above the clouds: a berkeley view of cloud computinga berkeley view of cloud computing ......
TRANSCRIPT
![Page 1: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/1.jpg)
UC Berkeley
1
Above the Clouds:A Berkeley View of Cloud Computing
Armando Fox and a cast of tens, UC Berkeley Reliable Adaptive Distributed Systems Lab
USENIX LISA 2009
© 2009
Image: John Curley http://www.flickr.com/photos/jay_que/1834540/
![Page 2: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/2.jpg)
2
Datacenter is new “server”
• “Program” == Web search, email, map/GIS, … • “Computer” == 1000ʼs computers, storage, network • Warehouse-sized facilities and workloads • New datacenter ideas (2007-2008): truck container (Sun),
floating (Google), In Tents Computing (Microsoft) • How to enable innovation in new services without first
building & capitalizing a large company?
photos: Sun Microsystems & datacenterknowledge.com
![Page 3: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/3.jpg)
RAD Lab 5-year Mission
Goal: Enable 1 person to develop, deploy, operate next -generation Internet application
• Key enabling technology: Statistical machine learning – management, scaling, anomaly detection, performance prediction...
• interdisciplinary: 7 faculty, ~30 PhDʼs, ~6 ugrads, ~1 sysadm
• Regular engagement with industrial affiliates keeps us from smoking our own dope too often
3
![Page 4: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/4.jpg)
How we got into the clouds
• Theme: cutting-edge statistical machine learning works where simple methods fail – Resource utilization prediction – Adding/removing storage bricks to meet SLA – Console log analysis for problem finding
• Sponsor feedback: Great, now show that it works on at least 1000ʼs of machines
4
![Page 5: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/5.jpg)
Utility Computing to the Rescue: Pay as you Go
• Amazon Elastic Compute Cloud (EC2) • “Compute units” $0.10-0.80/hr. $0.085/hr & up
– 1 CU ≈ 1.0-1.2 GHz 2007 AMD Opteron/Xeon core
• N • No up-front cost, no contract, no minimum • storage (~0.15/GB/month) • network (~0.10-0.15/GB external; 0.00 internal) • Everything virtualized, even concept of
independent failure 5
“Instances” Platform Cores Memory Disk Small - $0.085 / hr 32-bit 1 1.7 GB 160 GB
Large - $0.34/ hr 64-bit 4 7.5 GB 850 GB – 2 spindles XLarge - $0.68/ hr 64-bit 8 15.0 GB 1690 GB – 3 spindles
Options....extra memory, extra CPU, extra disk, ...
5
![Page 6: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/6.jpg)
Cloud Computing is Hot *sigh* “...weʼve redefined Cloud Computing to
include everything that we already do... I donʼt understand what we would do differently ... other than change the wording of some of our ads.” Sept. 2008
“Weʼve been building data center after data center, acquiring application after application, ...driving up the cost of technology immensely across the board. We need to find a more innovative path.” Sept. 2009 6
![Page 7: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/7.jpg)
A Berkeley View of Cloud Computing
abovetheclouds.cs.berkeley.edu • 2/09 White paper by RAD Lab PIʼs/students • Goal: stimulate discussion on whatʼs new
– Clarify terminology – Quantify comparisons – Identify challenges & opportunities
• UC Berkeley perspective – industry engagement but no axe to grind – users of CC since late 2007
7
![Page 8: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/8.jpg)
Rest of talk
1. What is it? Whatʼs new? 2. Challenges & Opportunities 3. “We should cloudify our
datacenter/cluster/whatever!” 4. Academics in the cloud
8
![Page 9: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/9.jpg)
