worth noting is that there is no data migration necessary to take advantage of this capacity increase, once you have upgraded to 3.0.3, you will find that all of your existing NMV's will be "converted/upgraded" to 1PB in size giving you the ability to increase you Vvol size if required.
NOTE: most operating systems at this stage will not be able to support 1PB capacity so please check before assigning large virtual volumes.
NOTE: As always, you will want to managed your raw NMV pool capacity wisely and not over commit to much to your application servers. Don't MAX OUT your credit card!
* Added SCSI Asymmetric Logical Unit Access (ALUA) support.
In the most basic explanation, storage solutions supporting ALUA (Asymmetric Logical Unit Access) give application servers (that support ALUA) more "path state" information allowing them to make educated decisions on path selection for optimal availability and performance with the specific storage vendors architecture mind.
Port States:
Path control under "Fixed pathing policy"
Uses the designated preferred path (if it has been configured), otherwise it uses the first working path discovered at system boot time. If the ESX host cannot use the preferred path, it selects a random alternative available path. The ESX host automatically reverts back to the preferred path as soon as the path becomes available.
source: http://blogs.vmware.com/storage/2009/10/vstorage-multi-paths-options-in-vsphere.html
(Other great information here on Vsphere4 pathing policy)
• Maintain the Persistent Reserve property for a virtual volume that has been split.
• Added Microsoft Windows 2008 R2 support.
This is support for Windows 2008 R2 to run SANmelody, as you should know Windows 2008 R2 as an application server has been supported for some time.
• Added support for HP82E Fibre Channel adapters.
HP Dual port 8Gb Fibre Channel Card support added.
• Added a new ‘offline’ button to the virtual volume properties page.
This sets all application server mappings for a virtual volume offline, providing more granual control of your Virtual Volumes state at the storage level.
Usage Case: You may have a volume that is being presented to a large VMware, HyperV or XenServer environment (many channel mappings) that you want to "offline" access to for some reason (unexpected capacity growth, storage pool full....). Instead of having to restrict access at each application server or unmap the volume at the DataCore Software level, the volume can be simply set offline.
NOTE: Be aware of your Application server states before offlining a volume as channel access to the volume will be set offline.
• Improve diagnostic event log messages for the DataCore capability driver.
• Added a new firewall rule for 'DataCore DCOM Connection' to allow access by DataCore VSS through a firewall without turning it off.
• Ability to Remove Raw disks from Pool if they are mirrored.
From what I can see this feature has been added but is not listed in the release notes, I have had a play with its functionality in my test environment. If you have mirrored your RAW disks in your NMV pool, one member of the mirror can now be remove from the pool without interruption.
BT.
* Added support for 1 petabyte volumes.
Also worth noting is that there is no data migration necessary to take advantage of this capacity increase, once you have upgraded to 3.0.3, you will find that all of your existing NMV's will be "converted/upgraded" to 1PB in size giving you the ability to increase you Vvol size if required.
NOTE: most operating systems at this stage will not be able to support 1PB capacity so please check before assigning large virtual volumes.
NOTE: As always, you will want to managed your raw NMV pool capacity wisely and not over commit to much to your application servers. Don't MAX OUT your credit card!
* Added SCSI Asymmetric Logical Unit Access (ALUA) support.
In the most basic explanation, storage solutions supporting ALUA (Asymmetric Logical Unit Access) give application servers (that support ALUA) more "path state" information allowing them to make educated decisions on path selection for optimal availability and performance with the specific storage vendors architecture mind.
Port States:
- Active Optimized
- Active Non-Optimized
- Standby
- Unavailable
- Transitioning
Path control under "Fixed pathing policy"
Uses the designated preferred path (if it has been configured), otherwise it uses the first working path discovered at system boot time. If the ESX host cannot use the preferred path, it selects a random alternative available path. The ESX host automatically reverts back to the preferred path as soon as the path becomes available.
source: http://blogs.vmware.com/storage/2009/10/vstorage-multi-paths-options-in-vsphere.html
(Other great information here on Vsphere4 pathing policy)
• Maintain the Persistent Reserve property for a virtual volume that has been split.
• Added Microsoft Windows 2008 R2 support.
This is support for Windows 2008 R2 to run SANmelody, as you should know Windows 2008 R2 as an application server has been supported for some time.
• Added support for HP82E Fibre Channel adapters.
HP Dual port 8Gb Fibre Channel Card support added.
• Added a new ‘offline’ button to the virtual volume properties page.
This sets all application server mappings for a virtual volume offline, providing more granual control of your Virtual Volumes state at the storage level.
Usage Case: You may have a volume that is being presented to a large VMware, HyperV or XenServer environment (many channel mappings) that you want to "offline" access to for some reason (unexpected capacity growth, storage pool full....). Instead of having to restrict access at each application server or unmap the volume at the DataCore Software level, the volume can be simply set offline.
NOTE: Be aware of your Application server states before offlining a volume as channel access to the volume will be set offline.
• Improve diagnostic event log messages for the DataCore capability driver.
