Hermes Wiki update

grc-iit · Oct 29, 2023 · b9a7c5b · b9a7c5b
1 parent 6020b4c
commit b9a7c5b
Show file tree

Hide file tree

Showing 21 changed files with 17 additions and 17 deletions.
diff --git a/docs/03-Hermes/Home.md → docs/03-Hermes/01-Hermes.md b/docs/03-Hermes/Home.md → docs/03-Hermes/01-Hermes.md
@@ -3,7 +3,7 @@
 [[/images/Hermes_hierachy.jpg|Deep Distributed Storage Hierarchy (DDSH)]]
 
 Consider an HPC cluster equipped with a deep, distributed [storage
-hierarchy](Storage-Hierarchy) (DDSH), the bottom layer of
+hierarchy](06-Hermes-components/10-Storage-Hierarchy.md) (DDSH), the bottom layer of
 which is typically a parallel file system (PFS). DDSH was introduced to
 boost or to at least improve the I/O (POSIX, MPI-IO, HDF5, ...)
 performance of applications performing poorly otherwise. Unfortunately,
@@ -44,24 +44,24 @@ with the following characteristics:
     where in DDSH a given data item is <b>best/well/optimally-</b>placed at
     that point in time.
   - To that end, the system consists of the following major components:
-      - [Strategies and algorithms](./Data-Placement-Strategies) that
+      - [Strategies and algorithms](06-Hermes-components/04-Data-Placement-Strategies.md) that
         implement policies and facilitate
         data placement decisions. Speculative data
         placement for read operations is also known as
-        [prefetching](./Prefetcher).
+        [prefetching](06-Hermes-components/09-Prefetcher.md).
       - These strategies work with (dynamic) sets of [buffering
-        target](Buffering-Target)s and are applicable more
+        target](06-Hermes-components/03-Buffering-Target.md)s and are applicable more
         broadly.
       - The physical buffering resources are managed in a distributed
-        [buffer pool](Buffer-Pool) (see also [Batching
+        [buffer pool](06-Hermes-components/02-Buffer-Pool.md) (see also [Batching
         System](Batching-System)).
-      - [Buffer Organizer](./Buffer-Organizer)
+      - [Buffer Organizer](06-Hermes-components/01-Buffer-Organizer.md)
       - [Profiler](./Profiler)
       - To separate concerns and for portability, system buffers are
         **not** directly exposed to applications. There is a set of
         intermediate [primitives](Primitives) targeted by
         [adapters](./Adapters) for different I/O libraries. A
-        generic [metadata manager](./Metadata-Manager) (MDM),
+        generic [metadata manager](06-Hermes-components/08-Metadata-Manager.md) (MDM),
         supports the bookkeeping needs of the various components.
   - The whole system is deployed in a server-less fashion.
 

diff --git a/docs/03-Hermes/1.-Getting-Started.md → .../02-Getting-Started/01-Getting-Started.md b/docs/03-Hermes/1.-Getting-Started.md → .../02-Getting-Started/01-Getting-Started.md
@@ -261,7 +261,7 @@ daemon, `LD_PRELOAD` a Hermes adapter and set some environment variables.
 
