Report on UGA Campus Data Backup and Recovery Needs
May 31, 2004
Chris Adcock, College of Agricultural and Environmental Sciences (Chair)
William Clayton, UGA Libraries
Wayne Crotts, College of Agricultural and Environmental Sciences
David Floyd, University Testing Services
Michael Hill, Franklin College
Eddie Hunter, EITS
Stuart Ivy, College of Veterinary Medicine
Martijn Middelplaats, University Health Services
Barry Robinson, College of Education
John Taylor, Terry College of Business
Larry Tucker, Computer Operations EITS
At the core of our business model and education delivery processes is the collection,
management, preservation, and redistribution of data in one form or another. Unlike
twenty years ago all of our data is not cloistered in one physical location under the
watchful oversight of a specialized team of systems analysts. Instead, a multitude of data
centers have emerged within the confines of individual colleges, departments, and service
units. The aggregate data storage capacity continues to expand in these diverse
computing environments. There are at least 90 such entities on the Athens campus and
each one must manage the storage and preservation of institutional data.
Managing the storage and preservation of data is not always highly esteemed within the
overall context of the various information analysis and delivery processes that comprise
our research, education, and outreach business models. However, it is extremely
important to the short term and long term continuance of these processes and the
effectiveness of our organization. Data backup and recovery is a necessary yet costly and
time consuming function.
Challenges abound. Conservatively, data center managers estimate that the storage
capacity growth rate is 18% annually. Increasingly stringent security measures and
regulatory requirements are being imposed requiring greater compliance diligence and
increased resources. Clientele expectation for uninterrupted access to information is on
the rise. Budget cuts have limited the ability to properly staff data centers and provide
adequate equipment and software tools to meet new access demands and regulatory
mandates. The need for a better understanding of these issues and research into possible
solutions and efficiencies stimulated the call for the formation of an exploratory
Committee Formation and Mission
The concept of forming a committee to explore campus wide backup and recovery
practices was first discussed at the August 2003 meeting of the Information Technology
Management Forum (ITMF). The spectrum of discussion regarding committee mission
and scope ranged from a full exploration of business continuity and disaster recovery
needs to elements of centralized backup and recovery economies of scale. It was
determined that investigating current backup and recovery practices and possible
economies of scale would be the logical first step.
ITMF strongly advised that the committee be small in number and include a diverse
membership representative of both small and large units across campus. It was
recommended that members be included from Enterprise Information Technology
Services (EITS) representing the functional groups of Security and Operations. In
addition, a representative from the Libraries was recommended because of their recent
experience in recovering from significant fire damage at their site. In September a call
for membership was issued and the committee was formed under the name Centralized
Backup and Recovery Exploratory Committee.
To seek to understand and quantify existing backup and recovery models currently in
practice on Athens Campus and to explore and recommend options for improving and
enhancing existing procedures with the goal of improving business continuity and finding
economies of scale.
• Develop a means of discovering existing backup / recovery practices, capacities, and
• Analyze current industry standard practices and procedures for centralized backup
• Research current industry standard technologies / infrastructure for supporting large
scale backup and recovery
• Document industry or institutional success stories / horror stories relevant to backup /
recovery / recovery and business continuity
• Document UGA campus communications infrastructure design and capacities
• Explore the possibilities of both internal and external collaboration
• Interview and analyze private offsite data management businesses
• Identify individuals / organizations who can provide us with best practices for
managing data security and privacy
• Study certain EITS large scale IT rollout models and ‘Due Diligence’ procedures
Phase One: Survey creation
Content review and revision by ITMF
Content review and revision by self application
Partner with Student Services for web delivery of survey
Partner with Paul Keck and ITMF for distribution list
Collection and analysis of survey responses
Phase Two: Consultation with outside data management vendors
Vendor: Data Survival International (DSI)
Vendor: Network Appliance
Phase Three: Consultation with internal resource personnel
Paul Keck: UGA Campus Network Management
Eddie Hunter: UGA Campus Data Communications Infrastructure
Jerry Nesmith: Research Services Computing Center
Phase Four: External site visit
Medical University of South Carolina Data Center
Phase Five: Compilation and presentation of findings and recommendations
Backup and Recovery Practices Survey Analysis
The Campus Backup and Recovery Practices Survey was conducted on the Athens
campus during the month of December, 2003. Invitation to participate was
communicated via ITMF and Domain Network Liaison distribution lists. Ninety one
respondents participated in this survey.