• Added a new firewall rule for 'DataCore DCOM Connection' to allow access by DataCore VSS through a firewall without turning it off.
• Ability to Remove Raw disks from Pool if they are mirrored.
From what I can see this feature has been added but is not listed in the release notes, I have had a play with its functionality in my test environment. If you have mirrored your RAW disks in your NMV pool, one member of the mirror can now be remove from the pool without interruption.
BT.
Monday, December 14, 2009
SUN x4500 - SANmelody iSCSI / Fibre Thumper
Recently had the chance to configure a SUN x4500 "Thumper" server with SANmelody 3.0.2 for connection to an VMware Vsphere4 environment. I was able to put together some basic benchmark numbers for this very interesting configuration which highlights the flexibility that DataCore software brings in designing storage solutions to fit specific budget and storage requirements.
New to DataCore SANmelody?
By utilising standard Wintel hardware with DataCore SANmelody software, we are able to create a iSCSI and Fibre channel Virtualised storage controller with enterprise features, functionality and performance.
48 x 500GB 7200 RPM SATA
Windows 2008 Standard Edition 64bit
DataCore SANmelody 3.0 Update 2
Configuration:
This was the VERY interesting part, we have a server with 48 SATA HDD's spread across 6 SATA controllers and no hardware RAID support. The plan was to take advantage of new features in SANmelody 3.0 PSP2 and mirror pairs of HDD together in the SANmelody NMV Pool (Thin provisioning Pool).
This provided the following Advantages:
Diagram: Disk Mirror into DataCore Pool.
From the diagram you can see that we paired together drives from different SCSI controllers to spread the workload across all 6 controllers as much as possible. 46 of the drives were mirrored into the pool, the remainining two were mirrored for the OS.
Firstly, this benchmark test was time restrictive so it is a very basic test to check the raw performance of the system from a very simplistic configuration.
Both the SANmelody server and VMware ESXi installation was a basic "default" install with all patches applied, no other fine tuning was made on either system.
The Configuration used to achieve the results below are displayed in the following diagram.
Diagram: Network configuration for test.
Performance Results:
NOTE: The IOmeter test performed was the same used for the "VMware unofficial performance" thread, so feel free to compare numbers. (NOTE: I have 3 workers for this configuration)
http://sanmelody.blogspot.com/2008/06/vmtn-sanmelody-performance-chart.html
Performance results explained
Throughput results were limited by the 3 x 1Gb ethernet connections which were flooded straight away during the tests. We achieved ~312MB/s across the three ethernet connections, however the RAW performance tested at the x4500 achieved ~700MB/s, I am very confident that with Dual 4Gb Fibre connections we would be able to achieve this number to the VMware hosts.
Hi IO workloaded looked to be limited by the CPU configuration on the Virtual Machine, with CPU load averaging 93% utilisation, however we were still able to achieve over 20,000 IO/s on IO intensive workloads.
Conclusion:
New to DataCore SANmelody?
By utilising standard Wintel hardware with DataCore SANmelody software, we are able to create a iSCSI and Fibre channel Virtualised storage controller with enterprise features, functionality and performance.
SUN x4500 Server
Dual CPU - Dual Core 3.0Ghz48 x 500GB 7200 RPM SATA
Windows 2008 Standard Edition 64bit
DataCore SANmelody 3.0 Update 2
Configuration:
This was the VERY interesting part, we have a server with 48 SATA HDD's spread across 6 SATA controllers and no hardware RAID support. The plan was to take advantage of new features in SANmelody 3.0 PSP2 and mirror pairs of HDD together in the SANmelody NMV Pool (Thin provisioning Pool).
This provided the following Advantages:
- Offload mirror functionality to DataCore Software (in turn faster host processors)
- Stripe data across all mirror sets in the pool, effectively creating a RAID 10 configuration.
- Simplify configuration by only using JBOD configuration at hardware level.
- Provide high resiliency against drive failure.
- Optimise storage utlisation and efficiency from a single management point.
Diagram: Disk Mirror into DataCore Pool.
From the diagram you can see that we paired together drives from different SCSI controllers to spread the workload across all 6 controllers as much as possible. 46 of the drives were mirrored into the pool, the remainining two were mirrored for the OS.
Firstly, this benchmark test was time restrictive so it is a very basic test to check the raw performance of the system from a very simplistic configuration.
Both the SANmelody server and VMware ESXi installation was a basic "default" install with all patches applied, no other fine tuning was made on either system.
The Configuration used to achieve the results below are displayed in the following diagram.
Diagram: Network configuration for test.
Performance Results:
NOTE: The IOmeter test performed was the same used for the "VMware unofficial performance" thread, so feel free to compare numbers. (NOTE: I have 3 workers for this configuration)
http://sanmelody.blogspot.com/2008/06/vmtn-sanmelody-performance-chart.html
Performance results explained
Throughput results were limited by the 3 x 1Gb ethernet connections which were flooded straight away during the tests. We achieved ~312MB/s across the three ethernet connections, however the RAW performance tested at the x4500 achieved ~700MB/s, I am very confident that with Dual 4Gb Fibre connections we would be able to achieve this number to the VMware hosts.