 We spawn a daemon on each node, then run our app with the appropriate
 environment variables, similar to the process described
-[above](./1.-Getting-Started#hermes-services-running-in-separate-process-as-a-daemon).
+[above](01-Getting-Started.md#hermes-services-running-in-separate-process-as-a-daemon).
 
 ```bash
 HERMES_CONF_PATH=/absolute/path/to/hermes.yaml

diff --git a/docs/03-Hermes/2.-Performance-Analysis.md → ...formance-Analysis/01-Performance-Analysis b/docs/03-Hermes/2.-Performance-Analysis.md → ...formance-Analysis/01-Performance-Analysis
@@ -30,7 +30,7 @@ First we perform a strong scaling study.
 2. Processes vary between 1 and 48
 3. Transfer size of 1MB
 
-| ![run workloads](images/performance/ssd-scale.svg) |
+| ![run workloads](../images/performance/ssd-scale.svg) |
 |:--:|
 |SSD strong scaling|
 
@@ -45,7 +45,7 @@ we want to measure the impact of garbage collection and OS caching.
 2. Processes fixed at 4
 3. Transfer size of 1MB
 
-| ![run workloads](images/performance/ssd-dset.svg) |
+| ![run workloads](../images/performance/ssd-dset.svg) |
 |:--:|
 |SSD dataset size scaling|
 
@@ -62,7 +62,7 @@ First we perform a strong scaling study.
 2. Processes vary between 1 and 48
 3. Transfer size of 1MB
 
-| ![run workloads](images/performance/nvme-scale.svg) |
+| ![run workloads](../images/performance/nvme-scale.svg) |
 |:--:|
 |NVMe strong scaling|
 
@@ -77,7 +77,7 @@ we want to measure the impact of garbage collection and OS caching.
 2. Processes fixed at 4
 3. Transfer size of 1MB
 
-| ![run workloads](images/performance/nvme-dset.svg) |
+| ![run workloads](../images/performance/nvme-dset.svg) |
 |:--:|
 |NVMe dataset size scaling|
 
@@ -99,7 +99,7 @@ processes to be between 1 and 48. Each case performs a total of 100GB of
 I/O with transfer sizes of 1MB using the Hermes native Put/Get API.
 This evaluation was conducted only over a single node.
 
-| ![multi-core scaling (NVMe)](images/performance/multicore-nvme-scale.svg) |
+| ![multi-core scaling (NVMe)](../images/performance/multicore-nvme-scale.svg) |
 |:--:|
 |NVMe dataset size scaling|
 
@@ -128,7 +128,7 @@ updated based on its current value.
 In addition, each workload contains a "Load" phase which perform insert-only
 workloads. Unlike an update, insert replaces the entire record.
 
-| ![load workloads](images/performance/ycsb-load.svg) |
+| ![load workloads](../images/performance/ycsb-load.svg) |
 |:--:|
 |Performance of KVS for the LOAD phase of YCSB|
 
@@ -139,7 +139,7 @@ KVS adapter, since insert operations replace data. In our KVS, a record
 directly to a single Put operation in the Hermes KVS. This demonstrates
 that Hermes can perform comparably to well-established in-memory KVS.
 
-| ![run workloads](images/performance/ycsb-run.svg) |
+| ![run workloads](../images/performance/ycsb-run.svg) |
 |:--:|
 |Performance of KVS for the RUN phase of YCSB|
 
@@ -158,7 +158,7 @@ In this evaluation, we run a multi-tiered experiment using Hermes native
 API. The workload sequentially PUTs 10GB per-node. We ran this experiment
 with 16 nodes and 16 processes per node. The overall dataset size is 160GB.
 
-| ![run workloads](images/performance/tiering.svg) |
+| ![run workloads](../images/performance/tiering.svg) |
 |:--:|
 |Performance of Hermes for varying Tiers|
 
@@ -183,7 +183,7 @@ synthetic workload. Each rank produces a total of 1GB of data, there are 16
