HostingCon 2013 – Eliminating Noisy Neighbours and Optimising Performance in the Cloud

HostingCon 2013 – Eliminating Noisy Neighbours and Optimising Performance in the Cloud

One of the exciting reasons for me to attend HostingCon this year was to participate in a discussion panel session. At HostingCon they have educational sessions, some are discussion panels involving a group of people (usually 4 – 5), some are key note speakers, and others are run by one person.

This year I was invited to participate in a discussion panel. The topic of the discussion panel was “Eliminating Noisy Neighbours and Optimising Performance in the Cloud” which was under the Technology session track.

Crucial was invited to attend this discussion panel as we have taken leaps and bounds to solving performance problems in the Cloud from a storage perspective. This is through the release of our Blaze Cloud platform which is explained further here.

The panel was made up of some key people in the industry including,

A little summary of the discussion topic was included as follows,

As the cloud enters its next stage of growth, most of us are not asking why and how we should run applications in the cloud. Now, we are looking to optimize and guarantee performance to meet service obligations and give the best possible service. In this panel session, infrastructure experts will discuss the technical hurdles you will face as you avoid the noisy neighbor effect and push your applications further. You will learn about the hardware and software pain points and how to overcome those obstacles, in a session that begins the new industry conversation centered on quality of service.

What is a noisy neighbour?

It doesn’t matter whether you have a standard shared web hosting account, a virtual server or a cloud server, this noisy neighbour issue has always been a challenge in one way or another.

With our standard business web hosting plans these are a shared hosting solution. This basically means we put your cPanel based hosting account on a server with a bunch of other customers. Now in most cases this is not a problem (as long as you don’t over fill your servers, which is not in-line with the Crucial vision), however it does leave the door open for one or two customers to potentially use up more than their fair share of resources such as CPU power, memory, or disk performance.

So what can happen is we see a customer install a brand new CMS on their cPanel hosting account and it needs a large amount of CPU power to run, this then leads to your hosting account potentially having issues.

So in other words, you are being impacted by your neighbour!

In this particular use case, we solved this problem by implementing Cloud Linux on all our Business Hosting and Reseller Hosting services. This limits each customer to a % share of the servers resources. Creating a fair and balanced environment for consumers to run their websites.

This translates to reliability and increased up time for you.

If you look at our Virtual Server or Cloud server offerings, you get your memory guaranteed (i.e you aren’t sharing that!) and you also get a guaranteed % of CPU resources allocated to your server. This is achieved by using the Xen Opensource as the underlying visualization platform. However we now have a new noisey neighbour issue to overcome, and that is disk performance.

Essentially you have servers all sharing the same “disk performance” so if one or two customers are running an application or task that chews up a large amount of the disk performance, this means there is less for other customers.

This is one of the downsides to the Virtual Server platform, inconsistent disk performance.

Once again, the noisey neighbour issue presents itself!

In this case we launched a brand new cloud platform called the Blaze Cloud, which combines the automation and control of the Onapp cloud software and the guaranteed disk performance of the Solidfire all SSD storage system.

By doing this we have been able to guarantee disk performance for any customers whom use our Blaze Cloud platform, removing the noisey neighbour problem at a disk performance level.

Key Points from the panel?

  • Disk performance has always been a common bottle neck that many hosting companies are dealing with. We had a number of people from the audience ask for examples on how this guaranteeing of disk performance was being achieved.
  • Moving away from hybrid solutions. Often these days cloud and hosting companies are trying to get performance to their clients websites, by using a combination of cloud, virtual and dedicated servers. This is considered a hybrid solution and often is inflexible or too complex. By introducing SAN solutions which can guarantee performance, big steps are being made to move away from the hybrid solution. This benefits the customer ultimately with less complex solutions, which reduces costs, and give the control back to the customer.
  • Being able to choose / select performance. One of the questions that came from the audience was a question on how we the hosting providers are giving the control back to the customer to choose or select their disk performance. Softlayer was not able to answer, nor was any other panel member besides myself. With our partnership with Onapp and Solidfire we have created a unique feature for customers to choose the level of disk performance they have, and they can change that whenever they choose. This impacts the level of guaranteed disk performance a customers virtual disk has access to.

My own takeaways

Here at Crucial we strive to ensure business’s succeed online. One of the key issues we identified was the lack of high end applications or websites being run on the cloud, instead they are being run on large physical servers which is costly, locked into contracts and inflexible. Not ideal for any customer trying to grow their ecommerce presence.

Crucial has looked into the market and partnered with Onapp and Solidfire who are two of the big upcoming players in the cloud market. Through this partnership we’ve been able to deploy a Cloud solution which gives the disk performance customers need to run their websites.

From the discussion panel that disk performance is still a big problem for a lot of different providers whether they are small, medium or large. This area of guaranteed performance is still new and the future is promising.

If you have any thoughts on the above, or were at the conference, let me know what your thoughts?