Hi IO workloaded looked to be limited by the CPU configuration on the Virtual Machine, with CPU load averaging 93% utilisation, however we were still able to achieve over 20,000 IO/s on IO intensive workloads.
Conclusion:
Very impressive performance results for a 46 SATA drive configuration with no hardware RAID, on a system that is now superseded by the SUN x4540 which promises even faster IO performance.
The performance numbers for "real life" workload (mixed random IO and read/write) were impressive.
Screen shots during setup:
Screen shots during setup:
For anyone wanting to see some screen shots from the SANmelody MMC I have included links below.
http://i282.photobucket.com/albums/kk255/berniebgf/Public/DataCore/x4500/x45-OSdisks.jpg
http://i282.photobucket.com/albums/kk255/berniebgf/Public/DataCore/x4500/x45-disks1.jpg
http://i282.photobucket.com/albums/kk255/berniebgf/Public/DataCore/x4500/x45-disks2.jpg
http://i282.photobucket.com/albums/kk255/berniebgf/Public/DataCore/x4500/x45-pool.jpg
http://i282.photobucket.com/albums/kk255/berniebgf/Public/DataCore/x4500/x45-pool2.jpg
http://i282.photobucket.com/albums/kk255/berniebgf/Public/DataCore/x4500/x45-pool4.jpg
http://i282.photobucket.com/albums/kk255/berniebgf/Public/DataCore/x4500/x45-pool6.jpg
http://i282.photobucket.com/albums/kk255/berniebgf/Public/DataCore/x4500/x45-rawdisks.jpg
http://i282.photobucket.com/albums/kk255/berniebgf/Public/DataCore/x4500/x45-systemdetails.jpg
http://i282.photobucket.com/albums/kk255/berniebgf/Public/DataCore/x4500/x45-OSdisks.jpg
http://i282.photobucket.com/albums/kk255/berniebgf/Public/DataCore/x4500/x45-disks1.jpg
http://i282.photobucket.com/albums/kk255/berniebgf/Public/DataCore/x4500/x45-disks2.jpg
http://i282.photobucket.com/albums/kk255/berniebgf/Public/DataCore/x4500/x45-pool.jpg
http://i282.photobucket.com/albums/kk255/berniebgf/Public/DataCore/x4500/x45-pool2.jpg
http://i282.photobucket.com/albums/kk255/berniebgf/Public/DataCore/x4500/x45-pool4.jpg
http://i282.photobucket.com/albums/kk255/berniebgf/Public/DataCore/x4500/x45-pool6.jpg
http://i282.photobucket.com/albums/kk255/berniebgf/Public/DataCore/x4500/x45-rawdisks.jpg
http://i282.photobucket.com/albums/kk255/berniebgf/Public/DataCore/x4500/x45-systemdetails.jpg
Hoping to have some more time on this machine to tweak the configuration for more performance, will also try and obtain a SUN x4540 for testing as a comparison.
BT.
Tuesday, November 3, 2009
SANmelody 3.0 Install time
Previously (http://sanmelody.blogspot.com/2009/05/sanmelody-30-just-little-patience.html)I have discussed long install times for DataCore Software install's, I found out through feedback on the blog that this can be caused by having no internet connection during install.
Why the delay?
During the software install all signed assemblies goes through a cert check against "crl.microsoft.com" in order to ensure that all of the assemblies are signed with a valid Microsoft cert, if there is no internet connection long delayes are seen.
Work arounds.
If you do not have internet access, before starting your install (or upgrade) you have following options to speed things up.
1. Unchecking "Check publisher's certificate revocation" under Internet Explorer. This setting can be found under tools menu, Options, Advanced, Security.
NOTE: Make sure you re-enable the setting once the installation is complete!!
2. I have not checked this however I have read that you can also simply add the following to your host file.
127.0.0.1 crl.microsoft.com
Why the delay?
During the software install all signed assemblies goes through a cert check against "crl.microsoft.com" in order to ensure that all of the assemblies are signed with a valid Microsoft cert, if there is no internet connection long delayes are seen.
Work arounds.
If you do not have internet access, before starting your install (or upgrade) you have following options to speed things up.
1. Unchecking "Check publisher's certificate revocation" under Internet Explorer. This setting can be found under tools menu, Options, Advanced, Security.
NOTE: Make sure you re-enable the setting once the installation is complete!!
2. I have not checked this however I have read that you can also simply add the following to your host file.
127.0.0.1 crl.microsoft.com
Sunday, October 18, 2009
DataCore Video - 6 years old - content more relevent today
Take a look at the following video from DataCore Software created over 6 years ago, the content in this clip is more relevent today than ever. It demostrates how far ahead DataCore Software was at that time and shows that their core message has not changed in over 12 years!
While watching this video, remember that they had been at it (storage virtualization, storage management, and storage networking) for 5 years already!