 ranks per node, and a total of 4 nodes. The total dataset size produced is
 160GB. We use a hierarchical setup with RAM, NVMe, and SATA SSD.
 
-| ![run workloads](images/performance/dpe.svg) |
+| ![run workloads](../images/performance/dpe.svg) |
 |:--:|
 |Performance of Hermes for varying DPEs|
 

diff --git a/docs/03-Hermes/3.-Hermes-Configuration.md → ...-Configuration/3.-Hermes-Configuration.md b/docs/03-Hermes/3.-Hermes-Configuration.md → ...-Configuration/3.-Hermes-Configuration.md
diff --git a/docs/03-Hermes/4.-Basic-Programming.md → ...05-Basic-programming/01-Basic-Programming b/docs/03-Hermes/4.-Basic-Programming.md → ...05-Basic-programming/01-Basic-Programming
diff --git a/docs/03-Hermes/Buffer-Organizer.md → ...-Hermes-components/01-Buffer-Organizer.md b/docs/03-Hermes/Buffer-Organizer.md → ...-Hermes-components/01-Buffer-Organizer.md
diff --git a/docs/03-Hermes/Buffer-Pool.md → ...es/06-Hermes-components/02-Buffer-Pool.md b/docs/03-Hermes/Buffer-Pool.md → ...es/06-Hermes-components/02-Buffer-Pool.md
diff --git a/docs/03-Hermes/Buffering-Target.md → ...-Hermes-components/03-Buffering-Target.md b/docs/03-Hermes/Buffering-Target.md → ...-Hermes-components/03-Buffering-Target.md
diff --git a/docs/03-Hermes/Data-Placement-Strategies.md → ...omponents/04-Data-Placement-Strategies.md b/docs/03-Hermes/Data-Placement-Strategies.md → ...omponents/04-Data-Placement-Strategies.md
diff --git a/docs/03-Hermes/Data-Staging.md → ...s/06-Hermes-components/06-Data-Staging.md b/docs/03-Hermes/Data-Staging.md → ...s/06-Hermes-components/06-Data-Staging.md
diff --git a/docs/03-Hermes/Hermes-Runtime.md → ...06-Hermes-components/07-Hermes-Runtime.md b/docs/03-Hermes/Hermes-Runtime.md → ...06-Hermes-components/07-Hermes-Runtime.md
diff --git a/docs/03-Hermes/Metadata-Manager.md → ...-Hermes-components/08-Metadata-Manager.md b/docs/03-Hermes/Metadata-Manager.md → ...-Hermes-components/08-Metadata-Manager.md
diff --git a/docs/03-Hermes/Prefetcher.md → ...mes/06-Hermes-components/09-Prefetcher.md b/docs/03-Hermes/Prefetcher.md → ...mes/06-Hermes-components/09-Prefetcher.md
diff --git a/docs/03-Hermes/Storage-Hierarchy.md → ...Hermes-components/10-Storage-Hierarchy.md b/docs/03-Hermes/Storage-Hierarchy.md → ...Hermes-components/10-Storage-Hierarchy.md
diff --git a/docs/03-Hermes/Tags.md → ...03-Hermes/06-Hermes-components/11-Tags.md b/docs/03-Hermes/Tags.md → ...03-Hermes/06-Hermes-components/11-Tags.md
diff --git a/docs/03-Hermes/Topology-Manager.md → ...-Hermes-components/12-Topology-Manager.md b/docs/03-Hermes/Topology-Manager.md → ...-Hermes-components/12-Topology-Manager.md
diff --git a/docs/03-Hermes/5. Adapters.md → docs/03-Hermes/07-Adapters/01-Adapters.md b/docs/03-Hermes/5. Adapters.md → docs/03-Hermes/07-Adapters/01-Adapters.md
diff --git a/docs/03-Hermes/Docker-Containers.md → ...-Hermes/08-Docker/01-Docker-Containers.md b/docs/03-Hermes/Docker-Containers.md → ...-Hermes/08-Docker/01-Docker-Containers.md
diff --git a/docs/03-Hermes/Developer's-Guide.md → ...-Developers-Guide/01-Developer's-Guide.md b/docs/03-Hermes/Developer's-Guide.md → ...-Developers-Guide/01-Developer's-Guide.md
diff --git a/docs/03-Hermes/7. Doxygen.md → docs/03-Hermes/10-Doxygen/01-Doxygen.md b/docs/03-Hermes/7. Doxygen.md → docs/03-Hermes/10-Doxygen/01-Doxygen.md
diff --git a/docs/03-Hermes/6.-FAQ.md → docs/03-Hermes/11-FAQ/01-FAQ.md b/docs/03-Hermes/6.-FAQ.md → docs/03-Hermes/11-FAQ/01-FAQ.md