Proposed storage Area Network Facilities

Glasgow Storage Area
Network (SAN) Update
Where We Were (April ’05)
Why Consider a SAN
Heterogeneous environment
Protects against SPF
Storage aggregation into virtual pools
Less complex management
Lower admin costs
More suited to DR
Potential for server consolidation
Significantly improves backup restore capabilities
Improves data security, accessibility and availability
Proposed SAN Model
A core fabric spanning 2 sites
A number of FC edge switches linked to the FC core
A number of disk arrays providing the storage requirements for:
Research projects
Mirroring
Core services
Other applications
A number of servers connected via FC Host Bus Adaptors (HBAs)
An enterprise class backup facility
NAS heads for file sharing functionality
iSCSI support for fabric extension to non FC attached clients
FC/IP support for native fabric extension
Proposed SAN Model
What Factors Need to Be Addressed
When Considering a SAN Solution?
Fabric costs
Fabric Design
Switch costs
HBA costs
Fibre provision
Manufacturers limits
Vendor and OS
support
Fabric extension
NAS heads
iSCSI and FC/IP
SAN management
Disk grouping
Security
Performance
monitoring
What Factors Need to Be Addressed
When Considering a SAN Solution?
Disk virtualisation
requirements
Virtualising file systems
Backup Regimes
Type – Snapshot, Full
and Incremental
Extent – wont be able to
backup all SAN storage
Media – Tape, staging
disk then tape
Costs – hardware and
media
Fabric Connection
Guidelines
Not an open SAN
Restrict to manageable
set of proven hardware
compatible systems
Funding
Initial SRIF
Requires significant
investment over time
Who pays?
What Happened
EU procurement process for SAN commenced with publication of official notice
in OJEC
Expression of interest from interested suppliers required by mid April ‘’04
Operational requirement and invitation to tender documentation produced and
reviewed by small working group
20 expressions of interest received
Initial shortlist drawn up and operational requirement/invitation to tender sent
on 11th may ‘ 04 with closing date for responses 10th June ‘’04
Responses received from usual suspects
Suppliers meetings held and shortlist refined
BAFO sent to short listed suppliers on 18th august ‘’04 with responses by 10th
sep ‘’04
Final shortlist of three suppliers agreed and detailed negotiations commenced
Decision on best overall solution and actual spend made by end of Nov ‘’04
Kit delivered between 23rd December and 10th January
Procurement process
OJEC Notice issued March ‘ 04
Expressions of Interest (20) by mid April; Initial shortlist agreed
(13 suppliers)
Produce and review OR/ITT and issue to short listed suppliers
11th may ‘04; responses by 11th June ‘04
Review responses arrange suppliers meetings, refine shortlist and
potential solutions (DELL, EMC, Hitachi, SUN, HP, SGI, IBM)
Issue BAFO on 18th Aug ‘’04 with responses by 10th Sep ’04
Review BAFO responses decide final shortlist and commence
detailed negotiations with Access (SUN), and Computacenter/OCF
(IBM))
Choose solution and place orders end of Nov ‘’04
Comptacenter (IBM)
Kit delivered between 23rd Dec and 10th Jan
What Did We Get
A lot of work – electrical and other
provisioning work
A lot of boxes and a lot of heavy boxes
Logistical nightmare – getting the right
boxes to the right location (saved by
Nano Fab refurb)
Some pictures
And the end result
Where to Now – Initial Allocations
Initial commitments are 8TB grid, 4TB Hatti,
3.5TB student support
This leaves 9TB approx for other purposes
Initial proposal is to allocate 1TB approx per
faculty but;
This is a file level allocation and will require
fileservers and decisions on the delivery
mechanism e.g NFS, Netware, SMB, CIFS
A suitable client/server model
Where to Now – Future Provision
Need to have an attractive costing model for future
buy in
Complicated by (DR) replication model and backup
requirements
Initial expansion will be costly requiring additional FC
switches, storage arrays and File servers
Thereafter additional capacity would be disks/expansion
units and servers
File level service requiring server consolidation to provision
Aiming for as low a cost per Megabyte as possible
Would not attempt to recover capital costs
Ask University to fund infrastructure with Fac/Dept funding
their own additional storage services
Is SAN Storage Really Cheaper Than
DAS
Difficult to estimate but studies highlight the
following:
DAS typically reaches 10-50% utilisation. SAN
typically 80-90%
Highest cost is Sys Admin some say by a factor of
7/1
Some studies indicate the watershed between DAS
and SAN is reached at around 12 TB
Need to include all costs not just headline costs
Is SAN Storage Really Cheaper Than
DAS
References:
http://www.dmreview.com/article_sub.cfm
?articleId=1000938
http://www.its.monash.edu.au/staff/syste
ms/dsm/technical/cost-recovery/costcomparison-das-san.html
http://www.infoconomy.com/pages/storage
-finance/group95520.adp