1. What is it? Whatʼs new?
• Old idea: Software as a Service (SaaS), predates Multics
• New: pay-as-you-go, utility computing – Illusion of infinite resources on demand (minutes) – Fine-grained billing: release == donʼt pay – No minimum commitment – Earlier examples (Sun, Intel): longer
commitment, more $$$/hour, no storage 9
![Page 10: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/10.jpg)
Unused resources
Cloud Economics 101
• Cloud Computing User: Static provisioning for peak - wasteful, but necessary for SLA
“Statically provisioned” data center
“Virtual” data center in the cloud
Demand
Capacity
Time
Mac
hine
s
Demand
Capacity
Time
$
10
![Page 11: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/11.jpg)
Unused resources
Cloud Economics 101
• Cloud Computing Provider: Could save energy
“Statically provisioned” data center
Real data center in the cloud
Demand
Capacity
Time
Mac
hine
s
Demand
Capacity
Time E
nerg
y
11
![Page 12: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/12.jpg)
Back of the envelope
• Server utilization in datacenters: 5-20% – peaks 2x-10x average
• C = cost/hr. to use cloud (.085 for AWS) • B = cost/hr. to buy server
– $2K server, 3-year depreciation: $0.076 • HW savings = (peak/average util.) – (C/B)
– in this example, save $$ if peak > 1.1x average – can also factor in network & storage costs
• Caveat: IT accounting often not so simple 12
![Page 13: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/13.jpg)
Unused resources
Risk of Overprovisioning
• Underutilization results if “peak” predictions are too optimistic
Static data center
Demand
Capacity
Time
Res
ourc
es
13
![Page 14: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/14.jpg)
Risks of Under Provisioning
Lost revenue
Lost users
Res
ourc
es
Demand
Capacity
Time (days) 1 2 3
Res
ourc
es
Demand
Capacity
Time (days) 1 2 3
Res
ourc
es
Demand
Capacity
Time (days) 1 2 3
14
![Page 15: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/15.jpg)
Risk Transfer vs. CapEx/OpEx
• Over long timescales, a dollar is a dollar
• CC is not necessarily cheaper, esp. if you have steady, known capacity needs
• But risk transfer opens fundamentally new opportunities.
15
![Page 16: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/16.jpg)
Risk Transfer: new scenarios
• “Cost associativity”: 1K servers x 1 hour == 1 server x 1K hours – Washington Post: Hillary Clintonʼs travel docs
posted to WWW <1 day after released – RAD Lab: publish results on 1,000+ servers
• Major enabler for SaaS startups – Animoto Facebook plugin => traffic doubled
every 12 hours for 3 days – Scaled from 50 to >3500 servers – ...then scaled back down
16
![Page 17: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/17.jpg)
Why Now (not then)?
• Build-out of extremely large datacenters (10,000s commodity PCs)
• ...and how to run them – Infrastructure SW: e.g., Google File System – Operational expertise: failover, DDoS, firewalls... – economy of scale: 5-7x cheaper than provisioning
medium-sized (100s/low 1000s machines) facility • Necessary-but-not-sufficient factors
– pervasive broadband Internet – Commoditization of HW & Fast Virtualization – Standardized (& free) software stacks 17
![Page 18: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/18.jpg)
UC Berkeley
2. Challenges & Opportunities
A subset of whatʼs in the paper
Both technical & nontechnical 18
![Page 19: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/19.jpg)
Classifying Clouds • Instruction Set VM (Amazon EC2) • Managed runtime VM (Microsoft Azure) • Framework VM (Google AppEngine, Force.com) • Tradeoff: flexibility/portability vs. “built in”
functionality
EC2 Azure AppEngine, Force.com
Lower-level, Less managed
Higher-level, More managed
19
![Page 20: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/20.jpg)
Lock-in/business continuity
Challenge Opportunity
Availability / business continuity
Multiple providers & datacenters Open API’s
20
• Few enterprise datacentersʼ availability is as good • “Higher level” (AppEngine, Force.com) vs. “lower level” (EC2) clouds include proprietary software
+ richer functionality, better built-in ops support – structural restrictions
• FOSS reimplementations on way? (eg AppScale)
![Page 21: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/21.jpg)
Data lock-in
Challenge Opportunity
Data lock-in Standardization
21
• FOSS implementations of storage (eg HyperTable)
• 10/19/09: Google Data Liberation Front
![Page 22: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/22.jpg)
Data is a Gravity Well
Challenge Opportunity
Data transfer bottlenecks
FedEx-ing disks, Data Backup/Archiving
22
• Amazon now provides “FedEx a disk” service • and hosts free public datasets to “attract” cycles
![Page 23: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/23.jpg)
Data is a Gravity Well
Challenge Opportunity
Scale-up/scale-down structured storage
Major research opportunity
23
• Profileration of non-relational scalable storage: SQL Services (MS Azure), Hypertable, Cassandra, HBase, Amazon SimpleDB & S3, Voldemort, CouchDB, NoSQL movement
![Page 24: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/24.jpg)
Policy/Business Challenges
Challenge Opportunity Reputation Fate Sharing Offer reputation-guarding
services like those for email
24
4/2/09: FBI raid on Dallas datacenter shuts down legitimate businesses along with criminal suspects
10/28/09: Amazon will whitelist elastic-IP addresses and selectively raise limit on outgoing SMTP
![Page 25: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/25.jpg)
Policy/Business Challenges
Challenge Opportunity Software Licensing Pay-as-you-go licenses;
Bulk licenses
25
2/11/09: IBM pay-as-you-go Websphere, DB2, etc. on EC2
Windows on EC2
FOSS makes this less of a problem for some potential cloud users
![Page 26: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/26.jpg)
UC Berkeley
3. Should I cloudify?
26
![Page 27: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/27.jpg)
Public vs. private clouds wonʼt see same benefits
Benefit Public Private
Economy of scale Yes No
Illusion of infinite resources on-demand Yes Unlikely
Eliminate up-front commitment by users* Yes No
True fine-grained pay-as-you-go ** Yes ??
