mhtirogla
Home
Systems
API
Pricing
About
Log in
← Back
Home
›
Systems
›
Cloud platform and compute
gpu-cloud
GPU cloud with pods and scheduling
RunPod, CoreWeave
System
Techniques · 6
Primitives · 25
gpu-cloud
Gang scheduling
distributed · medium: All-or-nothing co-scheduling of a job's tasks
Checkpoint and restart
hardware-os · medium: Periodically save enough state to resume a computation after a crash, preemption or move, with an interval that balances save cost against lost work.
›
Gpu virtualization
hardware-os · medium: Share GPUs across VMs or containers: SR-IOV, mediated passthrough, time slicing, MIG partitions
Lazy image pulling
systems · medium: Start containers before full download by fetching image chunks on demand
Priority and preemption
systems · medium: Priority classes, quota and evicting lower-priority work to admit higher
Topology aware placement
systems · medium: Place work by NUMA, rack, zone and GPU interconnect topology
DMA and IOMMU
hardware-os · low: How devices move data by direct memory access, and how an IOMMU gives each device its own address space, faults on stray accesses and lets user space or guests own a device safely.
›
Hardware virtualization kvm
hardware-os · low: VT-x/AMD-V via KVM: vCPU run loop, VM exits, MMIO emulation
Resource usage accounting
hardware-os · low: Cgroup/procfs/GPU counters sampled into per-workload usage
Bin packing
optimization · low: Pack items of given sizes into the fewest fixed-capacity bins; sort descending, then first-fit, gives at most 11/9 of optimal and runs in O(n log n).
›
Leases
distributed · low: A lock that expires on its own: bounded-time ownership that survives crashes and partitions, given bounded clock drift and fencing tokens on writes.
›
Cgroups and namespaces
hardware-os · low: Linux namespaces limit what a process can see; cgroup v2 limits and accounts what it may use. Containers, sandboxes and job schedulers are built from the two.
›
Rdma
networking · low: RDMA verbs over InfiniBand/RoCE: queue pairs, memory registration, one-sided ops
SSH protocol (transport, auth, chan…
security · low: A three-layer protocol on one TCP connection: encrypted packet transport, user authentication, then multiplexed channels with windows for shells, files and port forwards.
›
Distributed barrier
distributed · low: Rendezvous/barrier for a dynamic worker group: join, agree on membership and rank assignment, restart on change
Copy-on-write fork
hardware-os · low: fork() gives the child the parent's whole address space by sharing pages read-only and copying one only when someone writes it; vfork and posix_spawn skip even that.
›
Dirty page tracking
hardware-os · low: Write-protection or PML logs of modified guest/process pages
mmap and the page cache
hardware-os · low: How read() and mmap reach file data through the kernel page cache, what readahead and madvise do, and why databases use their own buffer pool instead.
›
Ptrace process control
hardware-os · low: Ptrace: attach, read and write registers and memory, single-step, intercept signals
Gpu priority preemption
hardware-os · low: GPU context priorities and compute or graphics preemption granularity for latency-critical work
Pci enumeration
hardware-os · low: Walk PCIe config space: BARs, bridges, capabilities, resource assignment
Seekable compressed stream
compression · low: Independently compressed blocks plus an index for random access (BGZF, seekable zstd, eStargz)
+ 9 more primitives
System
Techniques · 6
Primitives · 25
gpu-cloud
Gang scheduling
distributed · medium: All-or-nothing co-scheduling of a job's tasks
Checkpoint and restart
hardware-os · medium: Periodically save enough state to resume a computation after a crash, preemption or move, with an interval that balances save cost against lost work.
›
Gpu virtualization
hardware-os · medium: Share GPUs across VMs or containers: SR-IOV, mediated passthrough, time slicing, MIG partitions
Lazy image pulling
systems · medium: Start containers before full download by fetching image chunks on demand
Priority and preemption
systems · medium: Priority classes, quota and evicting lower-priority work to admit higher
Topology aware placement
systems · medium: Place work by NUMA, rack, zone and GPU interconnect topology
DMA and IOMMU
hardware-os · low: How devices move data by direct memory access, and how an IOMMU gives each device its own address space, faults on stray accesses and lets user space or guests own a device safely.