Creating software that can Unify, Optimise and Virtualise any vendors storage providing performance optmisation, advanced features and functionality that can traverse any disk technology from any vendor. Breaking down the "single box" storage barriers and short comings.
DataCore 5 year anninversary video
http://www.youtube.com/watch?v=5FJ2go6KHs0
BT.
While watching this video, remember that they had been at it (storage virtualization, storage management, and storage networking) for 5 years already!
Creating software that can Unify, Optimise and Virtualise any vendors storage providing performance optmisation, advanced features and functionality that can traverse any disk technology from any vendor. Breaking down the "single box" storage barriers and short comings.
DataCore 5 year anninversary video
http://www.youtube.com/watch?v=5FJ2go6KHs0
BT.
Thursday, October 15, 2009
Raw disk mirroring - Checking your mirror consistency
Background Ref: http://sanmelody.blogspot.com/2009/10/sanmelody-30-psp-2-released.html
When you are utilising "raw disk mirroring" you will want to become familiar with a new menu option for your NMV pools called "pool status", from this menu you will have access to the following important information.
1. The Status of your NMV raw disk mirrors
-Menu entry
BT.
When you are utilising "raw disk mirroring" you will want to become familiar with a new menu option for your NMV pools called "pool status", from this menu you will have access to the following important information.
1. The Status of your NMV raw disk mirrors
2. Mirror relationship information
3. Mirror build / recovery status
-Menu entry
-Pool Performance GUI
BT.
Raw disk mirroring - non-disruptive disk, array, or vendor substitution
With the release of SANmelody 3.0 PSP2 comes the addition of "raw disk mirroring" which I have touched on in the previous post. I really don't think this new capability has been given enough coverage when you look at the flexibility it provides.
The ability to replace "backend" raw disks without interfering with Virtual Volumes, NMV Volumes or the NMV Pool itself, all while I/O operations continue as normal is an incredible capability.
Let look at a scenario:
Client maintains a large VMware environment serving ~10TB of data across ~40 thin provisioned virtual volumes served from a SANmelody HA environment with "Vendor A" storage at the backend consisting of many 2Gb FC 146GB 7.5K drives... They are replicating this data to a DR site consisting of SANmelody with "vendor C" disk.
Customer has decided to de-comission "vendor A" storage, in place of "vendor B" which contains new 4Gb 300GB 15K rpm drives.
Solution: Utilising raw disk mirroring we are able to serve up new capacity from "vendor B" and mirror the raw disks with "vendor A" without touching any of our provisioned volumes. Once mirrors are established we can disconnect "vendor A" and continue as normal without any interuption to the business or to our DR replication solution.
But I don't run SANmelody or SANsymphony?..
The 2nd piece of this puzzle is "Proxy volumes support". If you have existing vendor storage and would like the type of flexibility discussed above to migrate data, it can be done. DataCore can "pass through" all of your existing served volumes (with file systems) and mirror the data while it is serving I/O..........
BT
The ability to replace "backend" raw disks without interfering with Virtual Volumes, NMV Volumes or the NMV Pool itself, all while I/O operations continue as normal is an incredible capability.
Let look at a scenario:
Client maintains a large VMware environment serving ~10TB of data across ~40 thin provisioned virtual volumes served from a SANmelody HA environment with "Vendor A" storage at the backend consisting of many 2Gb FC 146GB 7.5K drives... They are replicating this data to a DR site consisting of SANmelody with "vendor C" disk.
Customer has decided to de-comission "vendor A" storage, in place of "vendor B" which contains new 4Gb 300GB 15K rpm drives.
Solution: Utilising raw disk mirroring we are able to serve up new capacity from "vendor B" and mirror the raw disks with "vendor A" without touching any of our provisioned volumes. Once mirrors are established we can disconnect "vendor A" and continue as normal without any interuption to the business or to our DR replication solution.
But I don't run SANmelody or SANsymphony?..
The 2nd piece of this puzzle is "Proxy volumes support". If you have existing vendor storage and would like the type of flexibility discussed above to migrate data, it can be done. DataCore can "pass through" all of your existing served volumes (with file systems) and mirror the data while it is serving I/O..........
BT
Wednesday, October 14, 2009
SANmelody 3.0 PSP 2 released.
PSP2 for SANmelody 3.0 has just been released, there are some really great features that have been added which I have expanded on below.
Remember your DataCore Software is constantly being evolved, so make sure you keep up to date with product updates and take advantage of the extra features, functionality and fix's!.
Note: If you are looking to upgrade your environment or make substational changes I recommend you run your plans passed DataCore support as general best practice.
New features:
• Added a new Application Server type (Suse Enterprise Linux Server 11)
• Added support for Snapshot group operations (see Online Help for more information).
Great feature addition that will enable snapshot consistancy across multiple virtual volumes.
If you have database applications and have decided to seperate your DB and Log files (as you would), you will now be able to trigger snapshot group commands to ensure consistency across multiple volumes....
