Date post: | 18-Feb-2017 |
Category: |
Technology |
Upload: | scott-goessling |
View: | 234 times |
Download: | 1 times |
1 | ©2015 Burstorm Inc.
Automated Infrastructure and Cloud Modeling
Case Study – ExampleSeptember 2015
Scott GoesslingVP Cloud – Burstorm, Inc.
2 | ©2015 Burstorm Inc.
Current VS. Market
Client has asked Burstorm to provide data comparing the costs of Client’s existing infrastructure to the publically available market cloud services and set up a frame work to examine financial impact of future redesigns and potential optimizations for existing Client environments.
3 | ©2015 Burstorm Inc.
Building the Blueprint for Visualizing Current State
For the duration of the engagement you will have access to a Burstorm Cloud Services Specialist who will work with the Client team to gather data and visualize costs for existing targets
4 | ©2015 Burstorm Inc.
Blueprint (Step 1) – Gather Existing State Data
Inventory and In-house Cost Data
Actual Utilization Data
Network Contract Data
Colocation Contract Data
AWS Bill and Utilization Data Universal
5 | ©2015 Burstorm Inc.
Blueprint (Step 2) – Analyze Existing State Data
Utilization Patterns and Heat Maps
Data and Cost Normalization
Memory Utilization
CPU Utilization
Data Validation
6 | ©2015 Burstorm Inc.
Blueprint (Step 3) – Enter Contract/Svc Data
7 | ©2015 Burstorm Inc.
Blueprint (Step 4) – Import AWS Billing Data
8 | ©2015 Burstorm Inc.
Current State Blueprint
9 | ©2015 Burstorm Inc.
Blueprint DashboardNote: “Savings” box indicates annual savings of current market rate vs current contract rate for only the two U.S. colo contracts listed below.
Universal
Universal Technologies
10 | ©2015 Burstorm Inc.
Blueprint Interactive Contract Data
23 Active
11 | ©2015 Burstorm Inc.
Models and Scenarios
Determine Scenarios of Interest and organize data into associated models & scenarios based on client priority. Create hierarchy and alternate views via copy, import, export etc.
1. In-House services2. OpenStack3. AWS4. Organize by location, by app, by region
12 | ©2015 Burstorm Inc.
As-Is Environments (Legacy, OpenStack and Colo)
13 | ©2015 Burstorm Inc.
Legacy vs Market (As-Is Model)
A match is available for Legacy in current sizing if purchased as virtual servers as a service. The matching solution is 26% less than current OpenStack compute costs with TCC factor
Amazon infrastructure is currently deployed in smaller sizes and will match more often showing more options
Utilization based design and workload redistribution can have significant cost benefit (Ex. 93% or more in some models)
Current sizing is 100% custom in current market. No matches exist for current deployed sizing for any of the Client in-house environments in scope.
Universal
Universal
14 | ©2015 Burstorm Inc.
Two solutions were identified for OpenStack in current sizing, if purchased as virtual servers, as a service. The two solutions were 11% and 33% less respectively when compared to current OpenStack compute costs with TCC factor included.
Amazon infrastructure is currently deployed in smaller sizes and will match more often showing more options
Utilization based design and workload redistribution can have significant cost benefit (Ex. 93% or more in some models)
OpenStack vs Market (As-Is Model)
Current sizing is 100% custom in current market. No matches exist for current deployed sizing for any of the Client in-house environments in scope.
Universal
Universal
15 | ©2015 Burstorm Inc.
Legacy vs OpenStack (As-Is Model)
• Core costs for OpenStack are 40% higher than core costs for Legacy
• RAM costs for OpenStack are 13% higher than RAM costs for Legacy
• OpenStack is consistently the most expensive internal service by 27-32% (rises with additional core count)
+27%
OpenStack +27% vs Legacy
Universal
Universal
16 | ©2015 Burstorm Inc.
Colocation vs Market (SV5, As-Is Model)• Current colocation costs appear
to be much higher than current market
• Current environments are an average of 39% higher than market estimates with OpenStack currently 51% over market
• Per kW prices include space and power but do not factor in any NRC or installation requirements like PDUs.
