Low-Latency Caching for Cloud-Based Web Applications
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Many Web applications are now hosted in elastic cloud en-vironments where the unit of resource allocation is a virtual machine (VM) instance; entire VMs are added or removed to scale up or scale down. A variety of techniques can reduce the latency of communication between VMs co-located on the same server in, say, a private cloud. For example, par-avirtualized network mechanisms (e.g., vhost and virtio in Linux KVM) can optimize the number of protection bound-ary crossings. Inter-VM shared memory can further reduce boundary crossings after setting up a shared region. We present the design, implementation, and an evalua-tion of Nahanni memcached, a port of the well-known mem-cached that uses inter-VM shared memory instead of a vir-tual network for cache reads. As a widely deployed cache for back-end datastores and databases, memcached’s latency is important to the performance of many well-known web sites (e.g., Facebook, Twitter) and cloud platforms (e.g., Google’s App Engine). Although using shared-memory IPC is a well-known strategy, the recent introduction of the ivsh-mem inter-VM shared memory mechanism (also known as Nahanni) to Linux KVM makes the strategy practical for virtual machines. Using the Yahoo Cloud Serving Bench-mark, we confirm the intuition that Nahanni memcached can reduce the latency of cache read operations by up to 86%, and that given reasonable hit rates, this can reduce the total latency of read-related operations for a workload by up to 45 % compared to standard memcached. When using the experimental paravirtualized vhost networking mechanism in Linux KVM, Nahanni memcached offers a smaller, but still significant, advantage of 29%.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it