OPTIMIZATION IS FOR THE CLOUDS | ComsiConsa – Division de WebHosting


February 16, 2012


Optimization is for the Clouds
A primer on accelerating and optimizing
cloud computing on Amazon Web Services with Riverbed Cloud Steelhead
© 2011 Riverbed Technology. All rights reserved. 1

Introduction: Distance computing and skinny pipes

Information is a critical differentiator for businesses today. Innovative tools for creating and using information, and connecting people to it, can help set you apart from your competition. And it isn’t coincidental that the rise of cloud computing parallels the modern explosive growth of information—more resources are available to more people through more ways than ever before. Resources are only a click or an API call away, eliminating procurement delays that impede progress. The elasticity of cloud computing and its seemingly inexhaustible supply of storage continually call out to businesses and consumers to engage their creative processes and to share the results.
Of course, certain challenges arise when all compute and storage resources are remote. These are similar to what organizations faced when they consolidated dispersed branch office resources into central data centers: degraded application performance and overloaded WAN links. While the cost and security benefits of re-centralization are clear, user satisfaction and productivity both suffered—sometimes in financially measurable ways. Because raw bandwidth upgrades didn’t address application performance issues, WAN optimization quickly established itself as the solution to both challenges. For several years now, customers have appreciated the LAN-like performance and bandwidth savings that WAN optimization offers.
Fortunately, what worked for wide-area networks also works for cloud computing. This should come as welcome news to anyone who’s been reluctant to move applications to the cloud. Much of the buzz surrounding cloud computing tries to convince you that you have to throw away everything you’ve done and rebuild from the ground up. It’s true that many applications that run well in a tight LAN will behave erratically in the cloud. However, by optimizing and accelerating communications between your branch offices, mobile workers, and cloud computing resources, you can eliminate potential bottlenecks.
This paper begins with an outline of the characteristics of computing from a distance, describes how WAN optimization of cloud resources improves the user experience, explains how Riverbed’s Cloud Steelhead achieves this, and concludes with several deployment examples.

Moving workloads to the cloud

Early adopters of cloud computing often were starting new projects and could design their applications to take advantage of the cloud’s strengths. Popular workloads included scientific computation, development/test, and content encoding. Greater enterprise cloud adoption will happen as organizations become more comfortable with moving existing line-of-business workloads such as CRM, BPM, ERP, email, file sharing, and backups.
Most business applications exhibit characteristics that could undermine the user experience when distance is involved:

  • They’re highly interactive between the client and the server
  • They rely on chatty and inefficient application-level protocols
  • They move massive amounts of data between locations; this data is often redundant

In these and other cases, WAN optimization technology in the cloud improves the user experience and increases productivity in exactly the same ways it does for traditional branch office-to-data center deployments. And for applications that aren’t sensitive about placement but fail miserably over high-latency links, WAN optimization is the key to moving those off-premise: additional bandwidth won’t eliminate the problems of long round trip times.

Symmetric optimization brings the best benefit

Symmetric optimization employs technology at both ends of the network connection—close to the applications in the cloud and close to the users in branch offices or on mobile clients. This allows for more than simple file caching and helps achieve LAN-like performance by providing mechanisms to reduce the number of round trips over the WAN. Asymmetric optimization technologies, which require deployment at only one end of a network connection, can’t offer the same benefits that symmetric optimization provides.
Consider the case of optimizing only the client side (figure 1). Such devices act mostly as content caches; they’re less effective at reducing the amount of data flowing between the cloud and multiple branch offices and can’t help with latency at all. For example, say that two branch offices have a cached copy of a document. When a user in one branch modifies the document, the optimizer updates both its cache and the original in the cloud. When a user in a another branch requests the now modified document, the entire file is resent to that branch—the optimizer there has no way to determine what was updated and the application can’t ask the server in the cloud to deliver just the changed portion. Client-side asymmetric optimization doesn’t work well in point-to-multipoint deployments, which are actually very common deployment scenarios.