Better utilization (workload multiplexing) Yes Depends on size**
Better utilization & simplified operations through virtualization
Yes Yes
27
* What about nonrecoverable engineering/capital costs? ** Implies ability to meter & incentive to release idle resources
Consider getting best of both with surge computing
![Page 28: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/28.jpg)
So, should I cloudify?
• Why? Is cost savings expected? – economies of scale unlikely for most shops – beware “double paying” for bundled costs
• Internal incentive to release unused resources? – If not...donʼt expect improved utilization – Implies ability to meter (technical) and charge
(nontechnical)
28
![Page 29: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/29.jpg)
IT best practices become critical
• Authentication, data privacy/sensitivity – Data flows over public networks, stored in
public infrastructure – Weakest link in security chain == ?
• Support/lifecycle costs vs. alternatives – Strong appliance market (e.g. spam
filters) – “Accountability gap” for support
29
![Page 30: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/30.jpg)
Hybrid/Surge Computing
• Use cloud for separate/one-off jobs? • Harder: Provision steady state,
overflow your app to cloud? – implies high degree of location
independence, software modularity – must overcome most Cloud obstacles – FOSS reimplementations (Eucalyptus) or
commercial products (VMware vCloud)? 30
![Page 31: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/31.jpg)
Do my apps make sense in cloud?
• Some app types compelling – Extend desktop apps into cloud: Matlab,
Mathematica; soon productivity apps? – Web-like apps with reasonable database
strategy – Batch processing to exploit cost associativity,
e.g. for business analytics • Others cloud-challenged
– Bulk data movement expensive, slow – Jitter-sensitive apps (long-haul latency &
virtualization-induced performance distortion) 31
![Page 32: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/32.jpg)
UC Berkeley
4. Academics in the Cloud:some experiences
(thanks: Jon Kuroda, Eric Fraser, Mike Howard)
32
![Page 33: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/33.jpg)
Clouds in the RAD Lab
• Eucalyptus on ~40-node cluster • Lots of Amazon AWS usage • Workload can overflow from one to the
other (same tools, VM images, ...) • Primarily for research/experiments that
donʼt need to tie in with, eg, UCB Kerberos • Permissions, authentication, access to
home dirs from AWS, etc.—open problems
33
![Page 34: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/34.jpg)
An EECS-centric view
• Higher quality research – routinely do experiments on 100+ servers – many results published on 1,000+ servers – unthinkable a few years ago
• Get results faster => solve new problems – lots of machine learning/data mining research – eg console log analysis [Xu et al, SOSP 09 &
ICDM 09]: minutes vs. hours means can do in near-real-time
• Save money? um...that was a non-goal 34
![Page 35: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/35.jpg)
Obstacles to CC in Research
• Accounting models that reward cost-effective cloud use
• Funding/grants culture hasnʼt caught up to “CapEx vs. OpEx”
• Tools still require high sophistication – but attractive role for software appliances
• Software licensing isnʼt “cost associative” – typically still tied to seats or fixed #CPUs – less problematic for us as researchers
35
![Page 36: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/36.jpg)
Cloud Computing & Statistical Machine Learning
• Before CC, performance optimization was mostly focused on small-scale systems
• CC detailed cost-performance model – Optimization more difficult with more metrics
• CC Everyone can use 1000+ servers – Optimization more difficult at large scale
• Economics rewards scale up and down – Optimization more difficult if add/drop servers
• SML as optimization difficulty increases 36
![Page 37: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/37.jpg)
Example: “elastic” key-value store for SCADS [Armbrust et al, CIDR 09]
Capacity on demand +
Motivation to release unused =
Do the least you can up front
![Page 38: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/38.jpg)
CS education in the Cloud • Moved Berkeley SaaS course to AWS
– expose students to realistic environment – Watch a database fall over: would have
needed 200 servers for ~20 project teams – End of term project demos, Lab deadlines
• VM image simplifies courseware distribution – Students can be root – repair damage == reinstantiate image
![Page 39: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/39.jpg)
Summary: Clouds in EECS
• Focus is new research/teaching opportunities vs. cost savings
• Mileage may vary in other departments • Tools still require sophistication • Authentication, other “admino-technical”
issues largely unsolved • Funding/costing models not caught up
39
![Page 40: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/40.jpg)
UC Berkeley
Wrapping up...
