A Contributed Article by Adam Bogobowicz, Sr. Director of Service Provider Product
Marketing at Parallels
Service providers delivering infrastructure
cloud services today support their infrastructure either on storage area
network (SAN) or direct attached storage (DAS) solutions.
Both of these solutions come with
significant limitations. The problem with enterprise-class SAN solutions is
that they are very expensive, with both up-front and maintenance costs and
therefore only relevant in enterprise IaaS scenarios. The disadvantage of DAS in
addition to the non-cloudy characteristics is that it leads to a significant waste
of disk space and drives costs per GB of storage up, in extreme cases, matching
that of SAN solutions.
As the infrastructure cloud matures, the
ability to separate storage from the physical systems has become increasingly
important. This level of abstraction is now a requirement for delivery of the
infrastructure cloud features such as cloud level storage redundancy and
storage failover. Because of the high cost of dedicated SAN solutions, infrastructure
providers serving the SMB market are left out of the cloud game and need to
relay on DAS based non-cloudy solutions.
The right solution for this problem would
be a system that allows for separation of storage from the physical system,
delivers high performance, is based on existing hardware infrastructure, and
eliminates DAS storage utilization inefficiencies.
A block storage solution that meets these
requirements, Parallels Cloud Storage, was showcased recently by my company,
Parallels, at our annual conference in Las Vegas. You can see my related blog
on this at http://blogs.parallels.com/serviceprovider/2013/2/8/the-joys-of-cloud-storage-demoed-at-parallels-summit.html.
This is a distributed cloud solution that
separates compute from storage and as a result enables compute units to be re-provisioned
on a new physical server if the original physical server (running that compute
unit) becomes unavailable. At the same time, with the storage distributed
across multiple physical machines, a compute unit is no longer dependent on a
single physical machine managing the underlying storage and can run even if one
or more of the physical units in the clusters are lost.
To accomplish this, Parallels Cloud
Storage automatically chunks, replicates, redistributes and balances storage between
all available cluster nodes. You can think about this virtual storage solution
as similar to cloud level RAID system with improved placement intelligence and
compute unit management. In storing new chunks, Parallels Cloud Storage takes
into account both the amount of free disk space and the distribution of the I/O
load, in order to ensure high performance. Additionally, when you add new node
to a cluster, Parallels Cloud Storage automatically balances existing data
across all the nodes.
Parallels Cloud Storage, delivers not
only on the key enterprise class cloud features, but does it without the need
to purchase and maintain expensive SANs. Implementation of DAS-based cloud
storage has a potential of bringing the full power of cloud storage to the SMB
market and democratizing that technology across the industry.
Consequences of this new method of
storage abstraction however may be even more dramatic in the enterprise space
by delivering SAN-like cloud features at a fraction of the SAN cost and
allowing for cloud features to be offered in a standard form across all
enterprise workloads.
###
About the Author
Adam Bogobowicz holds the position of Sr. Director of Service Provider Marketing
at Parallels. In this role, he oversees the company's global service provider
product marketing function, covering the panel automation and infrastructure
technologies and products. Prior to joining Parallels, Adam worked at Microsoft
where he designed, and drove virtualization campaigns supporting Microsoft's
entry into the virtualization market. Earlier in his career Adam worked in IT
organizations where he managed and delivered hosted collaboration services.