Figure 1. Client-side asymmetric optimization
Or consider the case where optimization technology is deployed only at the cloud side (figure 2). Because there’s no cache close to the clients that can be shared by multiple users, cloud-side optimization is usually limited to some kind of compression that client applications already understand—which few applications actually do.

Figure 2. Cloud-side asymmetric optimization
Compare those cases with symmetric optimization (figure 3), which works without any special requirements of applications, clients, or servers. Data moving between the cloud and branch offices is de-duplicated and replaced by symbolic references, reducing the amount of traffic on the WAN. The cloud-side optimizer tracks all information exchanged with branch-side optimizers; a change made in one branch can be reflected to others without resending complete files back to the cloud or the branches.

Figure 3. Symmetric optimization with Riverbed between the cloud and a branch office plus a remote user

WAN optimization, cloud style

Because the cloud displays some of the same characteristics as WANs, the technology developed for WAN optimization suits cloud computing very well. What differentiates the cloud from other deployment methods are the aspects that make it truly compelling: elasticity, rapid on-demand provisioning, and pay-as-you-go pricing. Your WAN optimizer of choice will need to follow the same model. Allocating storage and doing compute in the cloud is nearly a friction-free process; optimizing the cloud should be, too.
Few organizations will move everything to the cloud. As have other new technical innovations, the cloud will take its place as one of many delivery models in the list from which to choose. To derive the highest benefit from WAN optimization, the same technology should be used at every location: in data centers, in branch offices, on mobile clients, and in the cloud. Using the same software code on the cloud-based optimizers ensures a consistent level of stability and robustness that you’re already accustomed to. Managing cloud-based optimizers should use the same interfaces and follow the same paradigms.
Riverbed Cloud Steelhead: the optimal choice
Cloud Steelhead runs the same RiOS (Riverbed Optimization System) software that backs all our Steelhead products. RiOS applies three independent optimizations to traffic:

  • Data streamlining. Eliminates redundant WAN traffic by de-duplicating data in all TCP flows. Data streamlining analyzes all traffic into indexed segments and stores those, along with references, in the local cache. When RiOS has seen a pattern once, during subsequent accesses it sends only the references. Each appliance, whether physical or virtual or in the cloud, has a single universal data store that’s shared across all applications.
  • Transport streamlining. Improves the performance of all TCP-based applications by adapting transmission characteristics such as window scale, loss handling, and congestion notification. Transport streamlining reduces the number of TCP packets required to complete a transaction, including those protected by SSL.
  • Application streamlining. Decreases the effects of latency by reducing the number of application-level round trips across the WAN. Application-specific optimizations remove much of the chattiness and overhead associated with common applications that normally behave fine on LANs, giving these applications a local-like feel even though they’re running at a distance.

Cloud Steelhead is packaged into a customized virtual machine running on Amazon EC2 and is delivered as an instance of an Amazon Machine Image (AMI), just like other virtual machines on EC2. Because it’s a virtual appliance, it can’t be placed physically in path. Instead it logically sits adjacent to your other compute instances. Those instances run a small in-memory agent that discovers Cloud Steelheads and redirects traffic through them.
To get started, you use the Riverbed Cloud Portal to provision Cloud Steelhead. The Portal creates your resources in Amazon EC2 and launches the Cloud Steelhead instance. After the instance boots it automatically contacts the Portal to obtain license information. Now fully configured, the Cloud Steelhead can begin optimizing and accelerating traffic to and from branch office, data center, and mobile Steelheads.
Cloud Steelhead is available in all Amazon Web Services regions and works with classic EC2 and with Amazon Virtual Private Cloud, including the VPC enhancements released in March 2011.