40
![Page 41: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/41.jpg)
Summary: Whatʼs new
• CC “Risk transfer” enables new scenarios – Startups and prototyping – One-off tasks that exploit “cost associativity” – Research & education at scale
• Improved utilization and lower costs if scale down as well as up – Economic motivation to scale down – Changes thinking about load balancing, SW
design to support scale-down 41
![Page 42: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/42.jpg)
Summary: Obstacles
• How “dependent” can you become? – Data expensive to move, no universal format – Management APIʼs not yet standardized – Doesnʼt (necessarily) eliminate reliance on
proprietary SW • SW licensing mostly cloud-unfriendly • Security considerations, IT best practices • Difficulty of quantifying savings • Locus of administration/accountability?
42
![Page 43: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/43.jpg)
Should I cloudify?
• Expecting to save money? – Economy of scale unlikely; savings more likely
from better utilization – But must design for resource accounting &
offer incentive to release – Does hybrid/surge make sense?
• Even if donʼt move to cloud...use as driver – enforce best practices – identify bundled costs => true cost of IT
43
![Page 44: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/44.jpg)
Conclusion Is cloud computing all hype?
No. Is it a fad that will fizzle out?
We think itʼs a major sea change. Is it for everyone?
No/not yet, but be familiar with obstacles & opportunities .44
![Page 45: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/45.jpg)
UC Berkeley
Thank you!
More: abovetheclouds.cs.berkeley.edu
45
![Page 46: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/46.jpg)
BACKUP SLIDES
46
![Page 47: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/47.jpg)
RAD Lab Prototype:System Architecture
Drivers Drivers Drivers
New apps, equipment, global policies (eg SLA)
Offered load, resource
utilization, etc.
Chukw
a & X
Trace (m
onitoring)
Training data
Ruby on Rails environment
VM monitor local OS functions Chukwa trace coll.
web svc APIs
Web 2.0 apps
local OS functions Chukwa trace coll.
SCADS
Director
performance & cost
models
Log Mining
Aut
omat
ic
Wor
kloa
d
Eva
luat
ion
(AW
E)
47
![Page 48: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/48.jpg)
CC Changes Demands on Instructional Computing?
• Runs on your laptop or class Un*x account
• Good enough for course project
• project scrapped when course ends
• Intra-class teams • Courseware: custom install • Code never leaves UCB
_____________________ • Per-student/per-course
account
• Runs in cloud, remote management
• Your friends can use it *ilities matter
• Gain customers app outlives course
• Teams cross UCB boundary • Courseware: VM image • Code released open source,
résumé builder ______________________ • General, collaboration-
enabling tools & facilities
![Page 49: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/49.jpg)
Big science in the cloud?
• Web apps restructured to “shared-nothing friendly” thru 90s; can science do same? – gang scheduling for clouds/virtual clouds? – rethink storage vs. checkpointing vs. code
structure – move to much higher level languages (leave
tuning to macroblocks/runtime, not woven into source code)
– Data-intensive (I/O rates & volume) needs of science apps
• Opportunity for “cost associativity”! 49
![Page 50: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/50.jpg)
SCADS: Scalable, Consistency-Adjustable Data Storage
• Scale Independence – as #users grows: – No changes to application – Cost per user doesnʼt increase – Request latency doesnʼt change
• Key Innovations 1. Performance safe query language 2. Declarative performance/consistency
tradeoffs 3. Automatic scale up and down using
machine learning
![Page 51: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/51.jpg)
Scale Independence Arch • Developers provide
performance safe queries along with consistency requirements
• Use ML, workload information, and requirements to provision proactively via repartitioning keys and replicas
![Page 52: Above the Clouds: A Berkeley View of Cloud ComputingA Berkeley View of Cloud Computing ... Datacenter is new “server ... • HW savings = (peak/average util.) – (C/B) – in this](https://reader034.vdocuments.us/reader034/viewer/2022042221/5ec783b8e474742fe2227fbe/html5/thumbnails/52.jpg)
SCADS Performance Model(on m1.small, all data in memory)
SLA threshold
5% writes 1% writes
99th percen6le
median