100kW EQX - Legacy EQX - Openstack Layer42 Digital Realty Market Est.
Per kW 376.87 448.65 285.02 190 220.00
Diff Mkt +42% +51% +23% -15% -
Universal
Universal
Universal
17 | ©2015 Burstorm Inc.
Colocation vs Market (AM2, As-Is Model)
• Quoted 24 cabinets @ 3.5kW power in secure cage (suite)
• Current colocation costs appear a bit higher than market by 14%
• Europe is still custom quote in nearly all scenarios
• There is some opportunity to improve via colo, minimal when compared against moving to service based compute and storage solutions.
84kW EQX – AM2 Digital Realty SingleHop Market
Per kW 323.49 240 330.00 285.00
Diff Mkt +14% -16% +16% -
18 | ©2015 Burstorm Inc.
Example Highlights for As-Is Environments
• Network Cost negligible. Current costs not out of market range. Any anticipated change in the mount of network spend vs total solution spend is not likely to skew any numbers, especially if geographies remain the same.
• Continuing with current server sizing and application layout/distribution will eliminate nearly all options for cost control in utilization based models and optimized infrastructure designs
• AM2 space and power seems to a bit higher than current market. • In both the U.S. as well as abroad, services like AWS have proven to be more
cost effective against models based on Legacy and OpenStack product sets.• AWS current state models, while more cost effective against current internal
costs on Legacy and OpenStack, show that significant cost benefit is available in the market when utilizing as a service type solutions with other solution providers
19 | ©2015 Burstorm Inc.
Optimization Models (Assumptions)
Burstorm models can only be as accurate as the data provided by Client. A few assumptions were required as Burstorm validated and normalized data. Models were built for as needed and ongoing comparison against the latest market data.
1. Client internal product sets were created for ongoing use which include searchable attributes for virtual, shared, dedicated, physical and metrics
2. Workloads can be redistributed and utilization patterns can be considered3. Normalized CPU/RAM costs across all environments4. Storage data was removed for consistency in costs and modeling5. Legacy to OpenStack migration not complete at the start of project
20 | ©2015 Burstorm Inc.
• Benchmark size is 2x32, a potential optimized size based on actual utilization data from Client. This same size could be deployed in Legacy as well as OpenStack.
• OpenStack is 93% greater than market cost leader at benchmarked size
• OpenStack cost is 23% greater than AWS market cost at benchmarked size
• Legacy is less expensive than AWS with other good options at lower costs than Legacy in appropriate geographies
OpenStack vs Legacy vs AWS vs Market
2x32 Linux OpenStack Legacy AWS Low CostMarket Cost (USD) 473.59 361.19 362.88 31.99
% Diff - -24% -23% -93%
AWS 23% Lower Cost
vs OpenStack
Low Cost Option 93% Lower
vs OpenStack
21 | ©2015 Burstorm Inc.
• Windows to Linux alternative brings in very good price/performance providers in appropriate geographies for Client
• Additional services accounts for ~50% of AWS bill• OpenStack is 378% higher cost for same solution at current sizing on AWS• Legacy is 248% higher cost for same solution at current sizing on AWS
AWS Current Bill vs Market
27% Lower Cost
22 | ©2015 Burstorm Inc.
Price/Performance Metrics• Other providers have better price/performance metrics vs AWS• US and Europe, (specifically Amsterdam) locations available• Performance, price, sizing and location directly affect optimal solution design• Resizing design and reviewing app layout may yield additional optimization
23 | ©2015 Burstorm Inc.
Example: Legacy and OpenStack (Utilization Base)
Current Utilization Differ %
Total $ 2,798,217.90 $ 543,867.58 -81%Cores 4156 208 -95%
RAM (GB) 32088 11893 -63%
Current Utilization Differ %
Total $ 543,190.50 $ 276,782.63 -49%Cores 536 147 -72%
RAM (GB) 5088 4054 -21%
Legacy
OpenStack