Cloud Steelhead use cases

If you’re already a Riverbed customer and are contemplating the cloud, we can help accelerate your migration. Read on for a few ideas; you’ll see how adding Cloud Steelhead fits in with your optimization strategy. If you aren’t a Riverbed customer, then we invite you to examine how we can help mitigate some of the challenges that accompany distance-based computing and data movement. The remainder of this paper describes use cases and applications ideally suited for Cloud Steelhead.

Optimizing migration of project workloads and data

Many organizations get started with cloud computing by duplicating an important (but not critical) application to the cloud for a period of testing time. If the testing is successful, the on-premise deployment is decommissioned and the cloud deployment becomes primary. Steelhead can help this project succeed in two ways:

  • The Steelhead appliance in your data center and a Cloud Steelhead in your AWS region can accelerate the initial transfer of applications, data, and even full server images using AWS VM Import
  • During the testing phase and after migration, the Steelheads will continually optimize communications between users on your network and applications running on instances in the cloud, providing LAN-like performance


To understand how much an improvement WAN optimization brings, consider an example in which you have two 500 GB virtual machines. Transferring these over a 10mbps connection takes 13 days; over a 100mbps connection, two days. Steelheads in your network and in the cloud can reduce this time by up to 90 percent.

Reducing cloud-bursting performance disparity

The cloud’s rapid elasticity and near-infinite scale attract many organizations needing additional resources on a temporary, periodic basis. Users expect LAN-like performance at all times; if an application bursts onto the cloud, some users will experience reduced productivity because the connection to the cloud has lower bandwidth and higher latency. Steelheads in the cloud and in the data center will restore LAN-like performance, thus eliminating the performance differential and allowing users to remain fully productive regardless of whether they’re accessing resources locally or in the cloud.

Optimizing cross-region transfer

Cloud Steelheads can communicate with each other, too. Global enterprises using AWS often need to transfer data from one region to another. Because AWS relies on public Internet connections for data transfer between regions, Cloud Steelhead can accelerate these transfers and reduce your bandwidth costs, often by a significant amount. When you configure a Cloud Steelhead in each region, the discovery agent will detect the Cloud Steelheads and ensure that cross-region traffic is optimized.

Using multiple regions can strengthen your disaster recovery plan and provide an even higher level of redundancy beyond spanning multiple availability zones within a single region. With Cloud Steelhead as an element of your design you can:

  • Reduce the time required to synchronize databases across more than one region
  • Quickly copy machine images and volume snapshots between regions
  • Accelerate data transfers from one region to the next in a “follow the sun” application design

Implementing a cloud-first strategy

A growing trend among younger small and medium enterprises is to dispense with centralized IT altogether and instead deploy everything in the cloud. In some cases, venture capital funding even requires this. The economic and technical benefits are clear: applications can be written to take full advantage of the cloud’s capabilities and resources can be added or removed to match demand. There’s no denying the fact that bandwidth usage in these scenarios can be quite high, though. This is an ideal opportunity to combine branch office Steelheads, mobile Steelheads, and Cloud Steelhead together to drive further cost reduction:

  • Application acceleration can lower your monthly telecommunications charges by allowing you to use lower-bandwidth connections
  • Network based data de-duplication can lower your monthly cloud transfer charges by reducing the volume of data traversing into and out of the cloud


Accelerating high-performance computing environments

AWS’s cluster compute instance type brings high-performance computing within reach of many organizations. These applications frequently operate on extremely large datasets that can take a while to transfer over even reasonably fast WAN links. Steelhead’s acceleration techniques can compress and de-duplicate this data often between 50 and 90 percent. By reducing the time required to transfer your datasets to your HPC cluster, you can start your project and obtain your results earlier.

Boosting one cloud with another

You can take advantage of AWS’s international peering agreements to accelerate traffic to other cloud providers that don’t offer Steelhead as an option in their data centers or on virtual machines. As an example, assume you’re a business based in Australia and you wish to consume PaaS or SaaS style cloud services from a provider whose closest data centers are in Singapore. Even under the best of network conditions, latency between these locations can be unbearably high. It’s likely, though, that AWS’s Singapore region is topologically very close to your ultimate destination, perhaps with latencies as low as five milliseconds. You can install Steelhead appliances in your Australia office, deploy a few Cloud Steelhead instances in AWS Singapore, and route traffic to and from your ultimate destination via AWS. Users will experience performance almost like that of a LAN because the long-distance, high-latency links are carrying only optimized traffic.

