4. AWS Regions and Availability Zones
Availability
Zone
Availability
Zone
Availability
Zone
Availability
Zone
Availability
Zone
Availability
Zone
Asia Pacific (Tokyo)
EU (Ireland)
US West (Oregon)
Availability
Zone
Availability
Zone
Availability
Zone
Availability
Zone
Availability
Zone
Availability
Zone
Availability
Zone
Availability
Zone
US East (N. Virginia)
Availability
Zone
Availability
Zone
Availability Zone
Availability
Zone
Availability
Zone
Availability
Zone
Availability
Zone
Availability
Zone
Availability
Zone
US West (N. Cal)
(Asia Pacific) Singapore
AWS GovCloud (US)
South America (Sao Paulo)
Customer decides where applications and data reside
Asia Pacific (Sydney)
5. AWS Ingest Options
AWS
Direct
Connect
AWS
Import/Export
AWS
Storage
Gateway
Dedicated
bandwidth
between
your
site
and
AWS
Physical
transfer
of
media
into
and
out
of
AWS
On-‐premises
storage
federa;on
with
Amazon
S3
and
Amazon
Glacier
6. AWS Ingest Options – One Common Theme:
1.
2.
3.
4.
Parallel Uploads
Multipart upload
Request rate optimization
TCP window scaling
TCP selective
acknowledgement
AWS has customers that ingest roughly 1 PB per day
7. AWS Ingest Options
AWS Direct Connect
•
•
Reduces costs for bandwidthheavy workloads
Private connectivity to AWS
– Physical connection – 1 Gbps or 10
Gbps port
– Logical connections (802.1q
VLANs)
§ Public: To AWS cloud (Amazon EC2,
Amazon S3 etc.)
§ Private: To VPCs
•
•
Consistent network performance
Compatible with all AWS services
8. AWS Ingest Options
AWS Direct Connect
Cost
•
•
•
•
1 Gbps port = $0.30/hour
10 Gbps port = $2.25/hour
Data transfer IN = $0
Data transfer OUT = $0.02 – 0.11 per GB
depending upon location
–
Can be a significant savings vs. Internet
bandwidth out
Locations
•
•
•
•
•
•
•
•
•
•
•
CoreSite 32 Avenue of the Americas, NY
CoreSite One Wilshire & 900 North Alameda, LA
Equinix DC1 – DC6 & DC10 - DC11, Ashburn, VA
Equinix SV1 & SV5, San Jose, CA
Equinix SE2 & SE3, Seattle, WA
Equinix SG2, Singapore
Equinix SY3, Sydney
Equinix TY2, Tokyo
Eircom, Clonshaugh
TelecityGroup Docklands, London
Terremark NAP do Brasil, Sao Paulo
9. AWS Ingest Options
AWS Import/Export
•
•
Rapidly move data into and out
of AWS
Portable storage device
shipment to AWS
–
–
–
•
Supports
–
–
–
•
eSATA
USB 2.0 and 3.0 (including USB flash
drives)
2.5 and 3.5 inch internal SATA hard drives
Amazon Elastic Block Store (EBS)
Amazon Simple Storage Service (S3)
Amazon Glacier
Use cases
–
–
–
Initial content migration
Content distribution via portable devices
Disaster recovery
10. AWS Ingest Options
AWS Import/Export
• Cost
– $80 per storage device handled
– $2.49 per data loading hour
– Standard pricing for
• Amazon S3
• Amazon EBS
• Amazon Glacier
11. AWS Ingest Options
AWS Storage Gateway
•
•
On-premises, virtual iSCSI
storage appliance
Local cache enables low
latency access to data
–
–
•
•
•
•
Gateway – stored volumes
Gateway – cached volumes
Copies data in the form of
Amazon EBS snapshots to
Amazon S3
Leverage Amazon S3 serverside encryption
Recent patch results in up to
5 TB of throughput per day
Recover to Amazon EBS /
Amazon EC2
12. AWS Ingest Options
AWS Storage Gateway
• Cost (N. Virginia – varies per region)
– Gateway pricing
•
$125 per activated gateway/mo.
– Volume pricing
•
$0.095 per GB per month of data stored
– Snapshot pricing
•
$0.095 per GB per month of data stored
– Tiered data transfer pricing
model
• Free inbound
• $0.12 - $0.05 per GB outbound
depending on tier
13. AWS Ingest Options
Gateway-Virtual Tape Library (Gateway VTL)
•
•
•
On-premises, virtual tape library
storage appliance
10 virtual tape drives / 1500
virtual tape slots
150 TB local cache
–
VTL – virtual tape library
–
VTS – virtual tape shelf
•
•
•
•
•
Restore in seconds from VTL
24 hour retrieval from VTS
Encryption in transit and at rest
Gateway VTL-AMI
In lab we achieved 55 MB/s
upload throughput and 90 MB/s
iscsi ingest rate per gateway
14. AWS Ingest Options
Gateway-Virtual Tape Library (Gateway VTL)
• Cost (N. Virginia – varies per region)
– Gateway pricing
•
$125 per activated gateway/mo.