›
Hardware virtualization kvm
hardware-os · low: VT-x/AMD-V via KVM: vCPU run loop, VM exits, MMIO emulation
Resource usage accounting
hardware-os · low: Cgroup/procfs/GPU counters sampled into per-workload usage
Bin packing
optimization · low: Pack items of given sizes into the fewest fixed-capacity bins; sort descending, then first-fit, gives at most 11/9 of optimal and runs in O(n log n).
›
Leases
distributed · low: A lock that expires on its own: bounded-time ownership that survives crashes and partitions, given bounded clock drift and fencing tokens on writes.
›
Cgroups and namespaces
hardware-os · low: Linux namespaces limit what a process can see; cgroup v2 limits and accounts what it may use. Containers, sandboxes and job schedulers are built from the two.
›
Rdma
networking · low: RDMA verbs over InfiniBand/RoCE: queue pairs, memory registration, one-sided ops
SSH protocol (transport, auth, chan…
security · low: A three-layer protocol on one TCP connection: encrypted packet transport, user authentication, then multiplexed channels with windows for shells, files and port forwards.
›
Distributed barrier
distributed · low: Rendezvous/barrier for a dynamic worker group: join, agree on membership and rank assignment, restart on change
Copy-on-write fork
hardware-os · low: fork() gives the child the parent's whole address space by sharing pages read-only and copying one only when someone writes it; vfork and posix_spawn skip even that.
›
Dirty page tracking
hardware-os · low: Write-protection or PML logs of modified guest/process pages
mmap and the page cache
hardware-os · low: How read() and mmap reach file data through the kernel page cache, what readahead and madvise do, and why databases use their own buffer pool instead.
›
Ptrace process control
hardware-os · low: Ptrace: attach, read and write registers and memory, single-step, intercept signals
Gpu priority preemption
hardware-os · low: GPU context priorities and compute or graphics preemption granularity for latency-critical work
Pci enumeration
hardware-os · low: Walk PCIe config space: BARs, bridges, capabilities, resource assignment
Seekable compressed stream
compression · low: Independently compressed blocks plus an index for random access (BGZF, seekable zstd, eStargz)
Fuse userspace filesystem
hardware-os · low: FUSE protocol: kernel forwards VFS requests to a userspace daemon; caching and splice
HTTP range requests
networking · low: Ask for byte offsets with Range and get a 206 slice back; the standard way to resume, seek and read big remote files piecewise.
›
Oci image layout
systems · low: OCI manifests, image index, layer tarballs and config, addressed by digest
d-ary heap
data-structures · low: Array heap with d children per node (d = 4 typical): half the binary heap's depth, siblings in one cache line, faster priority queues.
›
Graceful draining
systems · low: On a termination signal, stop taking new work, finish in-flight work until a deadline, tell clients to reconnect elsewhere, then exit; clients resume from a token.
›
Numa aware allocation
hardware-os · low: Allocate memory and pin threads on the local NUMA node; first-touch and interleave policies
Network topologies
networking · low: Fat-tree and Clos, dragonfly, torus and rail-optimized topologies: bisection bandwidth and diameter
Fencing tokens
distributed · low: A number that grows with every lock grant and is checked by the protected resource, so a paused ex-holder's late write is rejected.
›
Virtual memory and the TLB
hardware-os · low: How page tables and TLBs translate every address, what TLB reach means, and when 2 MiB or 1 GiB huge pages pay off on Linux.
›
Copy for AI
Similar systems
Share the most building blocks with this one.
cluster-scheduler
40 %
gpu-job-scheduler
23 %
ci-runner-fleet
17 %
serverless
16 %
cloud-dev-env
15 %