SAGA sits at the confluence of two strong 2025–2026 lineages: **program-level agent
scheduling (Autellix) and workflow-aware KV-cache retention** (Continuum, TokenCake,
KVFlow-lineage). Its distinctive move is to fold all three axes — distributed scheduling,
tool-call TTL, and provable fairness — under a single "workflow-as-schedulable-unit" thesis,
then anchor the whole thing to a Bélády competitive-ratio bound. The peer cluster below shows
this is less a novel primitive than a unification-plus-formalization of primitives that
each already existed in isolation.
The 8 peers fall into four bands by how directly they collide with SAGA's claims.
Band A — direct predecessors on the same problem (agent-serving scheduling + KV retention).
Both elevate the program (not the request) to the scheduling unit; Autellix does it via
program-level Least-Attained-Service (PLAS/ATLAS) over an arbitrary DAG
[2502.13965], SAGA via workflow-atomic dispatch of an AEG.
Autellix is non-clairvoyant (zero structural knowledge); SAGA is clairvoyant-ish
(observable AEG). This is the sharpest delta in the cluster.
TTL to pin KV across tool calls in multi-turn ReAct agents
[2511.02230], exactly SAGA's §4.2 mechanism. Continuum's headline is that
the dominant cost is per-turn queueing delay (58.2% of latency), not reload — a framing
SAGA under-emphasizes.
KV scheduler for multi-agent tool-call idle windows [2510.18586], a
memory-management peer to SAGA's WA-LRU + pressure-scaled TTL.
Band B — mechanism-complementary (KV reuse / speculation SAGA calls "orthogonal").
re-alignment + anchor-pool offset approximation [2510.12872] — a different
reuse axis (cross-prefix) than SAGA's within-session retention.
speculatively pre-executing the next tool [2512.15834]. SAGA itself flags
speculation as complementary (§1.2, §11 future work item 6): speculation trades wasted
compute for latency; SAGA schedules known work.
Band C — orthogonal-but-adjacent system framings.
88% of E2E latency [2511.00739] — a direct challenge
to SAGA's GPU-centric accounting (see §3 可攻击面).
same "route to the right resource" family but at the model-selection layer, not the
scheduling layer. Weak overlap; useful mainly as a fairness/cost-knob contrast.
Band D — scope/vocabulary peer.
[2505.02279]. Orthogonal to serving, but relevant: SAGA assumes
framework hints (LangChain/AutoGen callbacks) surface the AEG — the survey shows that hint
channel is not yet standardized, which stresses SAGA's best-case observability assumption.
Genuinely new in SAGA:
online policy gets to the offline optimum. Continuum's TTL is "expected-utility optimal" with
no competitive ratio [2511.02230]; Autellix explicitly leaves
program-level competitive ratio unproven and only simulates the gap to SRPT
[2502.13965]. SAGA's Bélády framing is the cluster's only attempt at an
optimality anchor for eviction [2605.00528].
anti-starvation but no formal bound; Continuum, TokenCake, KVCOMM have none. SAGA is alone
in providing a provable completion-time deviation bound [2605.00528].
competitive-ratio) is real: no single peer checks all four boxes.
Incremental / overlapping (SAGA is not first):
and Continuum's cost-benefit TTL is arguably more principled (explicit reload + queueing
terms with a memoryfulness factor η [2511.02230]) than SAGA's percentile-of-
log-normal TTL with linear pressure scaling. SAGA's §4.2 does not cite Continuum's queueing-
delay decomposition at all.
incremental contribution over Autellix is (a) making structure observable via AEG rather
than inferred from runtime, and (b) adding cache-retention + fairness on top.
Autellix additionally contributes batched swap (18× fewer ops [2502.13965]).
Potentially contradictory:
+54%* [2605.00528]. Continuum instead argues the scheduling/queueing*
bubble is the dominant cost and KV-reload is secondary [2511.02230]. These
two agree that co-location/scheduling beats raw cache mechanics — but SAGA foregrounds the
cache competitive-ratio as its "scientific contribution" while its own numbers say co-location
carries the win. See §3.
Attack 1 — the "1.31× of Bélády" headline over-claims relative to what carries the win.
SAGA calls the competitive ratio "the main scientific contribution" [2605.00528],
yet its own ablation shows removing session affinity costs +96% TCT vs only +54% for removing
workflow-aware eviction [2605.00528]. The marquee bound describes the
least load-bearing lever. A skeptic would say the real engineering win is plain session
co-location — which Autellix already gets from locality-aware routing without any Bélády framing
Attack 2 — the GPU-centric latency model may be measuring the wrong bottleneck.
SAGA attributes 6.0× latency inflation and 38% wasted time to KV regeneration
[2605.00528]. The CPU-centric characterization (2511.00739) shows that for
tool-heavy agents, tool execution itself (mostly CPU) is up to 88% of E2E latency
[2511.00739]. If SAGA's SWE-bench agents spend most wall-time inside
run_test/retrieval CPU tools rather than KV re-prefill, then eliminating KV regeneration
optimizes a minority of the latency budget — and the 6.0× figure would be workload-specific to
cache-regeneration-dominated traces.