The intent of the survey was to gather information on current data backup and recovery
needs and practices. Individual responses were compiled into aggregate views with
accompanying analysis as listed below. Individual department / unit results will not be
published in order to maintain confidentiality. See appendix A for a listing of the actual
When considering the following information be aware that although 91 individuals
responded to this survey not every respondent answered all of the questions.
Centralized Backup and Recovery Services Interest
The survey question asked, “If a non mandatory, secure, reliable, centralized backup and
recovery service could be put in place what elements of the service would you find
useful? Check all that apply.” Forty-six out of 91 respondents said they would be
interested in a primary backup service. Sixty-six percent recognize the need for off-site
storage. Thirty percent desire all three services. Only 16% replied that they had no
interest in any of the listed services.
Central Services Responses %
1. Off-Site Storage Facility 60 66%
2. Primary Backup Service 46 51%
3. Secondary Backup Services 57 63%
Interest in 1,2,&3 27 30%
Not Interested in Any Services 15 16%
Total Respondents 91
Server Platform Breakdown
From both a campus view and individual unit view, the survey data shows that our server
platform environment is quite heterogeneous. Of the 90 respondents to this question 39%
maintain Netware Servers, 57% maintain Microsoft Servers, 43% maintain UNIX / Linux
servers and 17% Other platforms.
Platform Netware Microsoft Unix Other
Response Response Response Response %
Quantities s % s % s % s
0 55 61% 39 43% 52 58% 75 83%
1 20 22% 15 17% 12 13% 11 12%
2 to 5 13 14% 28 31% 16 18% 3 3%
6 to 10 2 2% 6 7% 5 6% 0 0%
10 or more 0 0% 2 2% 5 6% 1 1%
Total 90 90 90 90
Data Storage Capacities
As shown in the chart below, respondents report available data storage capacity at 59.92
terabytes. Of this capacity 23.7 terabytes of data storage space was identified as being in-
use. It appears that roughly 39% of all hard drive storage capacity is currently being used
for active file storage needs. However, there is no guarantee that all of the additional
capacity can be classified as available for expansion. In practice, additional hard drive
capacity is used to facilitate systems maintenance, testing, and recovery needs. For
example, unused capacity may routinely be used to provide temporary space for data
manipulation, temporary buffer space for certain scheduled processes, hot backups of
selected data sets, or test bed space.
Data Windows Unix Novell Other Totals
Storage (TB) (TB) (TB) (TB) (TB)
Capacity 14.80 32.67 5.47 6.98 59.92
In –Use 4.78 12.74 2.05 3.33 23.70
Percentage 33% 38% 49% 48% 39%
Overall, survey respondent’s project data storage needs to grow at an annual rate of 18%.
Using the 2004 in-use data totals, 18% growth for 2005 would require a storage capacity
of 27.97 terabytes; a growth of 4.27 terabytes.
Full Backup Frequency
The responses to this question coupled with the information gathered about data storage and
capacities provide some insight into the expected bandwidth load that our data communications
infrastructure would need to accommodate. The survey question asked, “Which option best
describes how often you perform Full Backups?” The choices ranged from ‘daily’ to ‘never’.
Respondents were asked to reply per platform supported. The data shows that there was no
significant variation of full backup behavior based on platform type. The only notable exception
was that of the UNIX platform responses which indicate a 2-5 time greater frequency for
‘monthly’ full backups than was reported for the other listed platforms. It was also notable, and
somewhat surprising to discover that approximately 11% of the responses indicate that full
backups ‘never’ occur. As shown in the table below, independent of platform, the most popular
backup frequencies were daily (37%) and weekly (34%).
