powered by CADENAS

Manual

Manual

4.5.3. Performance Multi-Site Installation

Definition:

  • Latency

    Latency is the time a signal needs for a defined distance. In the following we refer to it as round-trip, that means from A to B and back. This is the time needed for each transaction to start: Client A has to communicate with Server B and gets an answer, then the transaction starts.

    For example, latency within Europe is about 40 ms, between Europe and South America about 250 ms. See following figure.

  • Bandwidth

    Bandwidth defines how much data can be transferred in parallel. E.g. a 10 Mbit line can transfer 1,25 Mbytes per second.

    Looking only at PARTsolutions, not all users are searching and selecting parts there at the same time.

    PARTsolutions alone in a 5 Mbit network: Few collisions = Full speed at most times

    PARTsolutions alone in a 5 Mbit network: Few collisions = Full speed at most times

    For customer own parts, there could be longer download times depending on the bandwidth.

    The available bandwidth is shared between all users and all applications at the same location (e.g. Mail, Browser, ERP, PDM, Remote Desktop, Citrix, …).

    PARTsolutions in a 20 Mbit network with other applications: Many collisions = Slow down

    PARTsolutions in a 20 Mbit network with other applications: Many collisions = Slow down

Conclusion: This means for many small transactions the latency is more important than the bandwidth. But for a few large transactions, the bandwidth is more important than the latency.

PARTsolutions speed comparison

The following table shows typical PARTsolutions actions in different latency / bandwidth environments.

[Note] Note

Due to many possible side effects the shown values can only be used as a rough estimation for a customer.

Data flow

In a standard setup the supplier parts, own parts and link database is only accessed through the PARTapplicationServer. Actions like browsing and searching only need a small bandwidth. Viewing a part needs to download the data locally which needs more bandwidth depending on the part complexity.

Cache Server

To compensate a small bandwidth or to save bandwidth it is possible to put a Cache Server between PARTapplicationServer and Client. The Cache Server will be placed at a company site to speed up all their clients.

CADENAS uses Squid as Cache Server, but if a company already has a HTTP Cache Server it can be used.