For the IT systems Chivers uses Evolven IT Operations Analytics, which applies analytics to configuration management. He purchased the software to assist with the company’s 2012 move from the building it called home for 48 years.
“What would have normally taken hours to figure out, with Evolven we just ran a report and found what was different,” says Chivers. The software is used on an ongoing basis
for compliance. What used to require a full time engineer to document is now done with an automated quarterly report.
2. Data centers have two completely
separate “brains”: one for monitoring
and managing the power and cooling
and the other for the IT hardware,
applications and services. But there
could well be value in joining them
together in order to improve the
analytics.
by Drew Robb
Left Brain vs.
Right Brain
Data Centers
July/August 2014 Data Center Management 2322 www.afcom.com
3. 24 www.afcom.com
One flavor of pop psychology asserts
that people are either left-brained
or right-brained. This concept arose out
of the work of Roger W. Sperry who sev-
ered the nerves joining the two sides of
people’s brains. Despite serious functional
problem among his experimental subjects,
Sperry achieved a lifetime achievement
award from the American Psychological
Association.
Later research demonstrated that either side
of the brain can perform the same analyti-
cal functions and that those skilled in math
used both sides. Although professionally
the concept has gone the way of the lobot-
omy and libido theory, the left brain/right
brain myth lives on in popular culture.
Data centers typically have two completely
separate “brains”: one for monitoring and
managing the power and cooling, and the
other for the IT hardware, applications and
services. Now the idea is to join those two
hemispheres in order to improve the ana-
lytics.
New technology providing highly granular
floor monitoring and analytics can help data
center managers fine-tune operations, pre-
vent hotspots, and respond instantly before
Power Usage Effectiveness (PUE) or thermal
boundaries are exceeded. Some of these
can marry up with IT monitoring tools to
ease analysis of what equipment or work-
loads are causing the spikes. While these
tools eventually will have the ability to take
automated actions to correct environmen-
tal anomalies, data center managers appear
reluctant to relinquish control.
“I’m not sure I’m willing to have a sys-
tem control my SLAs just yet,”says Bruno
Berti, director of product management
at RagingWire Enterprise Solutions, Inc.,
which is 80 percent owned by NTT
Communications Corp.
Analytics Options
Data center operators have three broad
options in setting up analytics on their
systems: using the analytics functions now
being built into Data Center Infrastructure
Management (DCIM) software, importing
their operational data into an analytics
package or building an in-house product.
Custom analytic tools might just mean an
Excel spreadsheet but can get far more
sophisticated such as the platform Digital
Realty Trust is developing for its more than
100 data centers around the world. In most
cases, however, data center managers will
acquire an off-the-shelf product and cus-
tomize it as needed.
“Data center managers already have a lot of
tools and information depositories in place,”
says Rhonda Ascierto, research manager of
data center technologies and eco-efficient
IT for 451 Research, “but there is really no
neat way to combine all that different data
and all the different tools that they have.”
This has led to the increasing use of DCIM,
but the market penetration is still low.
Ascierto estimates that no more than 15
percent of mid- to large size data centers
(3MW and above) have deployed DCIM
and a lot of deployments are in a single
data center, not across the whole portfolio.
While at one time DCIM tools did not live
up to their hype, the problem these days is
not with the technology but the processes
needed to manage the data center.
“The technology is doing what it promises
to do, but it will require process change
around the integration between the opera-
tional facilities staff and the IT department,”
says Ascierto. “Those departments have
different skill sets, staff, budgets, languages,
platforms and analytic tools.”
She says the two core components of DCIM
are monitoring of power and environmen-
tal systems, and change and configuration
management. Those two areas then lead
to the broad segmentation in the market,
between DCIMs that arise from vendors
such as Emerson Electric and ABB on the
power side, and those coming from the IT
side. In addition, there are startups address-
ing at least some of the functionality (Figure
1, 451 DCIM Suppliers Map).
