Point PAPER

Date

SUBJECT: DADMS Submission Request for Atlantis Storage Virtualization Solution for VDE.

BACKGROUND.
The Liquidware Stratosphere assessment taken as part of the Phase One (TRANET-U) installation
revealed a potential shortfall in network storage data response (IOPS) during specific daily time periods.
This shortfall in response time could manifest as undesirable performance delays to the students and
instructors accessing the virtualized desktops. Proposed solutions from storage vendors involved
procuring additional hardware storage devices at costs ranging between $80K to $400K.

Included in the proposals for the more aggressive Phase Two (TRANET-C) installation is an option for
an alternate storage solution employing a virtualization application titled Atlantis. By injecting this
application into the configuration, a significantly smaller SANS storage device can be deployed. For
Phase Two this option reduces the proposed cost of the project by nearly $150K. While the application
vendor reports of a successful deployment of Atlantis at an Army medical facility, the application has not
been entered into DADMs, thus is unavailable as an option to CID.

1. PURPOSE. The purpose of this paper is to request NETC action to add Atlantis to the DADMS
database and make it an option for CID to test for deployment as part of the VDE architecture installed
during VDE Phase Two.

2. DISCUSSION POINTS. A key maxim of knowledge management is that there are "few unique
problems". Although the deployment of virtualized IT infrastructure technology is just entering maturity,
commercial industry is already responding to situations where user growth/demand challenges the
resource availability initially designed into their virtual infrastructures. It is well recognized that the
most likely performance bottleneck within a virtual topology is the network storage substructure. Not
only is it is here that disk images are replicated, made available for distribution to users, all keyboard
actions made by the user pass into these stored images for processing. If storage memory becomes
burdened, the result is delays in response to keyboard strokes, screen refreshes and potential data errors
for the users.

The traditional response to increased IOPS demand has been adding additional network storage
capability. This example of "storage sprawl" is directly associated with the general "server sprawl" that
the new DoN CIO IT footprint consolidation policies have been implemented to eliminate. As a result,
IT departments find themselves in a position where server hardware is significantly difficult to procure,
including network storage devices.

Atlantis ILIO is a new virtual application developed address the "storage spawl" issue by creating a
virtual network storage device located just "outside" of the physical storage device, to capture and
replicate (de-duplicate) recurring input/output calls (IOPS) that normally would go directly to the virtual
images on the storage device, thus easing the burden on the physical device. Tested with both VMWare
and CITRIX, this solution reportedly increases performance by as much as 70 to 90 percent.

3. ASSESSMENT. Atlantis ILIO represents the opportunity to "use a virtualization approach to resolve
a virtualization problem".

4. RECOMMENDATIONS. Based on the assessment, recommend

actions the decision maker should take in anticipation of possible
outcomes. Recommendations should support the organization’s goals
and the decision maker’s intent. If the issue is complex, you may have
to coordinate the point paper before sending it to the decision maker.

5. POC. List point of contact, office symbol, telephone, and e-mail.

Applications and virtual machines run slowly. Problems can be traced to disk I/O bottlenecks and programs running out of disk space because 5
to 15 times more workloads are competing for the same storage resources

Virtually everything hinges on storage. VMware® vSphere™ and VMware® View™, in particular, place extraordinary demands on it. Too
often, disks slow down, interrupt or endanger these centralized IT operations not because they are poorly designed or built, but because
they are physically constrained. VMware Virtual Infrastructure helps you to overcome similar limitations in CPU and memory however
when it comes to advanced functions such as workload migration, load balancing, fail-over and disaster recovery, server virtualization is
completely dependent on highly available (HA) shared storage. You will be shocked at the high hardware costs and major overhaul
generally proposed to put such a storage infrastructure in place.
DataCore™ storage virtualization software delivers a radically simple, high availability solution to meet vSphere shared storage requirements.
The software abstracts your storage into idealized, virtual disks akin to virtual machines. It pools and mirrors disk blocks across available devices,
despite differences in make and model. In the process, it speeds up I/O response and throughput using extensive SAN-wide caching. This lets you
take optimum advantage of VMware’s full suite of capabilities without hesitation. A plug-in for VMware vCenter™ allows you to non-
disruptively provision, share, clone, replicate and expand virtual disks among physical servers and virtual machines. DataCore yields the highest
availability, fastest performance and fullest utilization from your storage assets, making it an essential element of your VMware Virtual
Infrastructure deployment

The solution processes storage IO traffic within the hypervisor to decrease the amount
of IO sent to storage and boosts application performance. As a result, storage is used
more efficiently as less is needed for a given application, and more applications can be
supported by the storage fabric.
Atlantis ILIO FlexCloud 1.0 can be configured to use server memory (RAM) as tiered
storage enabling clustered applications that require processing massive amounts of
data to operate efficiently in scale out shared storage environments.