E.G. DataBase1 = E: , DataBase2 = F: , DataBase Logs = L:
"DcsVssAdmin DcsEnable /for=E:,F:,L: /usinggroup=SnapGroupSQL1"
Hopefully, this will also mean that DataCore ASR "Advanced Site Recovery" will soon be supported.. Fingers Crossed!
Direct link to ASR PDF Document
http://www.datacore.com/downloads/ASR_DataSheet_July14.pdf
• Added redundant mirror path support (see Online Help for more information).
This is a very nice addition, we can set "redundant mirror paths" for fibre channel mirror links by configuring target/initiator for both mirror paths at both SDS's. This will allow for more throughput by taking advantage of "full duplex" while still providing redundancy.
E.G. Mirror path1 I/T <---> I/T, Mirror path2 I/T <---> I/T.
Correction: I had listed that mirror paths are no longer show under Vvol channels, I must have missed it because when I checked again they are still listed....
• Added the ability to set mirror volume recovery priority (see Online Help for more information).
More control has been provided to DataCore SANmelody administrators in choosing recovery priorities for mirror volumes based on a scale of 3 (high) --> 0 (low). Previously there was not control over which volumes would be recovered in which order. This is a great new feature.
• Added NMV disk mirroring (see Online Help for more information).
This is an interesting feature, we are now able to "mirror" raw volumes in an NMVpool "another dimension of high availability". I more look at this as another way of commission/de-commission my backend disk systems along side rhe existing method by working at a NMV volume level replacments.
From my testing, once a "mirror" is established between two raw disks, the original disk can be "unplugged" (failed within SANmelody) and then removed from the pool, making it easy to restructure, commission, decommission backend vendor disk technology.
Remember your DataCore Software is constantly being evolved, so make sure you keep up to date with product updates and take advantage of the extra features, functionality and fix's!.
Note: If you are looking to upgrade your environment or make substational changes I recommend you run your plans passed DataCore support as general best practice.
New features:
• Added a new Application Server type (Suse Enterprise Linux Server 11)
• Added support for Snapshot group operations (see Online Help for more information).
Great feature addition that will enable snapshot consistancy across multiple virtual volumes.
If you have database applications and have decided to seperate your DB and Log files (as you would), you will now be able to trigger snapshot group commands to ensure consistency across multiple volumes....
E.G. DataBase1 = E: , DataBase2 = F: , DataBase Logs = L:
"DcsVssAdmin DcsEnable /for=E:,F:,L: /usinggroup=SnapGroupSQL1"
Hopefully, this will also mean that DataCore ASR "Advanced Site Recovery" will soon be supported.. Fingers Crossed!
Direct link to ASR PDF Document
http://www.datacore.com/downloads/ASR_DataSheet_July14.pdf
• Added redundant mirror path support (see Online Help for more information).
This is a very nice addition, we can set "redundant mirror paths" for fibre channel mirror links by configuring target/initiator for both mirror paths at both SDS's. This will allow for more throughput by taking advantage of "full duplex" while still providing redundancy.
E.G. Mirror path1 I/T <---> I/T, Mirror path2 I/T <---> I/T.
Correction: I had listed that mirror paths are no longer show under Vvol channels, I must have missed it because when I checked again they are still listed....
More control has been provided to DataCore SANmelody administrators in choosing recovery priorities for mirror volumes based on a scale of 3 (high) --> 0 (low). Previously there was not control over which volumes would be recovered in which order. This is a great new feature.
• Added NMV disk mirroring (see Online Help for more information).
This is an interesting feature, we are now able to "mirror" raw volumes in an NMVpool "another dimension of high availability". I more look at this as another way of commission/de-commission my backend disk systems along side rhe existing method by working at a NMV volume level replacments.
From my testing, once a "mirror" is established between two raw disks, the original disk can be "unplugged" (failed within SANmelody) and then removed from the pool, making it easy to restructure, commission, decommission backend vendor disk technology.
• Added Proxy disk support Note: this feature is a licensable option (see Online Help for more information)
Extract from help file: "Proxied Volumes are most often used as a temporary measure to migrate data from existing storage to a new architecture or to accelerate a deployment to more quickly receive the benefits of SANmelody features."
My 2 cents.
Proxy disk support enables disks with existing file systems on them to be served up to SANmelody and presented through untouched under the control of SANmelody. Once a volume is selected as a "proxy volume" it is represented a a "Virtual Volume" within SANmelody, this volume can be treated like any other Virtual Volume.
The most common next step is to setup a "mirror" partner for the volumes so the original proxy volume can be removed.
This type of functionality allows customer to migrate large amounts of data to new storage infrastructure while the data is accessable to application server/s.
Some very nice updates here....... Start updating people!
BT.
Friday, October 2, 2009
Virtual Disk Structure - VMware planning
One of the most important aspects of your DataCore configuration is how you structure your disk configuration, time should be spent on planning your disk layout from top to bottom to achieve an optimised configuration.