矛盾根源: SAGA's instrumentation counts GPU-side time (KV regen, memory util)
[2605.00528]; 2511.00739 counts end-to-end time including CPU tool stages
[2511.00739]. Both can be internally correct: SAGA's 38%
is "of GPU-attributable time," 2511.00739's 88% is "of E2E time," and they overlap only if the
tool-idle windows are short. SAGA's own Table 1 (P99 web/API = 45s) suggests long tool tails
exist, which favors 2511.00739's framing for browser agents.
Attack 3 — Theorem 2 proves less than the abstract implies. The Lyapunov bound formally
covers only tenants with W_i ≤ C/N; heavy tenants lie outside the hypothesis yet the paper
reports comparable 99.1% SLO for them [2605.00528]. The empirical 99.2% is
called merely "directionally consistent" with the bound. So the "provable bounded-deviation
guarantee" in the abstract is a guarantee for the easy tenant class, validated empirically for
the hard one. Continuum is more honest here — it claims no formal guarantee at all
Attack 4 — TTL design is weaker than a cited-adjacent peer. SAGA's TTL = percentile of a
fitted log-normal, scaled linearly by memory pressure (§4.2). Continuum's TTL explicitly
maximizes P(hit)×Benefit − Cost, decomposing benefit into reload and out-of-order (queueing)
terms weighted by η = −Corr(k, N−k) [2511.02230]. SAGA's TTL ignores the
queueing-delay benefit that Continuum shows is the dominant term. A reviewer would ask why
SAGA didn't adopt (or at least compare against) Continuum's cost model.
Attack 5 — the anti-thrashing gating is fragile by construction. SAGA's steal trigger
T_idle=100ms is shorter than mean migration latency (230ms), requiring three separate guards
to avoid oscillation [2605.00528]. This is a self-inflicted design
tension; TokenCake sidesteps it with a single shared pressure-snapshot ordering protocol
[2510.18586], arguably cleaner.
Paradigm position. SAGA is a consolidation paper, not a paradigm shift. The paradigm shift
— "treat the agent program, not the request, as the schedulable unit" — was already articulated
by Autellix (2502.13965) [2502.13965] and Parrot (cited, out of scope). SAGA's
role is to (1) make the program structure explicit as an AEG rather than inferring it at
runtime, and (2) bolt on the two things Autellix lacked: cross-tool-call cache retention (which
Continuum/TokenCake independently supplied) and a formal fairness bound.
Where it advances the field. The Bélády competitive-ratio framing is a genuine conceptual
contribution: it reframes "how good is our eviction policy?" from a relative comparison ("beats
LRU") to an absolute one ("within 1.31× of the offline oracle") [2605.00528].
This is the kind of anchor the surrounding cluster lacks — Autellix's SRPT-gap simulation
[2502.13965] is the only comparable move, and it's about scheduling not
eviction.
Adoption evidence (weak). SAGA is HPDC'26, built on vLLM v0.6.0 + Ray, and claims
zero-agent-code-modification deployment [2605.00528]. But code is
未公开 (unlike Continuum, which ships github.com/Hanchenli/vllm-continuum
[2511.02230], and RouteLLM, which is fully open
[2406.18665]). This materially weakens reproducibility relative to
its closest twin. The reliance on framework-exposed AEG hints also depends on an interop channel
that the protocol survey (2505.02279) shows is still fragmenting
Deployment niche. Explicitly latency-sensitive interactive serving, accepting ~30% lower peak
throughput [2605.00528]. This is the opposite trade from Autellix (which
optimizes throughput at fixed latency, 4–15× [2502.13965]) —
so the two are complementary, not substitutes: pick SAGA for interactive multi-tenant TCT, pick
Autellix for batch throughput.
out-of-order/queueing benefit that Continuum quantifies via η [2511.02230].
A hybrid TTL = SAGA's log-normal percentile plus Continuum's memoryfulness-weighted queueing
term would likely tighten SAGA's own competitive ratio below 1.31×, since Continuum shows
queueing (not reload) is the dominant cost the current TTL is blind to.
future work item 6. Speculative Tool Calls (2512.15834) overlaps tool execution with
generation, capped at <2× by Lemma 1 [2512.15834]; SAGA schedules the known
work. Prefetching the argmax-successor KV (SAGA §4.3) and speculatively executing that
successor's likely tool would compound — but only for stateless tools, per 2512.15834's
constraint.
[2605.00528]. Given 2511.00739's finding that CPU tool time can be 88% of
E2E [2511.00739], an AEG node should carry a CPU-time estimate too, so
AFS urgency reflects true wall-clock remaining work. Nobody in the cluster co-schedules
GPU-KV retention and CPU-tool admission (COMB/MAS from 2511.00739) jointly.
(2510.12872) reuses KV across agent contexts via offset approximation
[2510.12872]. A multi-agent SAGA (SAGA §11 item 5) where sibling agents
share large prompt prefixes could layer KVCOMM's anchor-pool reuse under WA-LRU eviction —
turning "which session's cache to keep" into "which offset-approximated cache to keep."
static point [2605.00528]; Autellix's PLAS/ATLAS and RouteLLM's α-threshold
[2406.18665] both expose a tunable operating point. A workload-adaptive
controller that slides between throughput-optimal (BFS, Autellix-like) and TCT-optimal
(SAGA-hybrid) per tenant SLO is unexplored and technically straightforward given the shared
vLLM substrate all three build on.