 While selecting the right storage is important during all phases of server virtualization
deployment, it becomes increasingly critical as the scope of these environments widens to
include more business-critical applications and the virtual machine (VM) count reaches
the hundreds or thousands range.

As environments scale, previously manageable storage issues (e.g., issues related to cost
and complexity) can quickly become unmanageable, and in the process, can erode
important efficiencies gained on the server virtualization side.

A key concern is storage sprawl. Similar to server sprawl, storage sprawl results from
organizations having to deploy net-new systems to meet application demands for more
capacity and performance. In storage circles, “scale out” is being commonly used to refer
to storage architectures that scale in this fashion. “Scale up,” in turn, refers to
architectures that have the ability to meet the same capacity/performance requirements
natively – that is, within the same physical foot print. In other words, it keeps
organizations from having to build out unnecessary infrastructure (storage, network, etc.)
to support the virtualized server environment.
The less infrastructure there is to buy, manage, protect and maintain, the bigger the
potential cost-savings for organizations. The fewer infrastructures there are, the less floor
space that’s needed, the less energy that’s needed to power and cool the environment,
and, importantly, the less administration time that’s needed to manage the systems, and
so on.

Storage architectures that enable organizations to scale “up” and “out” have clear
advantages particularly in performance-demanding environments (e.g., environments
with more than 2,500 virtual machines). This type of architecture analysts contend
provide the best of both worlds: the flexibility of a “scale out” architecture to meet
specific application, site or geography requirements and the efficiency and cost-savings
benefits of a “scale up” architecture.

Heterogeneous storage virtualization is the key enabler of “scale up and out” storage
architectures. It provides organizations with access to storage capacity and performance
across their data center. Thin provisioning, if available, then doles out the capacity and
performance to applications as needed, and if available, dynamic tiering then stores data
on the appropriate tier of disk based on its importance to the organization, risk factors and
access patterns.

Side Effects of Server Virtualization

Economically Superior Storage Architectures

Price does not equal cost. In fact, in the case of storage, it’s estimated that acquisition
costs account for just 20 percent of its total cost. CAPEX and OPEX costs such as
hardware and software depreciation and maintenance; labor; floor space and power; data
protection and disaster recovery; and data management/mobility account for the other
80%.

This point is extremely important point for organizations to remember when selecting
storage systems for virtualized environments.
Features like thin provisioning, storage virtualization, dynamic tiering and data migration
can have a significant impact on CAPEX and OPEX costs; they can produce a level of
cost reduction and savings for organizations that is complementary, or additive to, the
benefits of server virtualization.

However, it is important to stress again that not all storage architectures are created
equally. Economically superior storage architectures are able to tier, thin and virtualize
heterogeneous storage environments.

Finally, it’s also important to note that cost savings are just one dimension of the
transformation this type of architecture can enable. Other benefits include improved
levels of efficiency, IT agility and improved application resiliency, for which end-to-end
(server to storage) virtualization is a prerequisite.

Improved Business Resilience

By reducing the overall number of physical servers needed to support its business,
organizations are able to extend protection to more applications and data types.

Doing so has important cost-saving benefits. In challenging economic times, having

proven, extensible disaster recovery and business continuity processes in place is more
important than ever before. Downtime from system failures translates to lower
productivity and lost revenue. In fact, an Enterprise Storage Group 2009 study finds that
85 percent of all companies that suffer a major data loss or significant downtime are out
of business within a year. Even brief outages can drastically affect customer satisfaction
and employee productivity.

While server virtualization is the primary means to improved business continuity and
disaster recovery and reduced business risk, storage is an integral component. For
example, how (e.g., synchronously or asynchronously) and where (e.g., at system- or
host-level) the replication is done can have significant business and IT implications,
affecting recovery capabilities as well as physical and virtual server performance.

Also, similar to server virtualization, storage virtualization reduces the physical resources
that must be deployed, managed, and protected. This gives organizations the ability to
extend protection to additional applications