Summary
- The Alabama Supercomputer Network survives as both a historical name and a visible internet identity, but the current service is the Alabama Research and Education Network, or AREN, overseen by the Alabama Supercomputer Authority.
- Project ELEVATE is a multi-year effort to move services from a professional-services contractor into in-house management; staffing, procurement and network work show a real transition under way, although the public record does not yet prove that every promised operational benefit has been delivered.
- The strategic test is not whether Alabama can substitute one contract for another. It is whether ASA can turn additional staff, supplier relationships and technical assets into durable operating knowledge, measurable resilience and faster recovery across very different member access paths.
Reliability becomes an ownership question
A network can meet an availability target while still leaving its operator uncomfortably dependent on someone else’s knowledge. That distinction is the centre of Alabama’s present experiment. The state is not simply refreshing bandwidth or announcing another fibre build. Through Project ELEVATE, the Alabama Supercomputer Authority, or ASA, says it is moving services formerly managed through a professional-services arrangement into its own organisation.
The transition therefore concerns who understands the network, who holds the runbooks, who can make a change at two in the morning and who is accountable when a school or university cannot reach what it needs.
ASA’s own April 2025 account introduced Project ELEVATE as a multi-year move from a professional-services contractor to internal management and linked continuity to an expansion of technical staff. That is a statement of programme design, not proof of completion, but it usefully defines the ambition: this is meant to increase the state’s own operating capacity, not merely reshuffle supplier invoices (Project ELEVATE launch account). By September, the Authority described work to bring services in-house, strengthen network operations and expand expertise, alongside procurement for AREN internet connections. The claimed benefits in that update should still be read as objectives unless later service measurements substantiate them (Project ELEVATE September update).
The importance of the move follows from the network’s constituency. AREN is not a single campus network with one technology department and one set of users. ASA says it oversees the service for schools, colleges, universities, libraries and selected agencies. Those institutions differ in budget, technical staffing, geography and tolerance for disruption. A research workload can be delayed by a bad route; a classroom assessment can be interrupted by an access failure; a library may depend on the same connection for public access and administration. Reliability is therefore not an abstract percentage.
It is the state’s ability to keep a heterogeneous public-service system usable and to explain what happened when it is not.
The relevant question is consequently broader than “insourced or outsourced?” Both models can work, and an in-house operator will continue to rely on carriers, equipment makers, fibre owners and specialist services. The better question is whether ASA is moving the boundary of control to the place where it can improve diagnosis, procurement and accountability. Success would mean knowing enough about the service to direct suppliers rather than simply relay their answers. Failure would produce the appearance of control while dependencies and blind spots remain unchanged.
One network, several names and a continuous public mission
The assigned historical identity, Alabama Supercomputer Network, should not be mistaken for a separate present-day operator. The current successor network is the Alabama Research and Education Network, commonly AREN, and its operator is ASA. That continuity matters because old and new names coexist in institutional pages, routing descriptions and public memory. Without a careful name bridge, a researcher could either split one evolving system into fictional organisations or collapse the operator, network and suppliers into a single entity.
A state-by-state educational telecommunications report from the 1996–97 period provides the oldest useful bridge in the available record. Its Alabama section describes a state-operated Alabama Research and Education Network that had formerly been known as the Alabama Supercomputer Network. It also records a T-1-era fibre backbone running from Huntsville to Mobile and describes the network’s educational, research and internet-access role at that time. The report is valuable for continuity of name and mission. Its capacity, funding, site counts and reach are historical facts, not current specifications (Hezel Associates report archived by ERIC).
That name connection is not confined to a retrospective national survey. The University of South Alabama says its gateway to AREN, formerly the Alabama Supercomputer Network, and the internet is located at the university’s Computer Services Center. This is a narrower claim than a network map: it confirms downstream institutional usage and the old-name/new-name bridge, but it does not establish statewide topology or performance (University of South Alabama technical support page).
The operator is a different layer. ASA says the Alabama Legislature established it in 1989, that it operates the Alabama Supercomputer Center, provides high-performance computing and oversees AREN. Because that description comes from the Authority itself, it is sound evidence of institutional identity and declared scope, not an independent grade for effectiveness (ASA institutional profile). The distinction can be expressed simply: ASA is the public authority; AREN is the current network service; Alabama Supercomputer Network is the assigned historical identity that remains visible in older and some technical contexts.
Names persist for practical reasons. Autonomous-system records, university pages and engineering practice move on different cycles. None implies that an unnamed new organisation took control. The Alabama Supercomputer Network directory record is best read as an entry point into that continuity, not as a claim that the historical name supersedes AREN.
