Log Analysis
Load Test
Site Check










E-insights High Performance Systems Limited.

Slide 1
Slide 2
Slide Explanation

Drivers to the Cloud.

Moore's Law drove processor and memory performance to such an extent that in the vast majority of dedicated server environments, the processors were idle and memory empty most of the time. Costs to maintain servers however (pipe, power and ping) did not decrease. This gave an economic motivation to 'share' multiple independent servers on a single physical environment by using 'vitrualization' technology.

Server Virtualization technology can trace its roots to the early days of Mainframe computrrs and IBM's Multiple Virtual Systems (MVS) technology. While MVS was performing a similar function to today's virtualization technology, the economic motivation at that time were in fact very different.

Keep in mind that virtualization does not make any of the underlying resources used in computing (CPU cycles, memory and IO) cheaper. Rather virtualization allows for much better utilization of such resources than is achievable in most discrete envionments.

The Cloud Promise

What you Need

Configure, and re-configure, the resources you need unconstrained by the need to buy new equipment or by the configuration options manafactures choose.

When you Need It

Respond to time varying needs with time varying resources, rather than with over capacity and idleness

Highly Reliable

Gain the benefits of highly reliable/redundant facilities without bearing all the cost.

Immensely Scaleable

No need to worry about out-growing the environment.

Pay for what you Use

Rather than for the capacity needed to run your maximum workload no matter how infrequent it is.

A Compelling Prospect!

Outlook - Not all Plain Sailing.

While cloud centered computer systems can deliver many benefits, it is not all plain sailing.

High Performance Shared Persistent Storage(File System)

The lack of high performance, persistent, shared file storage can be a challenge when migratings systems based on the assumption that such capability exists. You can try and build it, from scratch or more likely using existing open source software souch as Gluster. However, since such systems do not achieve much if any benefit from virtualization, hence the fact that they are not an offering and are unlikely to become one in the near future, re-desgning so that they are not relied upon should be investigated..

Load Balancers & Firewalls

Virtualized load balancer and firewall capabilities may not have the range of ancilliary functions physical ones do, necessitating system software changes to make up for the deficiency.

Disc Storage IO OPS & Initialization Times

Persistence of disc storage may not be what your are used to from the physical world, nor is the need to consider IO OPS, etc. in such detail.

A Change in Mindset is Needed

Don't try and replicate

When moving a physical environment to the cloud don't blindly build a replica of what currently exists. Take time to consider the limitations & constraints along with the new & enhanced capabilities the cloud environment has to offer.

For other than very simple environments this is hard to do without performing tests implementations. However, time spent coming up the 'learning curve' is time well spent.

Your Cloud vendor is a Mission Critical Vendor

It can be tempting to use vendor specific offerings that address particular needs of a specific environment. The down side can be vendor locked-in. While it may be the right thing to do, this should be a decision made thoghtfully. Deciding on a cloud vendor should be undertaken with the same rigor & considerations used for any mission critical vendor.

Understand & Consider all the costs before committing

Pricing in the cloud is much more fine grained than one would be used to from one's own physical environment. Variable costs may be associated with to resources that in the pyhsical world were supported by capital assets. Data to estimate costs in the cloud may not be readily available from the physical environment. Test implementations can be helpful for this, or roundabout approaches to bracket potential may be necessary.

Flexible & Complicated Pricing Options

Compute Cycles Still Cost

How much do you need

Computing resources can be much more finely selected and can be adjusted over time. By specifying only what you actually need, you can achieve cost savings.

When is it needed

How often are the resources actually needed. Every miute of every day, only at predictable times of each month.

How variable are the needs

Is there a significant degree of uncertainty arounc how much and/or for when the resurces may be needed.

While on demand pricing is an option, it is quite expensive if the resources are used a high percentage of time. The variety of pricing models available can be daunting. Nonetheless, to avoide excessive costs it is essential that the plans be chosen taking the demands and uncertainty into account.

© 2017, E-insights Ltd.