22. [email_address] DBS India, 1 st Floor, World Trade Towers, Barakhamba Lane, New Delhi 110001 6225 S. Mojave Rd. Suite B Las Vegas, NV 89120, USA. T: +1 408 459 0657
Editor's Notes
- definition of On-Cloud - remote resources: servers and software owned and managed by vendor, in vendor’s own data center. Nearly always shared, a.k.a. ‘multi-tenant’; can mean different things but key point is installation is not purchased/built for individual client, so vendor can build out capacity in advance - self-provisioned: client decides what they need; can add or remove resources, can specify details of setup. Thus, need minimal help from vendor. Is key to economy and moving quickly. - user operated: client actually runs the system, even though the infrastructure is managed by someone else. Gives client control. - think of as new delivery option, between In-house and outsource - in-house, are mostly dependent on IT for resources and provisioning, but user still operates. - outsource, are dependent on vendor for everything: resources, provisioning, operation - On-Cloud is intermediate: remote resources like outsource, but user operated like in-house. self-provisioning is new. - result is to free user from in-house IT without making hostage to external vendor. gives user greatest possible independence.
advantages - little/no fixed investment - up front. Makes easy to try new things. Often can fund from operating budget not capital budget. - in general, On-Cloud is not necessarily lowest long-term cost vs. stable on-premise application. - but for BI in particular, where requirements change constantly, may actually be cheaper long term as well because have lower cost of change - easy to expand / reduce capacity - due to self-provisioning, shared hardware. - especially important for BI, where can have huge short-term projects e.g. to test something new. - easy to use / no tech skills - this refers to deployment, since users still run the actual software - but is critical because tech skills for use can be very different from tech skills for deployment, & access to technical specialists for deployment can be critical stumbling block - quick deployment / easy to change - all the reasons above; result is fast time to value - especially important in BI where needs change rapidly - easy to underestimate advantage vs. conventional remote hosting, where still have to purchase, set up and manage own servers. - driven home to me the other day when were sitting in client’s office other day and listening to them buy a server – spent half a day, and was fast-growing SaaS company that couldn’t expand without it. Had they been cloud-based, would have been few minutes at control panel. - so now we’ve defined On-Cloud; let’s move to defining business intelliigence
- quick and dirty definition: ‘information used for decisions’ - includes any kind of data analysis; range from standard reports to dashboards to ad hoc exploration; - most recently includes operational / predictive / distributed BI, which guides front-line actions vs. management decisions / strategies - whatever the application, has same basic components - warehouse, which includes all the work of: - loading from source systems - checking for quality and integrating data from multiple sources - putting in proper structures for specific applications - note that different applications need different structures - presentation layer, which is what’s typically referred to as a BI system - this is tools used by data consumers: reports, dashboards, etc.; - key point is that most of the work is in the warehouse side - so when you think about removing obstacles to BI systems, you have to consider the warehouse too
- basic challenges to BI - need tech skills to set up - esp. the data warehouse part. - requires IT support - because they understand the source data and control the source systems, - also because the warehouse design is often tailored closely to applications - costly to deploy - mostly because of tech staff costs; somewhat due to hardware and software - time-consuming to implement - must take time to get design correct before moving ahead, for two reasons: - design must be tailored to system, so otherwise would rework - tech staff is very limited and expensive, so must use their time efficiently - staff is critical because are usually few people with those skills available & need for many projects, regardless of cost - may be short-term / one time project - hard to justify short-term projects because of big upfront investment - may require large amount of resources - again, not only is hardware expensive but it’s costly & time-consumer to set up good news: all of these are problems that On-Cloud addresses
- what we’ve done here is match previous slide on BI challenges vs. earlier slide of On-Cloud Advantages - happen to fit very nicely together - if had been live seminar, could have generated those lists on a white board and would have come out about the same - because those really are the issues and advantages that most people think of
So, let’s ask that question again: Does On-Cloud BI make sense. - yes, On-Cloud is good - yes, BI is good - yes, On-Cloud strengths match BI challenges - QED quod erat demonstrandum - but before we get too excited here, let’s remember that some people have some doubts about On-Cloud
- each of these objections has a reasonable solution - security: vendor processes are documented and auditable; often better than in-house; is core competency - process integration: APIs increasingly available - reliability: service level can be negotiated to some degree; also can pick projects that are not mission-critical. Most BI falls into this group.) - vendor lock-in: matter of understanding TOS and negotiating as needed - inadequate tools: pick the right vendor; also, is easy to move on if vendor doesn’t meet future needs
- this is more complicated than buying disk space in an online server farm or even configuring seats in a sales automation system - it’s why traditional BI projects are so difficult and take so much skilled labor - if On-Cloud BI still requires all that skill, then whether the system is on-premise or cloud-based makes very little difference So, to succeed, cloud-based BI must find ways to reduce complexity – that is, reduce the need for skilled labor -- not just shift the site where it’s executed
Two basic strategies are available: - make standard BI technology more efficient, or - use non-standard technology that is inherently less demanding (=more tolerant of complexity) 1. making standard technologies more efficient can mean - simplifying their operation; e.g., add a user interface that hides much of the complexity of importing data, defining data models, building queries, reviewing results. Wizards and data visualization tools are good examples. - automating their operation: let the system set up the BI environment and explore the results, maybe with some human guidance. Automated data profiling, matching and relationship discovery are good examples. - narrowing the scope: limit the system to well-understood data sources and analyses, so work can be done once using conventional methods and then reused. Systems that analyze data from Salesforce.com are one good example; systems that analyze bank or telco data to predict attrition are another. 2. using non-standard technology mostly means using non-standard databases, like columnar and in-memory systems, that don’t need data models to be tuned in advance for specific requirements, for example by defining data cubes or entity-relationship models
So far, we’ve identified two sets of BI problems that cloud-based BI can overcome 1. standard problems solved by cloud-based in general: - need tech skills to set up - costly to deploy - time-consuming to implement - requires IT support - may be short-term / one time project - may require large amount of resources 2. BI-specific problems - data preparation / CDI - data models - need for flexibility - analytical skills
But there’s also a third set of problems that even cloud-based can’t solve - data is not available (either it doesn’t exist or isn’t accessible) - moving data to an external system is too costly (because of volume) or too slow (because of latency requirements, mostly for operational BI) This is the set of problems that either your in-house IT team or an external resource, such as an outsourcing vendor, will have to solve. A cloud-based solution might solve them too, if it has the appropriate professional services available.
Taking these three sets of problems together, then, we can finally answer the question of when cloud-based BI makes sense:
- pick a project - clear goals - measurable value - reasonable scope - available data - quick completion, - limited risk - pick a vendor - technology (type, scalability, complexity, API, services as needed) - terms of service (data access, service levels, intellectual property) - professional services if needed - pricing: volume levels, contract period) - speed and ease of provisioning - execution - work incrementally - plan an exit strategy: know what you’ll do if you succeed (bring in house, put into production, expand scope, etc.) - capture what you learn