This history also changes how Project ELEVATE should be judged. Alabama is not creating a public research and education network from nothing. It is adjusting the operating model of an institution with decades of accumulated purpose and dependencies. The continuity is an advantage because the mission and constituency are established. It is also a burden because a live network cannot pause while an internal team learns it.
ASA’s operating surface is wider than transport
AREN is often described through connectivity, yet ASA’s operating surface spans more than circuits. The Authority links a statewide network to high-performance computing, specialist software, user support and programme administration. These functions do not prove that all of them share the same technical control plane, but they show why insourcing can have effects beyond the network operations centre. The people handling connectivity are part of an institution that also encounters researchers’ applications, schools’ funding constraints and members’ support requests.
On its managed internet and WAN page, ASA advertises dedicated symmetric access from 100 Mbps to 10 Gbps, more than 900 connected sites, seven AREN nodes in four locations and a backbone designed without a single point of failure. It also describes base service levels for different member classes. These are operator-published characteristics and can change. The “no single point of failure” formulation describes the backbone’s intended design; it has not been independently tested in the sources here and cannot be assumed to cover every local access circuit, building entrance or member configuration (ASA managed internet and WAN service).
The scope is important because resilience is end-to-end only when each relevant layer is resilient. A dual-path core cannot prevent an outage caused by a single access fibre into a school. Multiple upstream routes cannot repair a misconfigured campus firewall. A well-run backbone can still appear unavailable to a user if name resolution, identity services or local power fail. Conversely, an access provider may deliver its circuit correctly while an upstream policy problem damages reachability. ASA’s challenge is to understand where responsibility changes hands and to make those seams observable.
The network’s users also create unusual timing requirements. Education traffic is seasonal, but not uniformly so. Registration, testing, research deadlines and administrative cycles generate different peaks. Maintenance acceptable to one member may be disruptive to another. ASA’s announcements index includes maintenance notices alongside items about technical leadership, procurement and insourcing. The sequence provides public evidence that transition activity has multiple tracks, although an index is not an independent audit and individual notices may be revised (ASA announcements index).
This is why “bringing reliability back inside the state” should not be read as a promise that every physical component will be state-owned. The realistic goal is operational authority: the ability to see service conditions, make informed decisions, manage vendors and communicate with members from a position of technical knowledge. Fibre and transit will still cross organisational boundaries. The crucial change is whether Alabama controls enough information and expertise to manage those boundaries deliberately.
Project ELEVATE is a transfer of tacit knowledge
The visible components of ELEVATE—new hires, procurement notices, site work and supplier announcements—are easier to count than the capability the programme ultimately needs. The scarce asset is tacit knowledge. It includes how a network behaves under stress, which circuits have unusual escalation procedures, where configuration documentation is incomplete and how member institutions interpret a status message. Much of it cannot be handed over as a file.
The December 2025 Project ELEVATE update said work continued on initial-phase sites and that ASA’s staff had doubled to support the transition and ongoing services. Doubling is a substantial organisational signal, but the starting base is not supplied in the public reference, and headcount alone does not demonstrate readiness. A larger team can still lack role coverage, practical experience or clear authority. The update should be treated as a point-in-time first-party report about mobilisation, not a service-improvement result (Project ELEVATE December update).
The transition must also preserve service while authority changes. That creates a familiar operational paradox. The internal team needs hands-on work to become capable, but early hands-on work is when it is least experienced. A sensible progression would pair observation with supervised action, rehearsals with limited production changes and measurable acceptance criteria with each transfer. The public sources do not expose such a detailed transition plan, so it would be wrong to assert that ASA is following one. They do, however, show the conditions under which the programme should be evaluated.
Supplier relationships remain necessary during and after insourcing. A state team will not splice every fibre, manufacture routers or own every upstream network. The distinction is between dependency and dependence without leverage. A capable operator can specify the service, validate delivery, diagnose boundaries and change suppliers without losing its institutional memory. An incapable one may hold the contract while vendors hold the knowledge.
GDIT belongs within this narrow supplier history. A procurement preparation notice said a planned request for proposals would not be a renewal of services then in place with GDIT. That establishes GDIT only as the then-existing professional-services supplier in that procurement context. It does not provide a performance assessment, prove a final scope or say that a transition succeeded. Any account that turns the notice into praise or criticism of GDIT would outrun the evidence (ASA preparation notice for a second RFP).