Backup Frequency Responses %
Daily 52 37%
Weekly 47 34%
Every Two Weeks 7 5%
Monthly 9 6%
Every Two Months 5 4%
Every 6 Month 3 2%
Yearly 1 .7%
Never 15 11%
Total Responses 139
Should a campus based centralized backup and recovery service be established one could
speculate that between 37% and 100% of data center administrators could initiate full
backups on a daily basis. Given the data in-use capacities listed in the previous section,
approximately 24 terabytes of data could traverse the data communications infrastructure
during the same time period. In practice this would probably not be the case as it is likely
that a reliable backup and recovery model would be based on a routine of no less than a
weekly full backup coupled with daily differential or incremental backups. In any case,
should a centralized backup and recovery service be established there would inevitably be
new bandwidth loads introduced to our data communications infrastructure.
Retention of Backups
It appears that the majority of responding units retain full backups between one week and
six months. Only 14% retain full backups for 1 year or more and 9% are not retaining
backup copies at all.
Retention Responses %
1 week 26 19%
2 weeks 20 14%
1 month 23 17%
2 months 20 14%
6 months 16 12%
1 year 8 6%
1 year + 12 9%
Never 13 9%
Media Storage Location
The respondents were able to select all that apply for the question of where backup tapes
or other backup media is stored. Thirty one units responded that they are storing data off
site external to the campus. Seventy percent of the respondents indicate that they do not
store their backup media in a fireproof safe. The survey did not provide a sufficient
mechanism for identifying the level of protection against unauthorized access to media
stored at any location.
Location Responses Percentage
ONSITE 69 76%
OFFSITE/ON Campus 18 20%
OFFSITE/OFF Campus 31 34%
Total Respondents 91
Occurrence of Full Recovery from Tape
Survey data shows that approximately 52% of units have performed full recovery from
backup media over a one year period of time.
Full Recovery Occurrence Responses Percentage
Once Weekly 1 1%
Once Monthly 1 1%
Once 3-6 Months 7 8%
Once Yearly 38 42%
Never 44 48%
Total Respondents 91
Tape Backup Equipment, Software, & Media
Not all units use production grade tape backup and recovery equipment. For example,
smaller shops may simply dump to CD, DVD, or an out of production PC or server hard
drive. Totaling the reported campus wide unit costs for maintaining backup and recovery
equipment, software, and media calculates to $347, 882. Forty units on campus report
using Veritas Backup Exec as their backup and recovery utility. Should a reliable and
secure central backup facility be established and maintained a standard media type and
backup software could be selected. It is expected that bulk purchasing and site licensing
savings could be significant.
Top Four Equipment Brands Responses Percentage
Dell Powervault 17 20%
Seagate 5 6%
Sun 5 6%
Sony SCSI sequential 3 3%
Total Respondents 87
Top Six Backup / Recovery Utilities Responses Percentage
Veritas Backup Exec (all versions) 40 45%
Tar 8 9%
Retrospect 6 7%
Tivoli 2 2%
NT Backup 4 5%
Tapeware by Yosemite 1 1%
Total Respondents 88
Top Six Media Types Responses Percentage
DLT 32 37%
DDS 12 14%
LTO 8 9%
CD –R 5 6%
LVD 3 3%
DVD-R 1 1%
Total Respondents 87
Respondents to this particular question were able to check all that apply. The results
listed below indicate how wide spread the storage of personal information has become on
University of Georgia servers. Forty-three percent of respondents indicated that one or
all of the following data types are stored at their location: Social Security numbers, credit
card information, and student records. Regulatory compliance information applicable to
the storage and transport of personal, medical, and financial information was not
Stored Data Types Responses Percentage
Social Sec. No. 34 37%
HIPAA Data 3 3%
Credit Card No. 7 8%
FERPA Data 11 12%
Student Records 29 32%
Human / Animal Rsrch. 7 8%
Total Respondents 91
The majority of units responding to the survey indicate that they are connected to the
high speed Venus data communications network.
