RE: Why WAN optimization should be integrated with branch office routers, Part 1

I would tend to agree with Justin, in addition to adding a few more comments.

- There is at least one WAN Optimization appliance that I know of that performs "unattended install" quite well. Most vendor appliances are also managed via some sort of Central Manager which both aid in reduction of OpEx.

- Deploying a stand alone device allows greater flexibility in scaling. Seperating the OS is also a benefit, relying on IOS which can be "buggy" at times could cause issues with what the router does best - route. Some vendor appliances offer more disk space which helps build and keep a larger dictionary, and offer more memory to provide for specific optimization techniques.

- Deploying an appliance in an "off-path" solution (WCCP) also allows for "quick" remediation if there are any issues with the technology by simply disabling redirection.

For most vendors an off-path solution would size to just about any branch; whereas the technology deployed in a router would require a 2800 or 3800. This might not be feasible in smaller branch sites for some enterprises.

Just some thoughts from the experiences I have had with testing and evaluation.

I have a few insights based on this article.
Let's start by looking at what WAN optimization actually covers. This is a broad church taking in technologies such as (but not limited to) Web caching, CIFS protocol acceleration, compression, TCP Proxy.
Of course let's not forget reporting because if you can't report on what's going on then there's really no way other than subjective "user experience" assessment to understand if any of these technologies is providing a benefit.
When you consider this broad range of different services that can be deployed to provide an "optimization service", the actual customer requirements and the fact that some of these techniques do not always require equipment to be installed on branch sites, then the "economy of scale" argument for using services on routers starts not to stack up fully due to the potential complexity and cost of the system.
What is it that an enterprise is actually looking for when they start considering optimization services? It can be many things. They may want a managed service delivered by its Service Provider that guarantees key application performance or the company may want to use optimization as a mechanism to provide additional in-house management and understanding of how the network is used.
Bearing this in mind and looking at what services are on offer on routers at the moment, then it soon becomes evident that what's really required is a way to provide optimization services in a flexible, scalable and above all simple fashion, only utilizing hardware where there is a specific requirement to do so.
Most enterprises are looking to improve the performance of key applications with optimization. To do this successfully it's necessary to have session by session control of all the applications using the WAN to ensure that the resources available are used as efficiently as possible to support the critical applications and therefore the business. Once this control is in place, the enterprise can consider using additional technologies to provide additional benefits, for example TCP-Proxy acceleration to efficiently use high-speed or high-latency links or compression and caching to reduce overall data volumes. The key here is that if you can't ensure that the benefits of these acceleration technologies is re-applied to supporting the key applications, session by session, then you could be just providing more resources for the less important applications.
Optimization (per session application performance management) and TCP Proxy acceleration can both be delivered in the majority of situations without the need for hardware or router services everywhere. Other acceleration techniques that do require hardware or router services can then be deployed tactically within the overall scheme of network optimization and management to deliver their benefits where they are required.
One final point is that measuring and reporting on what's going on is vital to the success of any optimization or acceleration project. Therefore it makes sense to look to solutions that provide these capabilities as a core function to provide the basis of any WAN Optimization system.
Mark Burton

Ipanema Technologies

Your comment doesn't hold water at all. Even if you were to use a Cisco 3845 at every location it would still be much cheaper than using a generic router and a series of other appliances including switches, wireless, voice, and WAN optimization.

Furthermore, with appliance sprawl, your failure domain is TREMENDOUSLY larger than if you had everything in one system, because you have X number of processors, memory subsystems, power supplies, hard drives, network connections, and more. Couple that with the fact that an appliance is built using commodity components whereas a router is constructed from the ground up to provide high levels of availability and what you get is a solution that is integrated, consuming less rack space, with generally higher availability characteristics.

Also, these technologies integrate in a fashion whereby failure of a single component doesn't knock out the system. For instance, a WAAS module failing causes it to be removed from the WCCP service group (just like an appliance would) and it can then be replaced.

If you need high availability, you're going to need two of everything anyway, which makes the initial AND ongoing cost of the integrated solution FAR BETTER.

Here are my 2 cents:

As any technology oriented to WAN Optimization, the first step should be professional sizing of the expected results and a proper assessment of the existing infrastructure.

Sizing should include a clear understanding of focal effects directly related to productivity and savings as well as detailed technical planning, as commented elsewhere.

Once a proper sizing and planning is done, the WAN optimization deployment should fit pretty well in the infrastructure and performs usually as expected with minor adjustments and rare disruptions in the network traffic.

Regarding specifically Cisco’s WAAS solution and the ISR products, the fact of putting too much functionality in one box is an issue; the decision to use all of it is matter of choice.

This trend is not new neither in hardware nor in software. Modularity and multi functional devices are part of most technologies available today, including printers, cell phones, home appliances etc. and off course switches, routers, firewalls etc. Obviously the network architect’s choice determines which model to follow and what level of modularity fits her/his target SLA better.

WAN optimization benefits are as huge and undeniable as server centralization, consolidation and virtualization, which cause actually the hype about this technology. The ROI and TCO, even handled in a very conservative manner, show very appealing times and savings in bandwidth costs, productivity increases and bandwidth release for other uses, like real time applications. Even awfully made ROI analysis are attractive in less than a year projection, so, this technology will get financial support as well as high expectations. Be prepared for both.

Be aware to consider management for Wan Optimization: Policy distribution and MTTR could be a mess without it. All major vendors in this space have a sort of management appliance that relief configuration, reporting and performance automation.

As today there is no interoperability grounds to consider a mixed vendor scenario, I think it would be better to consider a single brand, and if the scale comes down to medium to very small sites, the ISRs would have the advantage of reducing gear that is located far from the enterprise IT reach, not meaning that the ISR is the only way to deploy WAN Optimization in small sites, but it just seems the wiser to me.

Regards:

Luis Guembes
Technology & Telecom Manager
Adexus Peru S.A.

If you needed the functionality of all of the modules than the economic benefit is validated. Moreover, the alternative cost of maintaining 4-6 devices--costs associated with real estate, cooling, maintenance,etc.--not to mention the cost of having staff trained on multiple platforms is significantly higher than maintaining redundant and/or efficient sparing of the Cisco ISR platforms.

If you needed the functionality of all of the modules than the economic benefit is validated. Moreover, the alternative cost of maintaining 4-6 devices--costs associated with real estate, cooling, maintenance,etc.--not to mention the cost of having staff trained on multiple platforms is significantly higher than maintaining redundant and/or efficient sparing of the Cisco ISR platforms.