StorSimple is a hybrid storage array by StorSimple Inc. which was a Santa Clara, California-based small company. In November 2012, Microsoft acquired StorSimple. The hardware portion of the solution is manufactured and distributed by Xyratex, a UK company, which was acquired by Seagate in December 2013.
- 2 RU rack mount unit
- This is a block device that serves out storage via iSCSI only
- Dual controllers running in active/passive mode
- Dual hot swap power supply units
- 4x 1 Gbps NICs
- 3 storage tiers:
- SSD tier
- SAS teir
- Cloud tier
- Automated tiering.
- Inline automated deduplication
- Automated data compression
- Choice of cloud provider. Major cloud providers like AWS, Azure, and Google are supported
- Local and cloud snapshot. A snapshot is a point in time recovery point.
- Cloud clone. A cloud clone is the ability to perform a cloud snapshot to multiple clouds in a single operation
- Fast recovery. Upon restoring a volume group, only the volumes’ metadata is downloaded, and the volume is unmasked to the host(s). To the host, the volume and its files appear and can be used. As a file is opened by a user, StorSimple fetches its blocks from the cloud (3rd tier). This is much quicker than waiting many hours to download a 20 TB volume for example before any of its files can be accessed.
- SSL encryption for data in transit, and AES-256 bit encryption for data at rest (in the cloud – 3rd tier)
- Protection policies can be setup in StorSimple to use combinations of local and cloud snapshots to match the organization RPO, RTO, and DR (Recovery Point Objectives, Recovery Time Objectives, and Disaster Recovery)
- Easy administration via simple web interface
- No-downtime controller failover, allows for firmware/software updates with no down time.
- Authentication to admin interface can integrate with LDAP directories like Active Directory.
In summary, this compact array makes it easy to move 400 TB of rarely used files for example, and move them seamlessly to the cloud. Best part is that IT folks don’t have to redesign applications or do anything major, other than mount an iSCSI volume on an existing file server. This offloads expensive on-prem primary storage. It removes the need for secondary storage, tape backup, and off-site backup. This provides a significant return on investment.
- 1 Gbps NICs (no 10 Gbps NICs) put a limit on the throughput the array can deliver. At a maximum, 3 out of the 4 NICs can be configured for iSCSI providing an aggregate bandwidth of 3 Gbps or about 10k IOPS
- Maximum outbound useable bandwidth to the cloud is 100 Mbps
- When planning a use case for the device, tiering to the cloud must be considered. This is desirable for a workload like a file share, but not recommended for an active workload where the entire volume is active such as SQL or Hyper-V.
- Wish list items include:
- No central management point if you have several devices. Each device has its own web interface.
- No Powershell support
- No Operation Manager Management Pack. Monitoring is via email and SNMP only.
- No SMI-S provider (needed for Software-Defned storage in VMM for example)
StorSimple 8000 series comes with new exiting features. One of them is the ability to have a virtual device. Microsoft offers the StorSimple 1100 virtual appliance as an Azure VM that can connect to a StorSimple 8000 series array’s Azure Storage account and make available its data in case of a physical array failure. There’s no cost for the StorSimple 1100 virtual device, but the VM compute and other IAAS costs are billed separately. So, the recommendation is to keep the StorSimple 1100 Azure VM powered up only when needed.
To setup a StorSimple 1100 virtual appliance:
1. In the Azure Management portal, click StorSimple on the left, then click on your StorSimple Manager service, click the Devices link on top, and click Create Virtual Device at the bottom:
Enter a name for the new StorSimple 1100 virtual device. Pick a Virtual Network and Subnet. Pick a Storage account for the new virtual device, or create a new one. Check the “I understand that Microsoft can access the data stored on my virtual device” box:
- A Virtual Network and Subnet must be setup prior to starting to setup a StorSimple 1100 virtual appliance.
- It’s recommended to use a separate Storage account for your StorSimple 1100 virtual appliance. If you use the same Storage account as your StorSimple 8000 series array, the data stored there will be subtracted from your array’s Storage Entitlement (more costly than a separate Storage account)
- The “I understand that Microsoft can access the data stored on my virtual device” checkbox is a reminder that data is encrypted at rest only. That is when it resides in the Storage account associated with your StorSimple 8000 series array. Once this data is accessed and used by a StorSimple 1100 VM, it’s not encrypted. Stored data is encrypted, compute data is not. In other words, if data in your Storage account is accessed by unauthorized individuals, it’s still safe due to EAS-256 bit at-rest encryption. However, if data in your StorSimple 1100 VM is accessed by unauthorized individuals, it’s no longer safe since compute data cannot be encrypted.
- This can take up to 15 minutes
2. Back in the Devices page, we now have the newly provisioned StorSimple 1100 device:
Note: StorSimple 1100 virtual device has a maximum capacity limit of 30TB.
Click on it, then click on the Complete Device Setup, or click the Configure link on top:
Also enter the device Administrator and Snapshot Manager passwords.
3. Similar to a StorSimple 8000 series, setup Volume containers, volumes, Access Control Records
To use a StorSimple 1100 virtual appliance in case of a StorSimple 8000 device failure:
1. In the Azure Manage portal, click StorSimple on the left, then click on your StorSimple Manager service, click on your StorSimple 8000 device, and click Failover at the bottom:
2. Select one or more Volume Containers:
Note: It’s recommended to keep total volumes’ capacity in any given Volume Container under 30TB since a StorSimple 1100 virtual device has a 30TB maximum capacity.
3. Select the target StorSimple 1100 virtual device to fail to:
Check the summary screen, and check the box at the bottom:
This has created a failover job. Jobs can be viewed in the StorSimple Manager service page under the Jobs link on top.
Once this failover job is complete, we can see the Volume Container that used to be on the StorSimple 8000 series array, now on the SVA:
So far we’ve made available the StorSimple 8000 series array’s volumes on the SVA we just spun up. The volumes in this Volume Container on the SVA will have the same ACRs as the original volumes on the StorSimple 8000 series array, which will make the volumes available to the original on-prem servers.
This is how the topology looks like before the failover:
and here how it looks like after the failover:
Considerations and limitation:
- In this scenario of failover, we have to mount the new volumes presented from the SVA onto the file server, and recreate the SMB file shares.
- SVA has a maximum capacity of 30TB. So, volumes larger than 30TB cannot be recovered without a replacement physical StorSimple 8000 series array.
- More than one SVA may be needed to provide enough storage capacity for all volumes that need to be recovered.
- SVA has one vNIC. MPIO is not supported.
- SVA does not have controller redundancy. So maintenance that requires SVA reboot or down time will cause volumes to be unavailable.
- Data on the SVA is accessible to Microsoft. If the SVA is compromised data is unprotected.
- Volume access speed will be slower than usual due to WAN link latency
Microsoft suggests taking it a step further and creating new Azure VMs as file servers:
This recovery scenario suffers from all the above limitations in addition to:
- Need to remap drive letters on all client computers for all recovered volumes
- Need to reconfigure all applications that use UNC paths that point to the original servers to the new VMs
- Azure virtual file servers will have to join the domain
- Additional time, costs, and complexity of creating new Azure virtual file servers
Some of these steps may not be needed if a name space is used to access network drives.