Venus Connectivity Status s Percentage
Off Venus 13 14%
On Venus 76 84%
Venus no response 2 2%
Time and Effort
The survey indicates that there is a wide variety of backup and recovery models in
practice on our campus. Equipment, software, data capacities, and operating procedures
effect both the time required to perform full backups and the personnel time investment
to administer the processes. Assuming that only 91 data centers existed on campus and
that the reporting of time investment was accurate it can be estimated that the aggregate
time dedicated to managing backup and recovery processes is 227.5 hour per week.
Human Resource Element Hrs.
Average hours to perform full backups 8.65
Average hours per week dedicated to managing 2.5
The challenges faced when considering a centralized backup solution include identifying
a solution that is:
• OS agnostic and preferably using open standard protocols. Departments/units on the
UGA campus employ a variety of operating system/networks including Microsoft
Windows, Novell, UNIX, Apple, etc and a number of vendors. The centralized
backup solution would need to meet the backup needs of these departments/units
regardless of OS and vendor used;
• Deployed within the UGA campus network infrastructure. (See the Campus
Infrastructure Overview section for more information);
• Scalable. It is anticipated that a centralized backup solution may be implemented in
phases or even as a collaborative effort between multiple colleges/units/departments.
The solution would need to be designed for not only data expansion, but also
allowance for additional parties to join at later dates;
• Manageable. A monitoring/management system would need to be in place so that:
o Successful backups of data can be verified by the college/unit/department;
o Successful recovery of data can be initiated and verified by the
o Ability for the college/unit/department to initiate new backup plans and/or edit
• Secure. The backup solution would need to meet established security criteria for
handling the transmission and archiving of the stored data.
• Cost Effective: A centralized backup solution should offer identifiable cost benefits
relative to current backup and recovery practices. Human resources, hardware,
infrastructure and software costs must be part of the overall TCO calculation.
The committee looked at the feasibility of current industry standard technologies for
supporting large scale backup and recovery. These included:
• SAN (Storage Area Network) Technology
• NAS (Network Attached Storage) Technology
• Vendor Proprietary Solutions (which may include a combination of the above
Although individual implementations differ, the most common SAN deployments consist
of storage cabinets with large numbers of disks, fiber-switches interconnecting these
cabinets, and interface cards (host adapters) connecting servers to the fiber-switches. This
hierarchical, scalable model makes SAN technology ideally suited for environments that
are expected to grow substantially over time. Redundancy in the links between servers
and switches as well as between individual switches and their attached storage cabinets
can provide extensive fault-tolerance, with split-second fail-over capability.
Typically, disks will be used to create (RAID) arrays, which can then be configured into
segments (aka Logical Unit Numbers) that are individually assigned/associated with a
specific server system. Arrays (and thus, LUNs) can be expanded upon dynamically, so
there is virtually no limit to how much storage space can be allocated for any attached
Management software usually offers extensive capabilities to create Point-in-Time copies
of LUNs, real-time mirroring (to in-SAN, local or off-site storage facility) and the ability
to create off-line backups to tape or disk.
Advantages of SAN Technology
• Largely OS agnostic
• Extremely scalable
• Lower TCO than DAS (Direct Attached Storage) for large deployments
• Centralized management
• Can be used for primary or secondary (backup) storage needs
Limitations of SAN Technology
• High initial costs
• Requires fiber for intra-campus or off-site connectivity (Fiber SAN Only)
• Management (of LUNs etc) can be challenging
iSCSI SAN Technology
This emerging technology eliminates the requirement for fiber optic interconnectivity.