“Data centers can gain visibility: if I do a
server refresh with this model of server,
what will it do to my operating environ-
ment, what will it to do my cooling require-
ments, how will the temperature, humidity
and air flow change,”says Ascierto. “From
there you can do a business analysis to
determine what are the power savings, and
the impact on availability, risk profile and
redundancy.”
Combining Disparate Systems
But getting to that point is a long-term,
multi-stage process, not just a matter of
getting the right software.
“We are experiencing a lot of growth, which
is one of the key drivers for us putting in the
DCIM system,”says Berti.
RagingWire, for example, operates data
centers in Sacramento, California and in
Ashburn, Virginia, about ten miles outside
the Washington, DC beltway. The two build-
ings in the Sacramento campus have about
500,000 square feet, 80MW of generation
and 37MW of IT capacity. A third building
with 80,000 square feet of white space is
coming on line later this year.
Although the new building is a mile away
from the other two, all three are connected
by fiber rings and appear to customers as
a single campus. The 150,000 square foot
Ashburn building has 28MW of genera-
tion and 14MW IT load. The company is
also breaking ground on a 78 acre parcel
in Ashburn and where it is permitted for
another million square feet.
Last year, Berti moved to RagingWire from
SunGard Recovery Services and took over
the project of setting up the company’s
DCIM. He ran the RFP process and selected
CA Technologies’DCIM, which supported
a secure, multi-tenancy model. Out of the
box, he said, it had a mobile client that
operations staff can use when out on the
data center floor.
While at one time DCIM tools did not live up to their
hype,theproblemthesedaysisnotwiththetechnology
but the processes needed to manage the data center.
4. 26 www.afcom.com
a 3D model, collecting power information
on all the branch circuits and the alarming
information from PDUs, UPSes, switch-
boards, generators and AHUs.
“CA has a standard set of reports that we
were able to use, including trending reports
and dial-based dashboards to see power
consumption, voltage and power factor,”
says Berti.
RagingWire is now rolling out the same
functionality to all buildings and adding
temperature and humidity monitoring to
manage hot spots. That will be followed by
the integration of physical security includ-
ing cameras, badging and other systems.
At this point, RagingWire is only using the
DCIM to analyze the infrastructure, but may
expand it to the racks.
“Before we were managing alert screens,
now we actually have graphical floor plans,
can look at all devices, click on the device
and see exactly the information collected in
the last poll as well as trend in the last hour,
24 hours, last week, month, year. We never
had that before,”says Berti.
Focusing on the IT Side
While RagingWire and Digital Realty Trust
use analytics on their infrastructure, Dave
Chivers feels that area is under control.
“We have APC by Schneider Electric racks
and cooling system and they have a pack-
age that monitors the entire environment,”
says Chivers, VP and CIO of VSE Corporation
(Alexandria, Virginia) which provides engi-
neering, development, testing, and man-
agement services to federal customers.
“There is no predictive data, just real-time,
but I will notice if my power consumption
goes up and can see when there are peaks
on the network and can tell when other
processors kick in heavily.”
For the IT systems Chivers uses Evolven IT
Operations Analytics, which applies ana-
lytics to configuration management. He
purchased the software to assist with the
company’s 2012 move from the building
it called home for 48 years to a new LEED
Gold building. The company has to comply
with a number of regulations, so when it
came time to move its 360 mostly-virtual-
ized servers onto new equipment at the
new location, it had to be able to validate
that each server, physical and virtual, was
identical to the one it replaced. EMC and
VMware had products to do the failover
from the old site to the new one, but even
so there were some errors. For example,
when one server came over, the older ver-
sion of a Dynamic Link Library (DLL) was
copied over the newer version when it
failed over and the server wouldn’t run.
“What would have normally taken hours to
figure out, with Evolven we just ran a report
and found what was different,”says Chivers.
The software is used on an ongoing basis
for compliance. What used to require a full
time engineer to document is now done
with an automated quarterly report.