DataCore has created a "Best Practices Guide" which is available through the DataCore support page with a login, simply search for "1348" or "best practices guide".
http://www.datacore.com/support
From a VMware perspective things can get a bit blury with regards to your virtual disk configuration and how you group your VM's, so I though I would share my experience.
NOTE: Every environment is different, the following should be use as a reference only to get you thinking.
Key advice:
1. No more that 5 VM's per Vvol - best practice
This is recommended for most storage vendors, not just DataCore. As DataCore provides a certain amount of cache per vVol, spreading your workload across more vVol will also optimise your cache allocation and utlisation.
2. Seperate your vmdk's based on workload and place them into the correctly size and configured pool.
example: os.vmdk = pool1, data.vmdk = pool2, db.vmdk = pool3
3. Group your specific workloads into specific pools.
When looking at your workload, I find it easier and more efficient to group my intended workloads into seperate pools based on their requirements.
Example: Pool1 = OS's, Pool2 = Data, Pool3 = DataBases, Pool4 = Archive / Backup
4. For each of your Pools, configure the apropriate amount of disks spindles and raid levels to suite the intended workload. The more spindles the better.
Items to concider: Disk Qty, Raid Level, Lun qty and how / when they are mapped to the pool.
5. Seperate you physical disks groups (raid/luns) into their own pools, don't overlap lun's from the same RAID pack to different pools.
Remember your storage is virtualised and thin provisioned (most likely) so creating new vVol's and moving workloads / data around are straight forward, AS LONG as you have spread your workload out and didn't put all your egg's in one basket!!
Key points to take away:
1. set your pool disk structure performance for intended workload
2. spread you workloads across more vVol (use the flexibility of virtulisation and thin provisioning!!)
3. Size your vVol to suite the intended size requirements (and allow for growth if you need)
Diagram of example layout
As stated, every environment is different and every workload is different please concider this a guide only.
BT
DataCore has created a "Best Practices Guide" which is available through the DataCore support page with a login, simply search for "1348" or "best practices guide".
http://www.datacore.com/support
From a VMware perspective things can get a bit blury with regards to your virtual disk configuration and how you group your VM's, so I though I would share my experience.
NOTE: Every environment is different, the following should be use as a reference only to get you thinking.
Key advice:
1. No more that 5 VM's per Vvol - best practice
This is recommended for most storage vendors, not just DataCore. As DataCore provides a certain amount of cache per vVol, spreading your workload across more vVol will also optimise your cache allocation and utlisation.
2. Seperate your vmdk's based on workload and place them into the correctly size and configured pool.
example: os.vmdk = pool1, data.vmdk = pool2, db.vmdk = pool3
3. Group your specific workloads into specific pools.
When looking at your workload, I find it easier and more efficient to group my intended workloads into seperate pools based on their requirements.
Example: Pool1 = OS's, Pool2 = Data, Pool3 = DataBases, Pool4 = Archive / Backup
4. For each of your Pools, configure the apropriate amount of disks spindles and raid levels to suite the intended workload. The more spindles the better.
Items to concider: Disk Qty, Raid Level, Lun qty and how / when they are mapped to the pool.
5. Seperate you physical disks groups (raid/luns) into their own pools, don't overlap lun's from the same RAID pack to different pools.
Remember your storage is virtualised and thin provisioned (most likely) so creating new vVol's and moving workloads / data around are straight forward, AS LONG as you have spread your workload out and didn't put all your egg's in one basket!!
Key points to take away:
1. set your pool disk structure performance for intended workload
2. spread you workloads across more vVol (use the flexibility of virtulisation and thin provisioning!!)
3. Size your vVol to suite the intended size requirements (and allow for growth if you need)
Diagram of example layout
As stated, every environment is different and every workload is different please concider this a guide only.
BT
Monday, September 21, 2009
VMware vSphere "HCL" listing - SANmelody / SANsymphony
Well, it's official....
Both SANmelody and SANsymphony have been "VMware ready Certified" for some time now with detailed support information available.
Now, both SANmelody 3.0 and SANsymphony 7.0 are listed on the VMware "HCL" for both iSCSI and Fibre Channel on VMware ESX 4.0.
Simply search the HCL for "DataCore" to view results.
http://www.vmware.com/resources/compatibility/search.php
BT.
Both SANmelody and SANsymphony have been "VMware ready Certified" for some time now with detailed support information available.
Now, both SANmelody 3.0 and SANsymphony 7.0 are listed on the VMware "HCL" for both iSCSI and Fibre Channel on VMware ESX 4.0.
Simply search the HCL for "DataCore" to view results.
http://www.vmware.com/resources/compatibility/search.php
BT.
Friday, September 4, 2009
SANmelody - 2.0.4 Update4 released
- New support for Quad port 8Gb HBA supported - Nice
- Other Various fix's.
- Other Various fix's.
Thursday, September 3, 2009
More storage vendors move to leverage "commodity" architecture
DataCore Software has been taking advantage of x86 commodity architecture, leveraging the latest technology to drive cost effective high performance virtualised storage solutions.