The term iSCSI refers to a process in which traditional SCSI commands, which are not
natively formatted in IP, are wrapped in IP packets and transported across Ethernet
networks. Since it uses SCSI block-based architecture, it allows efficient use of the
bandwidth of the Ethernet. This allows an IP-based data storage device to be accessed by
servers across an existing Gigabit Ethernet network. While iSCSI technology has been
around since before 2001, the Internet Engineering Task Force only recently approved
the iSCSI standard in 2003. In the last few months, iSCSI solutions have begun to emerge
from several storage vendors. Additionally, Microsoft, Novell, IBM, and others have
launched initiatives toward providing iSCSI drivers and support options to customers.
Advantages of iSCSI Technology
• Can be used over existing UGA Gigabit network
• Ease of integration in the open standards environment
Limitations of iSCSI Technology
• Bandwidth concerns. While efficient, the data is transmitted over the gigabit Ethernet
backbone. Bandwidth use as well as quality of service monitoring and packet shaping
may need to be in place to accommodate iSCSI traffic across the present UGA
network infrastructure. Also, currently iSCSI is limited to 1 gigabit Ethernet
capacity, though most experts believe this will increase to 10 gigabit capacity once 10
gigabit networks reach a more mature state in the next five years.
• Processing overhead. The CPU will have to process the TCP/IP packet header in
addition to the block data. The solution is a TCP Off-load Engine (TOE) or an iSCSI
Host Bus Adapter (HBA). These specialized cards perform all the TCP/IP overhead
processing with on-board processors instead of passing it to the CPU the way
standard NIC cards do. TOE cards and iSCSI HBAs each have advantages and
drawbacks. Whereas iSCSI HBAs deal only with iSCSI traffic, TOE cards can handle
all standard Ethernet traffic. However, iSCSI HBAs use fewer main CPUs than do
TOE cards for iSCSI operations, and they generally cost a few hundred dollars less
• Proprietary Management System: While iSCSI itself is an open standard, there are
concerns by industry experts that effective management systems may not be.
• Security: iSCSI was not designed to be a secured data transmission protocol. IP-
SEC/SSL or some other method of encryption will need to be deployed for iSCSI
traffic thereby further increasing the processing overhead.
• Software/license/hardware costs: In iSCSI methods reviewed, the cost for backup
software licensing was significant as licensing cost is based ‘per CPU’ of the server.
Also, hardware costs should be considered as TOE or HBA network cards are more
costly than general 100/1000 gigabit network cards. Most experts feel iSCSI
solutions will still be considerably less than fibre-optic solutions, but total cost of
ownership should be looked at very closely with any specific iSCSI solution.
Network Attached Storage:
NAS (Network Attached Storage) is defined simply as shared storage that is attached
to a LAN. Since it uses a familiar technology NFS (Network File System) and CIFS
(Common Internet File System), to transmit data, NAS resources can be managed by
existing IT staff with minimal training in storage management which may reduce IT
costs. Another benefit of NAS is flexibility -- since the storage unit can easily be
attached to the network and can support file services of different OS’s and networks. The
weaknesses of the NAS solution include not being a highly scalable solution and
requiring high bandwidth use thereby negatively affecting LAN performance. Due to
these weakness, it could be a very limited method to use in a centralized backup
initiative. However, vendor proprietary products offer NAS solutions which allow ‘block
to block’ data transfer as well as ‘snapshot’ technology. After a full backup of a system
has been completed, a number of subsequent full backups can be completed where only
‘disk blocks’ that have changed are transmitted over the wire. This radically improves
backup and recovery performance. Often these vendor products are available for fiber
and iSCSI SAN as well.
Pro’s/Advantages of NAS Technology
• Largely OS Agnostic. The storage unit can accommodate any system that uses NFS
• A NAS device is relatively easy to manage. Setup and implementation of a NAS
device is typically very easy.
• Secured. Traffic to and from a NAS device can be encrypted using SSL or IPSEC
• Cost Efficient: NAS devices are relatively inexpensive.
Cons/Limitations of NAS
• Cannot be practically implemented across the existing UGA Network infrastructure
as a centralized backup solution due to its high bandwidth consumption.
