Riverbed Hyper-converged Edge
Simplicity Pre HCI, a typical application delivery workflow required managing the network, storage and compute as separate entities each with their own management portals. The success and timeliness for the delivery of a new application depended upon the IT admin’s hope of all these segmented silos playing harmoniously with each other. Combining these made manageability of these components much simpler. A single pane of glass to manage all components of a service delivery stack meant less reliance on multiple groups and less chance of human error. Predictability From a business perspective, this model is a lot more predictable than any of its predecessors. Instead of over-provisioning...
Pre HCI, a typical application delivery workflow required
managing the network, storage and compute as
separate entities each with their own management
portals. The success and timeliness for the delivery
of a new application depended upon the IT admin’s hope
of all these segmented silos playing harmoniously with
each other. Combining these made manageability of
these components much simpler. A single pane of glass
to manage all components of a service delivery stack
meant less reliance on multiple groups and less chance
of human error.
From a business perspective, this model is a lot more
predictable than any of its predecessors. Instead of
over-provisioning scale-up architectures to account for
3-5 year growth, now one can design for what is required
and grow as needed i.e. pay for what you need today.
The “just in time sizing” makes budgeting decisions
easier as admins can predict when further scale
might be required based on organizational growth.
From a technical perspective, it also provides linearity
in both capacity and performance. Hyper-converged
infrastructure solutions are built as equal lego pieces
hence stacking more of these provides a linear and
predictable amount of all resources.
Consolidation with Convergence
HCI’s align really well with not only the software defined
next generation data centers but also with the CIO
initiatives of consolidation. Consolidation has always
been a primary point of focus since the beginning of the
first branch office. It gives you greater control and
governance of your intellectual property. It gives you
better bang for your buck for a particular resource.
Take storage as an example, almost 50% of all storage
is unused. Combining multiple buckets of storage means
less number of half empty buckets. Consolidation with
convergence of functions gives you the ultimate bang for
your buck. Instead of just consolidating your storage tier,
convergence gives you the same business benefits for
your entire IT infrastructure stack.
Remote Office Dilemma
We have seen that as IT administrators, Hyperconverged Infrastructures can provide a great
mechanism for architecting and delivering services
because they are inherently simple, scalable, and
predictable. But now let’s shift focus to the consumers
of these services - our customers.
© 2015 Riverbed Technology. All rights reserved.
Majority of the revenue-generating workforce resides
outside the data center at remote office branches. If we
design a remote office’s IT from a user’s point of view,
we find ourselves at a paradox as it conflicts with all the
benefits mentioned above. Let’s take a look at some
Performance equals productivity. Nowadays,
performance and experience are two of the biggest
factors in retaining a customer—internal or external.
If we are not giving LAN speed performance to our
users, they cannot function and hence productivity
suffers. If we put a conventional HCI at these locations
to solve the performance challenge, we forgo the
simplicity and ease of centralized management, data
protection, governance and support
WAN Dependency Dilemma
Many IT decision makers that end up consolidating
everything, try to balance the scales by provisioning
additional bandwidth and sometimes using traditional
WAN optimization solutions. These work for numerous
use cases and applications, however, as a user I am
still WAN dependent. My experience becomes
unpredictable depending upon the time of day and
consumption of services, not to mention WAN outages
that can completely inhibit my ability to get my
Stateful Branch Inflation Dilemma
As IT admins, usually stuck between a rock and a hard
place, we have to balance the consolidation-distributed
service delivery mechanism based on individual
applications. Some services do not work well over the
WAN, also certain applications are mission critical and
require near 100% up-time with no dependency on
WAN or data center infrastructure. These normally
end up at the branch. Once there is infrastructure to
support stateful applications in the branch, they tend
to grow and inflate beyond control. You start with one
service, and with every user complaint, you end up
placing more stateful apps in the branch, resulting in
more and more unmanaged and unprotected data at
these locations. Hence completely losing the control
and consolidation benefits of traditional HCI.
Looking side by side from both the deliverer (admin)
and the consumers (user’s) perspective, we can see this
Table 1 Pros, cons and tradeoffs of traditional HCI at Data Center & Remote Branch Office
• Traditional HCI ONLY at the Data
Center, nothing at the Branch
Service Deliverer (Admin in DC)
• Good Admin experience
• Consolidation. Simplicity
Service Consumer (User in Branch)
• Poor User Experience, Poor
Performance, Poor Productivity
• Zero risk of data loss
• Traditional HCI at Data Center
and at the Branch
• Poor Admin experience
• Distributed management
• Data at risk
• Stateful Branch take longer to
provision and recover
By now, we see the advantages and disadvantages of
conventional HCI solutions. It clearly favors the admin
and the data center, but is not best suited for the users
in the remote offices. Before we re-architect the solution
specifically for the branch, let’s take a look at what a
typical HCI architecture looks like and what architectural
benefits it provides.
• Good user experience,
performance and productivity
• Data Protection issues. High risk
of loss of user data.
Figure 2 Multiple Application Stack
Virtualization logically abstracted individual units of
compute and gave us a horizontal compute layer.
Making the stack looking like this.
If we look at a typical application stack, it looks
something like the following:
Figure 1 Single Application Stack
And architecturally, multiple applications, would look
something like this.
Figure 3 Virtualization In Application Stack
Storage and software defined networking vendors
did the same for the other two layers.
© 2015 Riverbed Technology. All rights reserved.