What are you looking for ?
Infinidat
Articles_top

2019: Myths and Truths – Virtual Instruments

Cloudbusting, container triumph and why AIO/s needs algorithmic IT.

Virtual Instruments, Inc.‘s SVP of products Tim Van Ash discusses the trends that will dominate the next-generation data centre, covering cloud, tiered infrastructure, high-performance storage, containers, DevOps and AIO/s.

2019: Myths and Truths

Different clouds for different situations: cloud gets real
In 2019, there will be greater recognition by organisations that cloud is not the panacea to all workload challenges. Its failure to meet performance and cost expectations has, for many enterprises, resulted in data repatriation. As a result, the blind, ‘Cloud First’ approach that occurred over the past few years, where organisations enthusiastically embraced cloud without any real idea of why or how they should migrate, will start to be replaced by more critical thinking leveraging workload analysis and simulation to better determine the optimal cloud strategy.

Enterprises will start to shift towards optimising cloud services for specific purposes. For example, companies might leverage a cloud service provider for a faster time to market, for the flexibility, or because the cost envelope for a particular application makes sense. There are some cases where the Cloud First approach can work, but it will be a strategy for specific capabilities, rather than for the entire organisation.

The trend we will start to see, is that there will be a trough of disillusionment about cloud, as organisations realise it is not as easy as first thought. The reality that is becoming understood is that performance SLAs cannot yet be guaranteed and cloud, as a whole is surprisingly expensive. In fact, it can be more expensive, the question organisations need to ask themselves is, “What are we trading off cost on the cloud for?”

With the exponential growth of cloud services, organisations will begin to diversify away from vendor lock-in and increasingly look to a multi-cloud strategy. Cloud-based approaches will continue to replace some of the most popular on-premises deployed application software such as Oracle and SAP. These software vendors are also now encouraging organisations to run these applications in their clouds, which may put them in competition with the major CSPs.

However, the current rate of innovation and raft of new capabilities from vendors like AWS, also brings a multitude of permutations with it, making it almost impossible for customers to determine what they should use. There are too many choices and not enough information on how to make those choices. The days when developers chose which path to the cloud they used is now gone. With the large array of potential options, the ability for development teams to manage that volume of combinations will become unviable. What is needed is a team that can focus purely on sourcing and producing a limited number of options for deployment.

In 2019, people will become better informed and start to realise that workloads should not be moved to the cloud without first going through the process of rationalising and optimising its application services. To fully gain the benefit of what will be a multi-cloud world, organisations must truly understand the resources required to meet performance levels, drive significant cost advantages and achieve capacity reclamation.

Container market maturation
Containers will enable workloads running within the infrastructure, with organisations ‘bursting’ to the cloud, rather than the cloud being the predominant method of hosting that infrastructure. They will empower the internal IT organisation to be more responsive and ultimately make better decisions based on the business need, whether it is agility, a cost profile, or accelerating go-to-market activities.

Applications that are ‘bursty’, meaning that they have significant peaks and troughs of activity, will benefit from a cloud structure as they can expand and contract as needed. Applications that have more or less consistent performance and growth should stay on-premises or be outsourced to a supplier for management.

Tiered infrastructure
Hybrid applications will become the dominant paradigm in 2019, as customers will find it easier to bolt-on new capabilities (for example, website based e-commerce), to their existing back-end databases, than to start from scratch, whether residing in the cloud, or on-premises. This will create new tiers within infrastructures that over time will become increasingly complex to manage.

High-performance storage
In 2019 we will start to see higher performance, higher capacity storage capabilities, such as NVMe over Fabrics impacting the next generation data centre. NVMe over Fabric (the integration of NVMe drives into a networked storage array) will be the harbinger to obvious and incremental performance improvements, but the flip side of that coin is that the full implementation of this protocol will also require organisations to upgrade their infrastructure, end-to-end. Although companies may be purchasing ‘NVMe-ready’ solutions, they will be nowhere near actual production deployment of NVMe over Fabrics for at least another two or three years. Organisations will want to hold back before implementing a solution that will disrupt their infrastructures, and many are still in the middle of life-cycle changes from the upgrade to all-flash arrays from 2016. The cost of upgrading their infrastructures right now would just be too high. This hesitancy will naturally impact heavily on the NVMe vendors, hoping for rapid adoption.

The balance of DevOps vs. ITOps
In 2019 DevOps will continue to focus on rapid deployment of new capabilities, but operating those capabilities will increasingly transition back to IT Operations, to standardise practice, tooling, and to minimise cost. Many organisations have hundreds of staff with DevOps titles dotted throughout the business, all with their own tools and processes, with no way for the company to properly leverage this disparity. This results in exorbitant costs in the running of applications, because DevOps lacks the standardisation and cost management that IT Operations brings to the table.

This means that a ‘single pane of glass’ management approach will become increasingly difficult to achieve in 2019 for most organisations that are struggling with the fact that it must rely on varying tools and metrics, all with differing granularity. Attempting to correlate those tools to pinpoint problems will become incredibly arduous without procuring instruments that provide collaborative approaches, regardless of the infrastructure they are managing.

Increasingly, companies will find themselves with a new set of silos, a new set of technologies, and more gaps between the silos than ever (a major cause of outages). This is caused by their existing disparate tools that lack the visibility needed to view one part of the application and tie that back to anything else. Application Performance Management vendors are all saying they offer end-to-end, single pain of glass solutions but none of them can meaningfully see past the server layer of the IT infrastructure that supports the applications. Organisations will look for ways to solve this problem in 2019, forcing them to rethink their tooling strategies.

AIO/s: what’s the buzz?
There has been a lot of buzz around AIO/s (Artificial Intelligence for IT Operations), but it is more hype than reality (if it is written in Python programming language it is machine learning, if it is written in PowerPoint it is probably artificial intelligence). The way AIO/s is being implemented today is by applying a level of machine learning to what is essentially an event management problem. Applying machine learning to a technique that is already failing does not make that technique any better. What people really fail to see is that AIO/s requires a different approach. We must rethink the problem, not band-aid the solution to make existing methods incrementally better, as the problem is not escalating in increments, but escalating exponentially.

To be meaningful, AIO/s requires the additional element of algorithmic IT, the expertise that comes from training the analytics. Although analytics can produce insights, rarely can it produce interpretations of those insights, which requires integrating sets of data that are not available today. AIO/s is going to need to be reinvented to incorporate real-time monitoring in order to meet the expectations that have been set.

In summary
The overwhelming need to drive and manage the infrastructure and how it is deployed has created many of the pain points we’ve seen in 2018. As digital transformation continues to progress in 2019, there will a greater requirement for the infrastructure to support and align the needs and processes of the business.

Articles_bottom
AIC
ATTO
OPEN-E