It should come as no surprise then that other vendors are "coming around" to this type of strategy.
http://www.enterprisestorageforum.com/hardware/features/article.php/3835961
DataCore Advantage remains.
The advantage of DataCore SANmelody (and SANsymphony) is that it is software, so you choose your commodity (high volume) server to act as the "Virtualised storage head" and save $$ in the process.
It should come as no surprise then that other vendors are "coming around" to this type of strategy.
http://www.enterprisestorageforum.com/hardware/features/article.php/3835961
DataCore Advantage remains.
The advantage of DataCore SANmelody (and SANsymphony) is that it is software, so you choose your commodity (high volume) server to act as the "Virtualised storage head" and save $$ in the process.
Monday, August 31, 2009
FCoE - Support anounced
Friday 28th August, DataCore Software anounced support for FCoE in both its SANmelody and SANsymphony products.
Specifically DataCore lists the support of "CNA" cards in "application servers" connecting to SANmelody or SANsymphony "Virtualised Storage heads"
Example: Emulex LP21002 (Dual Port 10Gb ethernet CNA card)
When used in an Application Servers, this card would appear as two PCI devices.
1. Dual port Ethernet adapter
2. Dual port Fibre Channel adapter
Encaptulation and decapsulation offloaded from the servers, handled by the Emulex card encapsulation/decaptulation engine.
You also have the ability to control (dynamically) bandwidth between "network" and "storage" traffic.
from what I can see, DataCore does not currently support CNA cards in their SANmelody or SANsymphony servers, so you will still need a capable switch to "converge" fibre channel and FCoE hence why support for the Cisco Nexus 5000 is also listed in this anouncement I presume.
So, you could use a CNA card in your application server for FCoE, connect it to a Nexus 5000 via 10Gb Ethernet port (utilising FCoE) ---> Connect your SANmelody or SANsymphony server to the switch via a Fibre Channel module --> then register the application server to SANmelody or SANsymphony as standard "fibre channel" channels. (While also using the same ports on the CNA card for your IP network traffic.
Make sense? ............ That's your "green field" installs I guess..... the other play is to connect your existing Fibre Channel storage fabric to a new "unified fabric" (Cisco's word's) capable switch and "go to town".
Specifically DataCore lists the support of "CNA" cards in "application servers" connecting to SANmelody or SANsymphony "Virtualised Storage heads"
Example: Emulex LP21002 (Dual Port 10Gb ethernet CNA card)
When used in an Application Servers, this card would appear as two PCI devices.
1. Dual port Ethernet adapter
2. Dual port Fibre Channel adapter
Encaptulation and decapsulation offloaded from the servers, handled by the Emulex card encapsulation/decaptulation engine.
You also have the ability to control (dynamically) bandwidth between "network" and "storage" traffic.
from what I can see, DataCore does not currently support CNA cards in their SANmelody or SANsymphony servers, so you will still need a capable switch to "converge" fibre channel and FCoE hence why support for the Cisco Nexus 5000 is also listed in this anouncement I presume.
So, you could use a CNA card in your application server for FCoE, connect it to a Nexus 5000 via 10Gb Ethernet port (utilising FCoE) ---> Connect your SANmelody or SANsymphony server to the switch via a Fibre Channel module --> then register the application server to SANmelody or SANsymphony as standard "fibre channel" channels. (While also using the same ports on the CNA card for your IP network traffic.
Make sense? ............ That's your "green field" installs I guess..... the other play is to connect your existing Fibre Channel storage fabric to a new "unified fabric" (Cisco's word's) capable switch and "go to town".
Protocol comparison courtesy: www.fibrechannel.org
Thursday, August 20, 2009
SANMelody - Software bundles
Was having a look through the Datacore site and notice a "new look" software bundles page.
Take 5 minutes out of your day and check it out.
http://www.datacore.com/products/prod_SANmelody_buy.asp
Take 5 minutes out of your day and check it out.
http://www.datacore.com/products/prod_SANmelody_buy.asp
Friday, August 14, 2009
FusionIO Results - With and Without DataCore
Was lucky enough to receive some fusionIO test results used with and without DataCore SANMelody, the test were performed using "H2Benchw" results below.
Sunday, August 2, 2009
Upgrade W2K3 - 2.0.4.2 --> W2K8 3.0.1
Recently had the pleasure of upgrading an SANMelody AIM Destination server for a client, thought I would share my experience.
Target Machine:
SANMelody 2.0.4.2 server --> AIM Destination role.
Hardware:
IBM x3650 with EXP3000 and DS4700 behind it.
Upgrade Proceedure:
Following the Supplied DataCore upgrade path, the proceedure to update the server was very straight forward.
NOTE: You Should follow the Upgrade guide available from DataCore support pages! (the following is just advice to take onboard)
Outside of the standard upgrade steps with regards to DataCore Software I have added some other steps I performed.
Roll forward......all SANMelody backup information collected, w2k3 server ready for shutdown...
