IT infrastructure isn’t static. As a business grows, it”s IT Infrastructure needs to grow with it. Changes in the business environment, application changes, equipment replacements all mean infrastructure changes. From time to time, the changes in workflow and storage are sufficient to mean a rethink about the overall infrastructure itself.
IT Infrastructure scaling comes in two flavours – horizontal and vertical. Horizontal scaling, or scaling out is adding more servers and physical resources to the infrastructure, thereby reducing bottlenecks and improving service levels at the expense of increasing its complexity.
Vertical scaling is increasing the processing power and memory of an existing IT Infrastructure.
In a cloud or virtual machine environment, horizontal scaling can mean upgrading the infrastructure to support more VM instances.
In summary, vertical means bigger instances, horizontal means more instances. You can of course do both simultaneously.
To look at each in turn:
Scaling up -Vertical Scaling
In the past, vertical scaling has been the more common approach, especially to in-house IT infrastructures.
Common infrastructure elements for scaling up are CPU power, network connectivity, and storage. The objectives are to increase capacity in the resources supporting the users and their applications, to improve performance and responsiveness, and to provide services where needed.
In CPU terms, it means replacing an existing server processor with a more powerful one, perhaps with a better performance profile, more onboard memory or better bus speeds. If that requires a motherboard change, then it may be better to replace the base systems unit, transferring RAM and other items to the new platform. It might also be driven by a need to upgrade the power supply to support a new processor.
Other items include adding more memory if the motherboard supports it, adding or upgrading storage or interface managers like graphics and sound cards.
These include upgrading switches, changing the existing network architecture by adding new switches to create new segments, and perhaps running new cables to currently unserviced areas.
In an in-house environment, this may take some time because of the need to schedule downtime. There may also be delays because of the need to secure a budget and procure the upgrades.
In an outsourced cloud environment, it should be very much quicker because the MSP (Managed Service Provider) will coordinate the upgrade.
Simply put, the overall effect of the upgrade is that you will have better service levels with an infrastructure supporting improved hardware resources.
Scaling Out – Horizontal Scaling
Scaling Out is becoming more and more common as organisations move to a hybrid cloud infrastructure, with in-house, public and private cloud components. It can arise when looking at scaling up and finding that adding more hardware resources won’t deliver the improvements needed.
The danger is over-commitment. Scaling for a current, temporary resource shortfall could lead to wasted spend.
In a nutshell, short-term issues need vertical scaling, longer-term issues, horizontal. Vertical scaling is often easier to implement since it is usually only a server upgrade, perhaps adding more RAM or replacing HDDs with SSDs. However, if the server cannot physically support an upgrade, then the upgrade scenario needs to be revisited.
In the past, having a structured approach to scaling was not so much of a problem, because all the IT Infrastructure was in-house. However, today, with many businesses establishing themselves on public, private and hybrid clouds, the approach is not so intuitive. For Managed Service Providers, it may even make sense at times of low demand to downscale or mothball excess infrastructure.
Careful assessment and planning of scaling are essential. If you don’t do either, then you could spend more than you need by planning for maximum peak usage, and having resources underused or idle. For example, you upscale to support the additional traffic after a marketing drive to promote Christmas sales. After Christmas, you will have idle resources that you are paying for.
Metrics must be used to evaluate existing resource usage at base and high loads. Many organisations have this kind of measurement policy as a part of normal operations. There are software tools, that given the appropriate data, can extrapolate it to show the effects of adding new resources.
The situation is made more complex if public clouds or MSPs are involved. On the face of it, the MSP or private cloud service supplier are responsible for providing the resources, but you will need to pay for the additional resources. Implementation is easier since that is not your responsibility.
The second area of consideration is staffing. Adding more resources might need additional staff to manage the new environment, provide technical support and oversee network and user security. Again, in an outsourced environment that is not your concern.
One point to note at this point is that if you change the operational environment with an outsourced service supplier, you need to upgrade your contractual terms with them. Any explicit timescales and costs may need to be reviewed and altered as necessary.