You wouldn't be surprised if i told you that we do live in interesting times. New business models are created at the same pace today at which older ones are being destroyed. Technology is no longer just an enabler for business, it has become the business for most organizations. In this session we will touch upon some of the challenges and opportunities that the cloud has to offer. The cloud (IaaS, PaaS, SaaS) as we know it offers organizations immense opportunity in terms of reducing time to market on delivering engaging customer experiences but with all of that agility a move to the cloud also brings numerous challenges, some obvious and some not very obvious. In this session we will go over challenges engineering systems for the cloud including a case study engineering a complex legacy application for the cloud.
3. Your Presenter
• Trevor is part of the ATA (Advanced Technology Architecture) Group within
Technology Consulting and leads the Performance Engineering Capability for ANZ
• In his current role Trevor works closely with accounts across ANZ advocating
Proactive Performance Management with a view to delivering applications that scale
first time around.
• Trevor Warren is passionate about challenging the status-quo and finding reasons to
innovate.
Name : Trevor Warren
Position : Technical Architect, Performance
Engineering Capability Lead, ANZ
Organisation : Advanced Technical Architecture
Contact # : 0423 607 688
Email : trevor.warren@accenture.com
4. Interesting Facts
• Been an aviation buff for a long time now
• Used to fly, but no longer
• Used to travel the country preaching the gospel
of Open Source & Free Software i.e. GNU,
Linux, etc.
• Organizer for the Melbourne Raspberry Pi
Hackers Club (CCHS, Hawthorn)
• Also run an Open Body Of Knowledge on
Systems Performance Engineering called –
Practical Performance Analyst
(www.practicalperformanceanalyst.com)
5. Agenda
• A Bit Of History – The Case For Change
• Bi-Modal IT – Managing Unpredictability & Delivering Required Change
• Cloud Computing Today - Opportunities
• Customer Expectations
• Costs Of Cloud Adoption - Challenges
• Case Studies
• What You Might Want To Do
28. In the new Digital world, performance is a key enabler of the customer
experience. Poor performance will make or break a digital business.
In A Digital World Performance Is An Enabler…..
Faster systems allow IT spend to be moved from license and hardware
spend to services spend.
29. Customer Experience : What Does Current Research Tell Us
• 0.1 second is about the limit for having the user
feel that the system is reacting instantaneously.
• 1.0 second is about the limit for the user's flow of
thought to stay uninterrupted, even though the
user will notice the delay.
• 3-5 seconds is about the limit for keeping the
user's attention focused on the dialogue. The
user’s flow of thought is interrupted and the user
doesn’t feel that the system is reacting
instantaneously.
• After 5s, users get impatient and notice that
they're waiting for a slow computer to respond.
• The longer the wait, the more this impatience
grows; after about 8 seconds, the average
attention span is maxed out.
Source : KissMetrics, Akmai, DynaTrace
30. Customer Experience : What Does Current Research Tell Us
Source : KissMetrics, Akmai, DynaTrace
31. It’s Still Your Problem
The customer does not care where your applications are hosted or how complicated your
architectures are. You still have to deliver 2 second response times !!!
35. Cloud Engineering Challenges - I
Limited visibility (Infrastructure Performance, Application Performance, etc.) across the stack on PaaS/SaaS platforms.
36. Cloud Engineering Challenges - II
Vendor’s ownership of Performance SLA’s is “generally” limited to within the data center.
37. Cloud Engineering Challenges - III
Distributed systems with components across multiple providers/regions makes obtaining an End to End view of
performance tough if not impossible.
39. Cloud Engineering Challenges - V
Slashdot Effect : Being featured on social media may result in very fast massive increases in traffic that may crash a
system that is sized for typical daily workload.
40. Cloud Engineering Challenges - VI
Customers expect to have a consistent experience across multiple device types and interfaces. Managing and
delivering on SLA’s are still your responsibility.
42. Web Performance Engineering - ShopZilla
• 100 Million impressions a day
• 8,000 searches a second
• 20-29M unique visitors a month
• 100 Million products
• 16 month reengineering effort
• Page load from 6 seconds to 1.2
• Uptime from 99.65% to 99.97%
• 10% of previous hardware needs
http://en.oreilly.com/velocity2009/public/schedule/detail/7709
!!! Revenue went up 12% !!!!
44. Enterprise Application Migration Into The Cloud
1. Background – Signed up to migrate SAP (ECC, BW, Portal, MII, PI, ~40 supporting applications) workload into the cloud.
Initially scoped as a lift and shift. Dev, Test, Pre-Prod and Prod to be migrated within ~10 months.
2. Environment complexity – Support for some core application performed by 3rd party vendors, application connectivity
requirements not clearly documented, legacy application mix, sheer number of applications, Unsupported systems, Outdated
platforms (e.g. Win2K3), BAU enhancements in flight, etc.
3. No prior evidence of PE or PT – Performance testing historically focussed on core ERP i.e. ECC for x100 VU, production
workload is in excess of x1000 VU. Poor representation of production workload for ECC, No PE or PT for other systems, lack of
a performance baseline for key production workload.
4. Lack of Non Functional Requirements and Mapping – NENFR’s or Non existent Non functional requirements. Lack of integrated
environments including process for validation performance for most of the key systems.
5. Migration Schedule – Complex phased cutover schedule. Key SAP systems migrated in a two step approach. Challenges with
ECP DB (~8 TB) and BWP DB (~6 TB) largely due to performance (R3 loads, custom code, etc.). In ability to meet cutover
schedule, custom migration code required with MS release custom patches, 5 x G5’s to scale out R3 loads, etc.
6. Engineering Environment – Challenges obtaining funding to stand-up and maintain production size environments. We were
required to stand up a copy of production within a matter of weeks to support engineering of systems. This also allowed for
triage and coming up with break fixes post go live.
7. Initial environment stability issues – Server reboots, storage performance issues, cluster stability issues, challenges completing
RCA, etc.
8. Custom code – Poor quality of custom code piled up over the years combined with lack of coherent design and architecture. Re-
engineering would be an ideal solution but could consume anywhere from ~18-24 months.
9. Challenges transition from old to new – Importance of having existing resources focused on keeping the lights on while having
to re-train support teams on managing the new systems, resources unable to support complex environments, challenging
dealing with un-foreseen challenges in production and make critical decisions on the go.
10. DB Performance – Move from DB2 (Aix) to MS SQL 2016 on Windows 2012. Design replaces BWA with Columnar Store.
48. Where Do We Go Wrong
Software Development
Life Cycle
Functional Requirements Gathering
Architecture & Design
Build Application
System Test,
System Integrated Test & UAT
Deploy Into Production
Performance Engineering
Life Cycle
Non Functional Requirements
Gathering
Design for Performance &
Performance Modelling
Unit Performance Test &
Code Optimization
Performance Test
Monitoring & Capacity Management
No NFR’s, No Workload Models
No PoC, No Consideration for
Performance, Will deal with
Performance during test
Lack of Diagnostics or code
profiling, Lack of continuous code
optimization & Unit Perf testing
Lack of well articulated NFR’s,
Poor or no workload models, lack
of Industry standard tools
Focus on System monitoring, Lack
of understanding of Capacity
Management or Perf Modelling
Performance Engineering =
Performance Testing