• NAS solutions that are utilizing ‘file to file’ data transfer formats (and not ‘block to
block”) are not very scalable. As a means to archive and be able to quickly recover
significant amounts of data, data transfer is significantly slower than the other
Vendor Proprietary Solutions
A number of vendors exist that provide proprietary data backup and recovery solutions.
Typically, this involves installing the vendor’s software client on the server/workstation
that requires data backup. The backup occurs over the existing Ethernet network to the
vendor’s storage site. Some vendors offer bandwidth friendly solutions as well as
encrypted means for transmitting data to the vendor’s storage facility.
Pros/Advantages of Vendor Proprietary Solutions
• Typically OS agnostic
• Typically can operate across existing network topologies including UGA campus
network infrastructure. (See the UGA campus network section for more
• Solutions exist that are scalable allowing both for future growth of existing users and
the provisioning for additional ones.
• Solutions exist that incorporate good management protocols and capabilities for both
building backup and recovery plans.
• Typically solutions include encryption methods for securely transporting data to and
from the archival site as well as physical site security protection measures.
Limitations/Challenges of Vendor Proprietary Solutions
• May have a higher cost of ownership
• May reduce interoperability with other technologies
In addition, the committee met with Jerry Nesmith, Director of Research Services, who is
coordinating the Research Computing Center project. This center, when online, will
provide IT consulting and resource services to the research community at UGA,
including housing a multi Terabyte data facility. Many of the network infrastructure and
storage challenges that RCC will need to address are identical to those of a centralized
backup and recovery solution.
Campus Infrastructure Overview
The University of Georgia network consists of three major components – core,
distribution and access layers. UGA’s core layer consists of 16 locations containing
Foundry routers interconnected via partially meshed fiber links. The distribution layer is
composed of a router or layer-two Ethernet switch location in the main wiring closet
(called the main distribution frame or MDF) in a building that connects to the core via
fiber and to building networks via fiber or Cat 5E twisted-pair copper cabling.
Networked devices in buildings typically connect to layer-two Ethernet switches in
intermediate wiring closets (called intermediate distribution frames or IDFs) via copper
cabling, and the switches connect back to the MDF via fiber or copper. This collection of
switches and cabling constitute the access layer.
The core and distribution layers have a fixed amount of bandwidth capacity, and the
access layer has varying capacity depending on the cabling that is used and more
importantly what network switches are used in IDFs. At the present time the theoretical
bandwidth capacity between core locations is two Gigabits per second (Gbps). However,
with network overhead, a conservative estimate of the actual network capacity is 80% of
the theoretical amount. This means that the inter-core capacity is approximately 1.6
Gbps. For most buildings on campus, the theoretical bandwidth capacity at the
distribution layer is 1.0 Gbps with an actual capacity of 0.8 Gbps (or 800 Megabits per
second [Mbps]). Networked devices in buildings typically connect at 10 or 100 Mbps (8
or 80 Mbps actual bandwidth), and switches in IDFs connect to the main building wiring
closet at 1.0 Gbps via fiber or 100 Mbps via fiber or copper (800 Mbps or 80 Mbps of
actual bandwidth, respectively).
It is important to realize that networked devices share the bandwidth capacity from their
IDF to the building MDF. All of the IDFs in a building share the capacity of the
distribution layer connection. Multiple distribution layer connections in turn share the
bandwidth capacity of the core layer network. Utilization data for inter-core and building
distribution connections indicate that the average amount of bandwidth on the most active
links is 18 Mbps and 13 Mbps, respectively (with peaks of approximately 30 Mbps).
EITS does not collect utilization data from access layer switches, so no data is available
for that layer.
Although the current core and distribution bandwidth utilization is not near its capacity,
future application needs such as research computing and centralized backup services
could potentially exhaust that capacity. EITS is in the process of planning UGA’s next-
generation core network which may consist of a fully-meshed 10 Gbps, four-core
network with redundant Gigabit links from buildings. Because of anticipated bandwidth
requirements for research computing, EITS, in a partnership with the Office of the Vice
President for Research, is contemplating the development of a separate physical network
to serve the needs of researchers.
