Use the Feature Cache.
When editing geometric networks in ArcMap, connectivity between features is maintained by the geometric network on the fly and is based on geometric coincidence. When a new feature is added to a network feature class, this geometric coincidence must be discovered. This is done by executing a separate spatial query on the server for each feature class in the network. If coincidence is discovered, then network connectivity is established. The cost of maintaining the connectivity on the fly is execution of these queries, which can be expensive.
To mitigate this cost, use the feature cache while editing the network; it will cache the features in your current extent so these spatial queries will not execute against the server. Using the feature cache when editing your network features will significantly improve performance when adding new features or connecting and moving existing features.
Use subtypes to minimize the number of classes in your geometric network.
While there is no limit to the number of feature classes that can be in a geometric network, the more classes you have, the more queries that will need to be executed while maintaining connectivity. Try to reduce the number of feature classes you have in your geometric network by lumping feature classes together using subtypes. If your feature classes carry different attributes, you can use relationships to manage subtype-specific attributes in different tables in the database, or you can keep all the attributes in the same table using nulls for those that don't apply to a particular subtype.