In many cases, it will be significantly slower
than the cluster scratch as it typically consists of just a single
disk. Some drawbacks are that the storage can only be accessed on that
particular node and that the capacity is often very limited (e.g.,
100 GB). The performance will depend a lot on the particular
implementation in the cluster. In many cases, it will be
significantly slower than the cluster scratch as it typically
consists of just a single disk.
removed double sentence.