Survey findings make it clear that there is a high level of interest in enhancing and
expanding backup and recovery capabilities on our campus. Eighty-four percent of the
respondents indicated that they would utilize elements of a centralized backup service.
Various levels of backup and recovery services are being performed at scores of locations
throughout the Athens Campus. Equipment, software, media, local infrastructure, and
personnel must be in place to perform this necessary function. Thus, ongoing backup and
recovery costs are a significant portion of IT expenditures. Survey respondents reported
that they cumulatively spend approximately $348,000 annually on equipment, software,
and media. This does not include human resource costs. Total annual expense for the
campus may be considerably higher.
There are no official UGA institutional standards for administration of backup and
recovery services. Data security practices, data management methodologies, retention
and recovery practices vary from unit to unit.
Forty-six percent of our respondents do not include off-site data storage as part of their
data backup management practice. If no other means of data redundancy exist then these
sites would stand a much higher probability of complete data loss should a localized
catastrophic event occur. To mitigate this risk there is unquestionably a need to provide
institutional off site storage.
Technology is available to address the need of centralized backup and recovery services,
(See Technology Overview). A combination of NAS, iSCSI and Fiber SAN technologies
can be put in place to meet UGA centralized backup needs. At present Fibre Channel 1-
Gbps and 2-Gpbs provide a more robust technology but require a fiber optic network
infrastructure. iSCSI SAN technology has 1-Gbps limits, however its ability to use IP
over a standard Gigabit Ethernet network infrastructure makes it a very attractive
emerging technology because it eliminates the need for a separate fiber network.
Likewise, NAS devices that offer ‘block to block’ and ‘snapshot’ features could offer
definite advantages within the overall centralized backup and recovery model.
There are advantages to establishing a centralized backup and recovery service. These
include the opportunity for: overall reduction in operating costs; better positioning to
meet regulatory compliance; improved service levels; improved and more consistent
performance; standardization; enhanced monitoring and analysis; greater security control;
and positioning to link to true offsite facilities. Should a centralized solution be
implemented, startup and maintenance costs will vary based upon expectation levels, end
point technology selection and transport technology selection. These decisions will be
based on business need requirements and influenced by several factors such as
institutional security and retention standards, clientele demand, and budget
As part of an overall institutional business continuity-disaster recovery plan, we
recommend that the University initiate a full cost benefit analysis of a centrally managed
backup and recovery implementation. The study should include: the development of
institutional guidelines and acceptable practices, the impact of consolidating existing
parallel backup and recovery processes, human resource investment / reallocation,
training, and risk mitigation.
Investigation should be made into whether using the existing Athens campus network
infrastructure for a centralized backup and recovery medium is feasible and scalable.
Should it be necessary to enhance the existing infrastructure the following options could
Dense Wave Division Multiplexing. DWDM would allow 8 times the current
bandwidth over the existing network fiber. This additional network throughput
could eliminate the need for additional fiber cabling to handle a central backup
Incorporation of unused ‘dark’ fiber cable. There is unused (dark) fiber already in
place on the Athens campus. However, it is unclear how much is available and its
criteria for use.
The committee is aware that the Terry College of Business (TCB) has entered into an
agreement with the Research Computing Center (RCC) to use a portion of the data
storage capacity of the research facility. RCC is purchasing 10 TB of high performance
RAID-DP protected storage from Network Appliance as well as 16TB of lower
performance serial ATA, RAID-DP storage. The latter will primarily be used for backup
capacity. TCB is purchasing an additional 8TB of lower performance media to add on to
the RCC configuration. Testing will be performed by Terry College of Business to
measure the effectiveness and efficiency of using a centralized backup model for a
portion of their operations. It is recommended that these findings be reported to the ITMF
committee when appropriate.
