It depends on your goals and workload, it really really does.
If you're just going for high-availability, multiple servers is better than a single big server. The multiple servers can be mid class, or even low class depending on whether or not the cluster's services can be broken up, and how well they take to load-balancing.
If you're looking for computational resources, not all workloads parallelize well and those would be better served by a big machine with a few fast processors. Other workloads parallelize very well, and those would be served by multiple machines. Whether or not the multiple machines are netbook class performance machines or mid-range servers again depends on what you want to do with them and how fast you want to get there.
It depends on many things; primarily, what do you need the server(s) for, and what exactly do you mean by 'low end'? But just some junk to throw out there:
One High-End server Disadvantages:
single point of hardware failure (lack of redundancy)
Can you give us detail about what you are trying to do? Are you clustering Windows, Linux, vmWare, Xen? What is you application: web servers, Database servers, email servers, etc.
An advantage of low-end over high-end is distributed resources and easy replacement but is there enough power for the service you are providing? I can't tell with the info you have given.
A disadvantage with low-end would be a requirement for better operational conops and potentially more staff if you deployed a lot of smaller hosts.
If you are virtuallizing you may want big machines to host your ESX/Xen/Hyper-V and then cluster them, with additional clustering at the OS level.
Cluster
Single Bigger Box
It's this last item that matters most to me, and why I never buy a single box for production systems.
It depends on your goals and workload, it really really does.
If you're just going for high-availability, multiple servers is better than a single big server. The multiple servers can be mid class, or even low class depending on whether or not the cluster's services can be broken up, and how well they take to load-balancing.
If you're looking for computational resources, not all workloads parallelize well and those would be better served by a big machine with a few fast processors. Other workloads parallelize very well, and those would be served by multiple machines. Whether or not the multiple machines are netbook class performance machines or mid-range servers again depends on what you want to do with them and how fast you want to get there.
Cluster of low-end servers:
Single high-end server:
It depends on many things; primarily, what do you need the server(s) for, and what exactly do you mean by 'low end'? But just some junk to throw out there:
One High-End server Disadvantages:
Cluster Disadvantages:
Can you give us detail about what you are trying to do? Are you clustering Windows, Linux, vmWare, Xen? What is you application: web servers, Database servers, email servers, etc.
An advantage of low-end over high-end is distributed resources and easy replacement but is there enough power for the service you are providing? I can't tell with the info you have given.
A disadvantage with low-end would be a requirement for better operational conops and potentially more staff if you deployed a lot of smaller hosts.
If you are virtuallizing you may want big machines to host your ESX/Xen/Hyper-V and then cluster them, with additional clustering at the OS level.