– Virtual tape shelf storage
•
$0.01 per GB per month of data stored
– Virtual tape library storage
•
$0.095 per GB per month of data stored
– Retrieval from virtual tape shelf
• $0.30 per GB
– Virtual tape deletes
• Free
15. AWS Storage and Archive Options
Amazon
Elas@c
Block
Store
(EBS)
Amazon
Simple
Storage
Service
(S3)
Amazon
Glacier
High-performance block storage device
Highly
scalable
object
storage
Long-‐term
object
archive
1 GB to 1 TB in size
1
byte
to
5
TB
in
size
Extremely
low
cost
per
gigabyte
Mount as drives to instances with snapshot/
99.999999999%
durability
99.999999999%
durability
cloning functionalities
16. AWS Storage Options
Amazon Elastic Block Store (EBS)
• High I/O block storage for Amazon EC2
• Predictably scale to 1000s of IOPS per
Amazon EC2 instance
• Automatic replication within the Availability
Zone
• 10x more reliable than commodity disk drives
• Point-in-time snapshots
•
Amazon S3 durability (11-9s)
• Point-in-time snapshots across regions
• Amazon CloudWatch
•
Exposes Amazon EBS performance metrics
17. AWS Storage Options
Amazon Elastic Block Store (EBS)
Costs (US East)
Amazon EBS standard volumes
§ $0.10 per GB-month of provisioned storage
§ $0.10 per 1 million I/O requests
Amazon EBS provisioned IOPS volumes
§ $0.125 per GB-month of provisioned storage
§ $0.10 per provisioned IOPS-month
Amazon EBS snapshots to Amazon S3
§ $0.095 per GB-month of data stored
18. AWS Storage Options
Amazon Simple Storage Service (S3)
•
Synchronous in and synchronous out
object storage
•
Designed for 99.999999999% durability
•
Authentication mechanisms ensure data
is kept secure
•
Multiple encryption options
–
Amazon server-side encryption
•
Standard storage
•
Reduced redundancy storage (RRS)
20. AWS Storage Options
Amazon Simple Storage Service (S3)
Costs (US East)
Standard Storage
Reduced Redundancy
Storage
First 1 TB / month
$0.095 per GB
$0.076 per GB
Next 49 TB / month
$0.080 per GB
$0.064 per GB
Next 450 TB / month
$0.070 per GB
$0.056 per GB
Next 500 TB / month
$0.065 per GB
$0.052 per GB
Next 4000 TB / month $0.060 per GB
$0.048 per GB
Over 5000 TB / month $0.055 per GB
$0.037 per GB
21. AWS Archive Options
Amazon Glacier
•
•
$0.01 - GB per month
Retrievals:
–
5% of monthly average storage (pro-rated daily) free
•
•
•
•
•
Synchronous in
3–5 hour asynchronous retrieval
Designed for 99.999999999% durability
AES 256 encryption at rest
Highly scalable
•
•
Reliable
Authentication mechanisms ensure data is kept secure
22. AWS Archive Options
Object Lifecycle Management: Amazon S3 → Amazon Glacier
•
Seamlessly move data from Amazon S3 → Amazon Glacier
•
3-5 hour asynchronous retrieval
•
Data lifecycle policies
•
$0.01 per GB for Amazon Glacier costs
→
23. Partner Ingest Options
Aspera
CloudBeam
Signiant
Up
to
1
Gb/s
per
instance
to
AWS
SaaS-‐based
file
transfer
into
and
out
of
AWS
High-‐speed,
network-‐efficient
file
transfer
–
up
to
200X
faster
than
FTP
with
95+%
network
efficiency
24. Partner Ingest Options
Aspera On-Demand
•
•
•
•
•
•
•
Achieve file transfer speeds that are 1000s of times faster the FTP
In, out, and across the cloud with enterprise-grade security
End-to-end security
Speeds of up to 1 Gbps per AWS instance
10 TB per 24 hours
Scale-out architecture
Web, mobile, embedded clients
25. Partner Ingest Options
Attunity CloudBeam
• Simplifies, automates, and
accelerates the loading and
replication of files from onpremises, heterogeneous
sources to and from
Amazon S3
• Common Use Cases:
–
–
–
–
Content availability and distribution
Data analysis (Amazon EMR Hadoop)
Backup, disaster recovery, and archiving
Region-to-region replication
26. Partner Ingest Options
Signiant Media Shuttle
•
•
•
•
•
AWS-based fast file transfer
as a service
200X faster than FTP
Separates control layer from
the data layer
Multiple sources and targets
including Amazon S3
Firewall-friendly transfers
with autoselecting UDP,
TCP, and HTTP transport
options
NAS (CIFS, NFS)
DAS / SAN
27. Partner Ingest Options
Cycle Computing DataManager
•
•
•
•
Move data from any NAS / file
system to Amazon S3 and/or
Amazon Glacier
Clean up expensive, on-premises
disk
Maintain full access to all content
Reduce or eliminate future data
migrations upon hardware refresh
28. Partner Storage and Archive Solutions
Avere
Systems
Panzura
Record
performance,
scale-‐out,
single
file
system
NAS
Cloud-‐integrated
local
NAS
capabili;es
for
the
globally
distributed
enterprise
31. Partner Storage Options
Avere Systems – Comparing 1,000,000 IOPS Solutions
•
•
•
•
Add high-performance, scale-out clustering with any NAS
Automated tiering
Separates performance scaling from capacity
Avere offers the leading $ per IOPS for NAS
Avere
$2.3 / IOPS
– $2.3/IOPS
•
•
•
•
80% less total equipment than traditional NAS systems
Fastest scale-out, single file system (NAS) available
Linear scaling to millions of operations/sec
Tens of GB/sec of throughput
150ms
Cloud
Latency
32. Partner Storage Options
Avere Systems
•
Amazon S3 integration by EOY 2013
– 3-step process:
1. Leverage Avere to accelerate current NAS System
2. Nondisruptive migration to Amazon S3 / Amazon Glacier
–
3.