According to the survey 45 percent of respondents use Veritas backup software to help
manage data backup and recovery. Licensing agreements are negotiated and purchased
on a unit by unit basis. A centralized approach to Veritas licensing should yield an
overall savings as compared to our existing autonomous practice. It is recommended that
MSD explore campus site licensing options for backup software. Independent of a
centralized backup solution a campus wide licensing agreement with Veritas could
reduce overall costs and offer greater flexibility previously unaffordable at the unit level.
Based on the range of backup practices, sensitive data categories, and equipment
identified by the survey, we recommend that the ITMF explore with University Training
and Development and EITS the development of a workshop or training class to promote
recommended backup and retention standards for the campus.
Campus Backup and Recovery Practices Survey
1. Please provide use with your contact information.
2. Is your site connected via UGA campus’ VENUS network?
3a. How many Novel Netware based servers do you maintain at your site?
3b. How many Microsoft based servers do you maintain at your site?
3c. How many Linux / UNIX based servers do you maintain at your site?
3d. How many ‘other platforms’ based servers do you maintain at your site?
4. What is the total hard drive storage capacity, measured in Gigabytes?
5. How many Gigabytes of storage are required to perform a full backup of all data?
6. Which option best describes how often you perform full backups?
7. Which option best describes how long you retain your full backups?
8. On average, how often does a situation occur that requires a full recovery of any
system you maintain?
9. Describe your backup equipment. Please include functional description, make, model,
media type, interface, throughput (in GB/min), and total raw capacity.
10. How much time in hours, is required to perform a full backup of all data at your site?
11. What day of the week and time of day do you normally initiate full backups?
12. Are there certain times of the day when data must be captured? Why and what times?
13. Do any of your backups include the following information types? Check all that
14. Where do you store your tapes/media?
15. How much time per week, in hours, does staff at your site spend managing backup
processes and media management?
16. What backup software do you use at your site and how much does it cost annually?
17. Over a three year period of time estimate that annual cost of 1) equipment, and 2)
media, (but exclude backup software) associated with providing local backup and
recovery services at your site. Please itemize these if possible.
18. Does your site store backup tapes in a fire proof safe?
19. Give your best estimate of how much additional storage capacity will be required for
next fiscal year.
20. Do you have any special backup and recovery needs beyond what has already been
described above? If so please identify those needs in the space below.
21. If a non mandatory, secure, reliable, centralized backup and recovery service could be
put in place what elements of the service would you find useful? Check all that apply.
References / Contributors
The committee wishes to express our great appreciation to all those who invested their
time, ideas, and effort into this project. Their participation has contributed to a better
understanding of existing campus backup and recovery operations, needs, and potential
Bobby Lathrop, Ph. D.
Student Life Studies, University of Georgia
Campus Resource: Provided consultation and hosting services for campus wide survey
Information Analyst II
Student Life Studies, University of Georgia
Campus Resource: Provided survey hosting support
EITS, University of Georgia
Provided overview of campus networking and futures
Aided with committee communications
Office of Research Services, University of Georgia
Provided knowledge transfer regarding grid computing /data center project
Executive Director – Admin
EITS, University of Georgia
Campus Resource: Provided consultation and EITS resources
Major Accounts Manager
Arranged site visit to the Medical University of South Carolina
Provided knowledge transfer regarding Dells data center direction
Enterprise Systems Consultant
Provided knowledge transfer regarding Dell / EMC Storage Area Network technology
David L. Mills
VP and Director of IT
Data Survival, Inc.
Provided knowledge transfer on offsite data management provision
Account Executive, (GA & TN)
Network Appliance, Inc.
Provided knowledge transfer regarding Net. Apps. data center direction
Architectural Specialist, Consulting Systems Engineer
Network Appliance, Inc.
Provided knowledge transfer regarding data storage technology solutions
Data Storage Manager
Hosted sub-committee visit to explore the Medical University of South Carolina’s data