MOSS has highly configurable and flexible architecture that can easily accommodate some of most challenging collaboration requirements in large enterprises. One of the biggest benefits to leveraging MOSS as opposed to developing your own architecture for portal/collaboration solutions is the prepackaged saleable design of MOSS that you get out of the box. However as with many large enterprises one of the more complex design choices is on how to make the collaboration solution available to its geographically distributed workforce . Broadly the choices would be to either adopt one of the following options
- Centralized solution
- Decentralized solution
In most cases a centralized solution would be recommended so that the operations, governance, High Availability(HA) and Disaster Recovery(DR) can be managed from a single location and a tighter policy can be enforced, however in some cases the network may not be as reliable or scenarios where centralization may have challenges with high latency between the branches and the central/ low bandwidth etc may cause the enterprise to adopt a decentralized solution. But how do you know which is the optimal MOSS deployment strategy in your case?
Last week we were working with a large financial service provider in North America to demonstrate the feasibility of deploying MOSS as a central solution, the strategy to improve the end user experience for a centralized MOSS solution was to:
- Provide a highly scaled up + scaled out MOSS farm - Based on the role of the MOSS server i.e. Web Front End(WFE) or Index Server it can be either scaled up or scaled out only. Providing the right amount of scale ensures that the centralized solution can cater to the requests from users across the enterprise.
- Optimal logical architecture and information architecture - the right design ensures that a single ContentDB is not the bottle neck for all requests.
- Compression - compression improves performance over the wire
- Caching - Caching can drastically improve the end user experience, a smart caching solution will ensure that only delta changes from previous requests are being sent over the wire.
Both 3 and 4(above) can be achieved by leveraging a SharePoint aware network accelerator such as Certeon, other general purpose network accelerators may be used as well, however leveraging a SharePoint aware one is always better.
As with any other large project undertaking its important to benchmark the system performance before rolling it out into production, metrics captured during the benchmark would help validate if a centralized solution would be the right choice, additionally it could provide metrics that can be used to decide if additional bandwidth needs to be arranged etc. The way we benchmarked end user experience before rolling the entire solution in production was by leveraging Shunra. Shunra can be used to capture the network conditions (latency, packet loss) between the various branches and central office, once the data is captured it can be used by the Shunra device to simulate WAN conditions in a lab. Additionally Shunra can be used to simulate what-if scenarios that may occur in your WAN. Running test scripts simulating end users and having Shunra simulate various WAN conditions can provide valuable metrics that can be used to make a more informed decision on MOSS deployment.
Performance tuning and benchmarking can be a complex undertaking, and many a times the interdependence of various network parameters during simulation and actual load on the MOSS box(es) can be confusing and perplexing, if not holistically viewed and planned it can skew the final metrics - the entire endeavor would be a waste of time if not panned and executed correctly.
Benchmarking the MOSS deployment and other design choices and rollout strategies can significantly contribute towards a successful enterprise SharePoint(MOSS) deployment. However at the end of the day its still the onus of the enterprises IT group to understand and execute the required tests prior to rollout to ensure the success of SharePoint solution. Its unfortunate that in many enterprises limited interpersonal dynamics within the IT group and lack of communication to the portal stakeholders leads to a less than desirable MOSS solution, IT group seldom looks into what the end users needs. No wonder IT is still seen as a "cost center" rather than a value center. " Technology produces its best results when an organization has the doctrine, structure, and incentives to exploit it" - a famous quote on a different context (the 9/11 commission report) however its applies to how every enterprise can strategically leverage IT.

1 comments:
Having problems with MOSS response times so this should be very helpful. Citrix have somethign similar as well called NetScaler & WanScaler.
Post a Comment