A new service that leverages Microsoft’s Azure cloud for large-scale extended storage could enable more, and perhaps smaller, businesses to host their own applications in a hybrid cloud configuration.
SIOS Technology’s DataKeeper Cluster Edition is a service that could compel CIOs to think differently about “the cloud” than just that place where all the Dropbox files hang out together.
Most discussions of cloud storage in the pages of CMSWire are about file repositories, file sharing and document-based collaboration. For some CIOs, “the cloud” is the general name for Dropbox or Google Drive or whatever space all those various shared documents cohabitate.
The Success of Failover
So topics like failover clustering don’t usually merit a lot of space here.
That may change because of the following: Many of the on-premise databases maintained by CMS applications are not, in the formal sense, data warehouses. Or data garages or data lakes or any of the other myriad metaphors intended to infuse this subject with interest and intrigue.
But these databases (as we can attest to first-hand) are getting larger. So now we are seeing failover clustering — a common technology for replicating data at the block level to ensure availability — being extended to the cloud.
This trend began quietly in February 2014 with Amazon’s AWS Storage Gateway being adapted to support Windows Server Failover Clustering (WSFC) and VMWare ESX iSCSI Initiator.
This enables applications hosted in VMware’s vSphere virtualization environment to recognize the AWS cloud as its failover cluster — which, in turn, enables smaller businesses with fewer internal data center resources to embrace more powerful on-premise applications, including in the CMS category.
Last week, SIOS Technology began competing with this approach by making its DataKeeper service connect to Microsoft’s Azure using the same WSFC protocol.
This way, SharePoint, SQL Server, Dynamics and other high-volume Windows software (the kind you find running these days in vSphere) can be easily configured to recognize Azure-based storage in a more failure-proof, “shared-nothing” architecture.
It’s the “shared-nothing” part that should get your attention.
“If I asked somebody to draw on a whiteboard what a cluster should look like, they would draw two or more servers and then some type of shared storage, typically that would be a SAN,” said Tony Tomarchio, SIOS’ director of field engineering, in an interview with CMSWire.
Of course, he’s referring to a storage-area network.
Naturally, administering a SAN oneself carries costs.
“But technically, a SAN is a single point of failure,” Tomarchio continued.
“If the shared storage, which all the cluster nodes are connecting into, has an issue, it could take down your entire cluster.”
In a shared-nothing architecture such as a SIOS SANless cluster, all the storage nodes are isolated and independent from one another.
It’s a failover cluster that, at least architecturally speaking, more legitimately earns its title because it’s more failure-tolerant.
SIOS had been producing block-level data replication software that pools together local storage devices into SANless, but shared, arrays. This way, higher-speed servers using local SSD devices (all-flash memory) could be leveraged as failover clusters — a lot less expensive than, say, an all-flash SAN.
“We present our replicated disk as a cluster disk, which looks and feels like it’s shared storage to the cluster,” explained Tomarchio, “when in reality, it’s not.
Which application you decide to protect within the cluster, is completely up to you.”
How This Relates to CMS
CMS systems such as Drupal cast their databases in an abstract fashion, so that they’re not restricted to single database formats. This abstraction means that drivers are used to connect Drupal to common database management systems such as SQL Server.
So it is no coincidence that the data most often replicated on SIOS SANless clusters, as Tomarchio told us, belongs to SQL Server. Highly available (HA) file-shares or HA Hyper-V virtual machines and SAP Central Services are other known use cases for DataKeeper, he noted.
“Because DataKeeper is a replication technology, we integrate with the cluster,” he said. “So this is an excellent solution for any clusterable service or application — not just SQL Server.”
View this article at CMS Wire