We present in this article TranspeerToken, a new approach for managing transactions for WEB2.0 applications, based on a previous work, Transpeer. Contrarily to the latter, TranspeerToken is able to manage peak loads efficiently, i.e., when most of transactions try to access the same data concurrently. This bottleneck problem happens because Transpeer exploits a distributed system (DHT) for storing metadata used for transaction routing, which raises data contention in case of peak loads. We leverage it by replacing the DHT by a token ring-based mechanism, where each token carries a metadata, avoiding thus locking metadata. Furthermore, the token mechanism tolerates the failure of k consecutive nodes of the ring and guarantees that transactions requests are not lost in case of failure. Some preliminary experiments on top of SimJava simulator have given promising evaluation results.