Chivers says that if he pushes a patch out,
the software
will analyze
whether the
systems that
need the
patch get it
and if not it
will alarm. Or
it will notify
Chivers if a
processor is
operating at
90 percent
capacity 60 percent of the time. It will then
monitor that process for the next month or
two and if the issue persists, advise that a
faster processor is needed.
Comparative Analytics
For Digital Realty Trust (DRT) an off the shelf
product wasn’t an option. The company
owns and operates about 135 data centers
worldwide: some are custom designed for a
single customer, others use the company’s
1.2 MW Turn Key Flex (TKF) pods and still
others are existing facilities of various types
of configurations that DRT manages. Last
year, it hired David Schirmacher as senior
vice president of Portfolio Operations where
he oversees building and deployment of its
DCIM.
The package developed for DRT is called
EnVision. DRT started rolling it out in the
summer of 2013 and by this spring was
deploying it at four or five locations month-
ly. A single TKF can return 29,000 data
points per minute, and there might be 20
TKFs in a single data center. By the end
of the year, they expect to have Envision
deployed at all U. S. locations and to have
started on overseas data centers.
Schirmacher says that while existing DCIM
systems are good at gathering information
and presenting it in terms of absolute val-
ues they don’t give what he terms “compar-
ative analytics.”
“Returning lots of absolute values doesn’t
give the true realization of the power of
DCIM,”he says. “It is more important to see
the performance of that device relative to
the hundreds or thousands of other devices
that are doing the same job.”
For most data center managers, that isn’t
an issue as they only manage one or a few
locations. But given the scope of DRT’s port-
folio, they have enough data to do internal
benchmarking between locations and see
where they can make improvements. This is
used by DRT’s managers and is also provid-
ed to customers.
“If you were my customer, what if I could
say, of our other 100+ locations, the value
is consistent or better than the average of
those 99 other locations,”says Schirmacher.
“Not only the absolute value, but confi-
dence that the value is consistent with
good performance.”
Data center managers can monitor real time
information, but also use the comparative
data to see which data centers are achiev-
ing the best results so that those practices
can be adopted elsewhere. Schirmacher
says that, while Envision is capable of ana-
lyzing workloads, the software is not set
up to do so. DRT provides the data center
space and infrastructure, but customers
generally own and operate their own IT
gear. But the customers do have their own
interface into the system to view any data
related to their own equipment.
EnVision is designed to be used by a wide
variety of users. As a result, it only provides
information, not the ability to control any of
the monitored systems. Sales engineers can
use it for selecting the best place for new
equipment. Finance can use it for billing
customers
for power.
Customers
can view their
power con-
sumption and
infrastructure
status. Data
center man-
agers can use
it for predict-
ing future
problems.
“Rather than waiting for something to over-
load, which might drive the temperature
up or might drive electrical capacity to its
brink,”says Schirmacher, “we can spot the
trends where there is a growth trajectory
well before that problem occurs.”
Drew Robb is a freelance writer based in
Florida.
“It is more important to see the performance of
that device relative to the hundreds or thousands
of other devices that are doing the same job.”
-David Schirmacher, Senior Vice President of
PortfolioOperations Digital Realty Trust
RagingWire had put branch circuit monitor-
ing on every customer circuit. But there was
no single point where all information was
collected. If a device alarmed, admins would
have to log into that system to investigate.
“Having data in disparate systems does not
give you the ability to make decisions”says
Berti. “With that data in one place we can
use analytics such as measuring our PUE
across campuses.”
Using DCIM, the company is presenting
customers with dashboards to view data on
their own equipment such as power usage.
To date, Berti has purchased and installing
the hardware and software; created tem-
plates for sensors and back office systems,
normalized the data put it in a graphical
format. This included putting all the floor
maps, including customer enclosures, into
Baseline Comparison shows a comparison of a production server to 2 test servers.
The time filter is set to“any time” in order to include historical differences.
Check Consistency shows the changes made to a system and were checked against
the testing environment. Note that one DLL file was not updated.
July/August 2014 Data Center Management 27