Optimizing specific applications

In addition to the scenarios above, certain popular applications deployed in the cloud are ideal candidates for optimization with Cloud Steelhead. These benefit from Riverbed’s powerful application-aware optimization technology, and demonstrate the true potential of layer-7 acceleration.

  • Microsoft SharePoint Server. Distributed organizations that rely on SharePoint for collaboration will truly appreciate how well Steelhead accelerates its performance. Steelhead’s SharePoint optimization incorporates all HTTP optimizations plus a deep understanding of the specifics of SharePoint traffic. These extend to HTTPS, too, which is crucial when the servers are accessed over a public network, as is the case with SharePoint running in the cloud.
  • Microsoft Exchange Server. Steelhead’s MAPI optimization works for both encrypted and unencrypted sessions between Outlook and Exchange. Users benefit from the full suite of optimizations, including data reduction and application protocol acceleration. For users in branch offices, MAPI pre-population “warms” the Steelhead in the branch office with Outlook data that arrives overnight; when users log in the next morning, they experience no delay when opening email.
  • Web-based applications, including ERP/CRM. Any application that uses a web-based front-end to communicate with a database will benefit from the improved performance and responsiveness provided by Steelhead’s HTTP optimizations. Of particular interest to cloud computing are ERP and CRM applications, as these tend to be popular products for cloud deployments. Beyond the standard HTTP optimizations, Steelhead provides additional acceleration of SAP and Oracle deployments. This is especially important for organizations that choose to deploy hybrid cloud models, because such deployments often involve continual transfers of large amounts of data between cloud and on-premise resources.
  • Virtual desktop infrastructures. Performance challenges have thus far slowed the growth of cloud-based VDI. Steelhead’s support for Microsoft RDP and RemoteFX, Citrix ICA, and VMware View improve responsiveness such that applications feel local to the user’s desktop.



The cloud works best when it’s invisible to end-users. For many scenarios it’s an ideal deployment model—but one that can introduce network-related performance and productivity problems. Solving these problems requires WAN optimization services that are well-integrated with cloud providers and their business models. Riverbed Cloud Steelhead, built on proven RiOS technology and designed specifically for the cloud, allows you migrate a wider range of applications and data without incurring performance problems normally associated with computing at a distance. Give the cloud an on-premise feel today with Cloud Steelhead.

About Riverbed

Riverbed delivers performance for the globally connected enterprise. With Riverbed, enterprises can successfully and intelligently implement strategic initiatives such as virtualization, consolidation, cloud computing, and disaster recovery without fear of compromising performance. By giving enterprises the platform they need to understand, optimize and consolidate their IT, Riverbed helps enterprises to build a fast, fluid and dynamic IT architecture that aligns with the business needs of the organization. Additional information about Riverbed (NASDAQ: RVBD) is available at www.riverbed.com.

Riverbed Technology, Inc.
199 Fremont Street
San Francisco, CA 94105
Tel: (415) 247-8800

Riverbed Technology Ltd.
One Thames Valley Wokingham Road, Level 2
Bracknell. RG42 1NG
United Kingdom
Tel: +44 1344 31 7100
Riverbed Technology Pte. Ltd.
391A Orchard Road #22-06/10
Ngee Ann City Tower A
Singapore 238873
Tel: +65 6508-7400
Riverbed Technology K.K.
Shiba-Koen Plaza Building 9F
3-6-9, Shiba, Minato-ku
Tokyo, Japan 105-0014
Tel: +81 3 5419 1990

Comments are closed.


Contact us

Proximos Eventos

  • No upcoming events