1. Once the SANMelody 2.0.4.2 W2K3 environment was in a shutdown state ready for upgrade to SANMelody 3.0.1 on W2K8, I disconnected external fibre attached storage to be sure not not to overwrite pool allocated volumes during the W2K8 install.
As for SCSI attached devices (internal or DAS), I typically leave these connected as to not caused any "import configuration" requirements for the RAID cards once they loose connectivity to their disks. Or remove the RAID card (if not OS disks) during the OS install.
2. Once OS is install (Basic W2K8 install) I reconnected all storage devices (fibre and DAS). I then performed and update on all drivers and firmware for the IBM server and its DAS devices.
I used the "IBM Update Express system pack installer" to automate all of the driver and firmware updates for this configuration.
http://www-947.ibm.com/systems/support/supportsite.wss/docdisplay?lndocid=SERV-XPRESS&brandind=5000008#uxspinstall
3. Finished following DataCore Upgrade instructions.........
Results of this upgrade were all AIM volumes resuming transferes, (buffered on source server while offline) and all configuration information and data retained.
In the back of my mind I was thinking this would be a cumbersome task (but well worth it), but in the end it was a VERY straight forward proceed.
Shame on me.
BTout.
Target Machine:
SANMelody 2.0.4.2 server --> AIM Destination role.
Hardware:
IBM x3650 with EXP3000 and DS4700 behind it.
Upgrade Proceedure:
Following the Supplied DataCore upgrade path, the proceedure to update the server was very straight forward.
NOTE: You Should follow the Upgrade guide available from DataCore support pages! (the following is just advice to take onboard)
Outside of the standard upgrade steps with regards to DataCore Software I have added some other steps I performed.
Roll forward......all SANMelody backup information collected, w2k3 server ready for shutdown...
1. Once the SANMelody 2.0.4.2 W2K3 environment was in a shutdown state ready for upgrade to SANMelody 3.0.1 on W2K8, I disconnected external fibre attached storage to be sure not not to overwrite pool allocated volumes during the W2K8 install.
As for SCSI attached devices (internal or DAS), I typically leave these connected as to not caused any "import configuration" requirements for the RAID cards once they loose connectivity to their disks. Or remove the RAID card (if not OS disks) during the OS install.
2. Once OS is install (Basic W2K8 install) I reconnected all storage devices (fibre and DAS). I then performed and update on all drivers and firmware for the IBM server and its DAS devices.
I used the "IBM Update Express system pack installer" to automate all of the driver and firmware updates for this configuration.
http://www-947.ibm.com/systems/support/supportsite.wss/docdisplay?lndocid=SERV-XPRESS&brandind=5000008#uxspinstall
3. Finished following DataCore Upgrade instructions.........
Results of this upgrade were all AIM volumes resuming transferes, (buffered on source server while offline) and all configuration information and data retained.
In the back of my mind I was thinking this would be a cumbersome task (but well worth it), but in the end it was a VERY straight forward proceed.
Shame on me.
BTout.
Tuesday, July 28, 2009
SANMelody 3.0 Update 1
SANMelody 3.0 Update 1 - Now available
Various fix's, added support for Sun-rebranded Qlogic qle2532
New proceedure for Updating from SANMelody 3.0 to 3.0 U1.
The update actually Uninstall the old version and installs the updated verions with configuration information save and reloaded. (nice)

In previous versions, updates would be applied over the top of existing installations.
BT.
Various fix's, added support for Sun-rebranded Qlogic qle2532
New proceedure for Updating from SANMelody 3.0 to 3.0 U1.
The update actually Uninstall the old version and installs the updated verions with configuration information save and reloaded. (nice)
In previous versions, updates would be applied over the top of existing installations.
BT.
Sunday, July 12, 2009
OT - Mark Webber Wins German GP!
Congratulation Mark Webber on his first GP win at the German GP!!!
Go Mark!
Next...... Win the Championship!
Go Mark!
Next...... Win the Championship!
Saturday, May 23, 2009
SANMelody 3.0 - Just a little patience ....
Update: Confirmed, install can take extra time if you do not have an active internet connection. If you are connected to the internet the install takes ~2-4minutes.
When installing (and especially uninstalling) SANMelody 3.0 from a Windows 2008 server, have patience.....lots of patience. I have found it can take a substantial amount of time for the inistall/uninstall to complete. There is allot of work going on in the backend. (Adding roles, firewall rules, installing services, driver modules......)
When installing (and especially uninstalling) SANMelody 3.0 from a Windows 2008 server, have patience.....lots of patience. I have found it can take a substantial amount of time for the inistall/uninstall to complete. There is allot of work going on in the backend. (Adding roles, firewall rules, installing services, driver modules......)
Also, don't rush to open the SANMelody MMC snapin when you have just started the SANMelody server up (usually during initial installs). With both of the tasks above, it can take some time for all of the services and drivers to load at the backend. You will most likely received long delays / no response from the MMC if you do..
If your performing these tasks, maybe sit back and enjoy the following clip. ;-)
BT
Subscribe to:
Posts (Atom)