FlashMove
Switch mode in Avere to enable primary NAS operations
–
Retire older NAS gear
Client
Workstations
Core Filers
Avere FXT Edge Filer
Purpose-built for cloud
Enterprise-class scaling
Lowest TCO
Compute
Farm
Amazon
S3
Legacy NAS
Show as complex w/
RAID, volume limits, low
utilization, mirror
schedules, etc.
WAN
Amazon
Glacier
On Premise
AWS
33. Partner Storage Options
Panzura
•
Panzura enables:
• Global sharing – On-premises, hybrid, across AWS
regions
• Panzura Amazon Machine Image (AMI)
• Small physical footprint
• Separation of data and metadata
• Data protection
• NAS centralization
• Shift ratio of Opex vs. Capex
34. TCO: On-Premises Cost Considerations
1. Primary storage hardware (primary / remote site)
2. DR / Remote site storage hardware
3. Raw to utilized storage (both primary and DR)
4. Storage growth (cost of upgrades)
5. Storage management software and 3rd party tools
6. Professional services
7. Hardware maintenance
8. Software maintenance
9. Backup software
10. Backup hardware (primary / remote site)
11. Offsite tape storage / vault
12. Archive software
13. Archive hardware
14. Power
15. Cooling
16. Space
17. Labor
18. Cost of capital
19. Training
20. Asset depreciation
21. Migration
22. Decommission / remove
23. Recycle
38. iN DEMAND Intro
• World leader in providing transactional
entertainment delivered through television
• Joint Venture – owned by Comcast, Time
Warner, & Cox
• Pay-per-view programming – MLB, NBA, NHL,
boxing, MMA, & Howard Stern
39. The Problem
• 1.5 PB video archive
– World War Z
– Ice Pirates
– Titanic II … Tsunami AND an Iceberg
• Tape storage
–
–
–
–
Tape corruption and bit rot
Lost tapes
Physical storage – 1.5 PB is a lot of tapes
Legacy tape formats – LTO-1, 2, 3, 4, 5, etc. etc.
40. The Problem (cont.)
• Manual asset tracking
– Typical backup system stores file name, date, size
– Important metadata is tracked separately, e.g. bit rate, aspect
ratio, closed captioning, dual language, codec.
– Inventory issues –What bit rates do we have for Spider Man?
– Multiple storage – “Put it on tape just to be sure”
• Manual archive and restore
– Wait for operator to handle restores – not 24x7
41. The Problem (cont.)
• Expensive
– Tape operator
– Tape storage
– Yearly tape library maintenance
• Limited scale
– Limited by tape library capacity
– Limited by physical space
42. The Solution – Mini-DAM
• Limited digital asset management system for
Glacier
–
–
–
–
Web UI
Glacier storage
$50 K
Hosted at AWS – EC2, Amazon RDS, Amazon SNS,
Amazon SES
– Over 300 TB in Glacier to date
– Adding about 2 TB / day
43.
44.
45.
46. Tips & Tricks
• Concurrent downloader required
– Users want files FAST!!!
– .NET and JAVA AWS SDK have only a single-threaded downloader – MAX
download c.a. 160 Mbps
– iN DEMAND wrote a multithreaded downloader
– Added to AWS SDK for Python (BOTO) – MAX. download 600 Mbps
• Per archive Glacier overhead
–
–
–
–
Every Glacier archive has a 32 kb overhead for metadata
You are charged for this overhead
For small files that 32 kb starts to add up
Zip up small files before uploading
47. Tips & Tricks
• Download request time outs
– 24 hours to download archive
– Queue up requests to ensure that files are downloaded within
the 24-hour timeout
• Add the extra encryption to make management
happy
– The MPAA loves encryption
– Management loves encryption
– AWS automatically encrypts files at rest in Glacier
48. Tips & Tricks
• Checksum files before you upload
– Save MD5 checksum
– Check that file hasn’t already been uploaded to Glacier
– Avoid file duplication
• Track who requests downloads to manage costs
– Fee associated with each download
– Keep employees honest
49. Please give us your feedback on this
presentation
MED301
As a thank you, we will select prize
winners daily for completed surveys!
Thank You