Cloud computing and latency

by Peter Green on April 29, 2010 4:42 PM.

For about as long as people have been writing about cloud computing, they have been complaining of the problem of latency in cloud computing. These latency complaints largely fall into two categories: latency in storage and latency in virtualization. Let’s take a look at each problem and see what we might be able to do to mitigate them, or, even better, eradicate them altogether.

Latency in storage

First is the problem of latency in storage. Consider this quote from a 2-year-old CNET article:

We tend to run applications close to the data they operate on for a reason. That’s because application performance is often largely a function of how quickly it can read and write the data that it’s working on. And data stored on a local hard disk can almost always be accessed faster than that same data sitting at the other end of a network pipe hundreds or even thousands of miles away.
http://news.cnet.com/8301-13556_3-10024650-61.html

In essence, one obstacle to cloud computing and its adoption by your organization is that it sits “far away” from where your data resides. That presents a number of problems:

  • When the data and the cloud are far apart, it takes longer for the cloud to be able to use that data. Consider your database of constituents: your website running in the cloud will take a little longer to access that database than if the website were running right alongside the database.
  • When the data and the cloud are far apart, it introduces potential security concerns. Any time your data travels over public internet lines, you must ensure it does so over an encrypted channel. This introduces complexity into your overall infrastructure; it may not even be possible, if you don’t have the IT staff available to set up such encrypted channels.

In addition, the implication in the referenced article is that the cloud is fine for storing transient data or simple files like images or web pages, but that mission-critical data is better left to “traditional” storage methods (e.g., an in-house SAN, which provide more feature-rich and performance-driven storage). These are all valid concerns… which we address by simply providing better cloud services.

We solve the problem of latency between your storage and your cloud by providing a cloud that can operate as an enterprise-class storage system. By building your cloud to order, we can specify the industry’s fastest drives, or largest drives, or even solid-state storage: whatever your application needs. Suddenly, we can make a compelling case to move your most disk-intensive needs to your cloud, rather than leaving it on an offsite traditional storage array, effectively negating the distance (and associated concerns) between your critical files and your website.

Perhaps you already have a SAN where everything is stored and you’re happy with that. If that’s the case, we can still bring the power of the cloud to your organization. Through our colocation services, we can often colocate your equipment right alongside your cloud in one of our data centers. Or we can send one of our clouds to your in-house data center. Either way, you can use your existing infrastructure with a private cloud to get the best of both worlds.

Latency in virtualization

The other big problem with latency in the cloud is in the virtualization layer. Here’s a quote from an article in last December’s Scientific Computing:

Unfortunately, cloud computing depends heavily on the Internet communications infrastructure or the network within a virtualized datacenter to link all the computational nodes (along with the user) to the cloud. While the investment in the communications infrastructure for the Internet is clearly substantial, it also was designed as a latency-tolerant network. Similarly, the network investment in a virtualized datacenter may not be designed to meet the needs of scientific and HPC computing.
— http://www.scientificcomputing.com/articles-HPC-Cloud-Computing-Pie-in-the-sky-120109.aspx

This article specifically addresses the adoption of cloud computing in high-performance computing in various scientific communities, but the challenge is a valid one for everyday cloud computing as well. The concern is that the infrastructure that makes up the physical layer of a cloud might not be optimized for speed, and may suffer at the virtualization layer as a result.

The good news is that our cloud software was designed specifically to address this concern. As a result, our clouds run at or near “wire speed” on private gigabit networks and virtually eliminate any potential for delays on the virtualization/networking layer. If gigabit ethernet isn’t fast enough, our clouds have been certified on 10-gigabit networks as well, and will support faster standards as they come out. And with our ability to build your private network to your personal specifications, we can support virtually any configuration, putting any claims of “latency in virtualization” firmly in your rear view mirror.

Summary

Despite all of this, reports continue to be released that show a relatively high reluctance to adopt cloud computing. To this, I say: drop me a line. Tell me what you need, tell me what your cloud concerns might be, and let me tell you how we can bring it all together for your organization.

Bookmark and Share

Comments

Leave a comment