RosettaHub™ for Research
Compute without the queue
The cluster is
the easy part now.
Research groups usually get cloud access one of two ways: a ticket queue that makes the cloud slower than the machine under the desk, or an open account that nobody is watching until the money is gone. Neither is necessary.
What a researcher starts,
without asking anyone
Described once, launched on demand, and taken down when the work is finished. The environment comes up the same way every time, which is also what makes the result reproducible.
HPC clusters
A scheduler, a head node and compute that grows with the queue and shrinks when it drains. Spot capacity for batch work that can be requeued.
More →Spark and Hadoop
The distribution and applications your jobs expect, your bootstrap script, and separate spot bids for the workers that make up most of the cost.
More →Notebooks and desktops
Jupyter, RStudio and a remote desktop in the browser, so visualisation happens next to the data rather than after a download.
More →A grant is
the whole budget
In most organisations an overspend is embarrassing. On grant-funded work there is often nowhere for it to go at all, which is why the budget has to be something more than a number on a dashboard.
Refused, not reported
A launch that would take the project over budget is refused by the cloud’s own API. The overspend does not happen, so nobody has to explain it.
Attributed to the project
Cost lands on the person and the grant that incurred it, so the funder’s question is answered with a report rather than an investigation.
Idle work reclaimed
A cluster left running over a weekend is the classic way research money disappears. Idle resources are found and switched off across twenty eight kinds of waste.
Every dollar not spent on an idle machine is a dollar still available for the work.
Across institutions
Research crosses
institutions
A researcher belongs to a home institution, a consortium and a grant-funded project at the same time, each with its own role and its own budget. That is not an edge case to be worked around, it is the normal shape of the work, and it is how membership is modelled here.
Environments, datasets, images and clusters are shared directly with a person, a group or a whole organisation. Access appears when you share and disappears when you stop, with nobody in the middle writing a policy.
See how sharing works →Your accounts, your rates,
your credits
Everything runs in your own cloud accounts and is billed to you by the provider. Whatever you have already negotiated continues to apply, research credit programmes keep working, and no data leaves your estate to reach us.
“It facilitates the provision of resources and services in the cloud for complex projects in the research and innovation department, eliminating the need to invest in costly hardware.”
Common questions
How can researchers get cloud compute without going through central IT?
They launch it themselves, inside limits set once. HPC clusters, Spark and Hadoop clusters, GPU machines and notebook environments are all described and started from the portal, the command line or the API. The guardrails are what make self-service safe to grant, so the ticket queue stops being the safety mechanism.
Can a grant budget be enforced rather than monitored?
Yes. A budget is checked when a resource is created, so a launch that would take the project over is refused by the cloud’s own API. For grant-funded work that matters more than in most settings, because there is rarely anywhere to move an overspend to.
How do researchers at different institutions work together?
A person can belong to several unrelated organisations at once, each with its own role and its own budget. A home institution, a consortium and a grant-funded project can all coexist for the same researcher, and environments, datasets and clusters are shared directly with a person, a group or a whole organisation.
Whose cloud account does the work run in?
Yours. Everything is created in your own cloud accounts and billed to you by the provider at your own negotiated rates. Nothing runs on our infrastructure, no data leaves yours, and existing cloud credits and grant programmes continue to apply.