The same restraint should apply to ASA’s language about improved response or stability. Insourcing can reduce the distance between a member’s report and an engineer’s decision, but only if staffing, tools and authority align. It can also create new delays if vacancies persist or processes remain unclear. ELEVATE’s strongest public case will eventually be made through comparative operational measures: time to acknowledge incidents, time to isolate faults, restoration time, repeated-failure rates, change success and member experience.
Until those appear, the evidence supports an active capacity-building programme, not a completed reliability verdict.
Procurement reveals the network’s dependency map
Procurement notices are sometimes treated as administrative background. For a network transition, they are closer to an architectural map. The categories ASA buys reveal where its own authority ends and an external service begins. They also show that insourcing network management does not mean eliminating the market; it means becoming a more informed buyer within it.
One ASA notice says the Authority issued a request for proposals covering IP transit, backbone transport, member-hub transport and WAN transport through an Alabama Buys sourcing project. Those four categories describe different layers of reach. IP transit connects the network to broader internet destinations. Backbone transport links major parts of the state network. Member-hub transport brings institutional aggregation points into that structure. WAN transport extends managed connectivity toward members. The notice confirms that ASA sought these services. It does not reveal responses, scores, prices, awards, contract performance or actual route diversity (ASA network-services RFP notice).
The preparation notice for the second RFP adds two institutional constraints: supplier registration and E-Rate considerations. These details show that purchasing for a public education network is not simply a technical optimisation. It must fit state sourcing procedures and federal support rules. Such constraints can lengthen a transition, but they also provide formal checkpoints. The programme’s operators need enough technical knowledge to translate service requirements into procurement language and enough administrative knowledge to preserve eligibility.
Procurement can become a repository of operational learning if incident history informs the next contract. Suppose a member repeatedly experiences long restoration times because responsibility is disputed between access and backbone providers. An internal team that records those incidents can add clearer demarcation, telemetry access, escalation and restoration expectations to the next solicitation. If the knowledge remains with a departing contractor or in individual inboxes, the state may buy the same ambiguity again.
There is another reason to avoid reading award outcomes into preparation documents. Supplier selection is only an intermediate step. Real assurance comes from implementation design, physical-path validation, testing, monitoring and sustained performance. A low-cost bid can become expensive if operational visibility is poor. A technically attractive bid can underperform if local dependencies are not understood. Public procurement gives Alabama a mechanism for competition and accountability, but it cannot substitute for engineering acceptance.
ELEVATE’s procurement phase is therefore evidence of the work, not proof of its result. It indicates that ASA is attempting to recompose the service boundary while building internal capability. The mature version of that model would keep strategy, service knowledge and accountability inside ASA while using external providers for well-defined infrastructure and specialist functions. Whether the current process reaches that state remains an empirical question.
Alabama Fiber Network is an infrastructure partner, not AREN’s operator
Alabama Fiber Network, or AFN, is prominent in the physical transition story, but its role must remain precise. AFN says it was selected to support Project ELEVATE by connecting more than 300 locations, nearly 100 hub sites and three of five high-speed backbone connections among major facilities. It also says its statewide middle-mile fibre was completed in October 2025 and identifies local partners. This is meaningful supplier-reported evidence of intended scale. It is not independent verification that every connection was accepted, physically diverse, complete or performing as intended (AFN partnership announcement).
Most importantly, AFN is not thereby the operator of AREN. ASA retains the operator role described in its institutional and service materials. AFN is an infrastructure partner, and local partners or carriers involved in delivery are dependencies rather than alternate owners of the directory identity. Keeping those roles separate prevents a common category error: treating the company that supplies transport as the institution that governs the public network.
The distinction also helps explain the insourcing model. ASA can bring network management in-house while purchasing middle-mile capacity or related transport from AFN. In fact, a disciplined internal team may make supplier partnerships more effective. It can define handoff points, require evidence, compare alarms across domains and recognise when a fault sits outside the supplied layer. Internal competence is not the opposite of commercial partnership; it is what allows the public customer to use that partnership on clear terms.
AFN’s stated counts should nevertheless be treated as claims awaiting corroboration. “Connected” can refer to several stages: planned, constructed, provisioned, tested, accepted or carrying production traffic. “Backbone connection” says little by itself about shared ducts or facilities. “Completed” may describe the supplier’s middle-mile build rather than the readiness of each AREN member path. Contracts, ASA acceptance reporting or observed service would be needed to resolve those questions.
That does not make the announcement unhelpful. It identifies a potentially large operating surface that ASA’s growing team will need to incorporate. Hundreds of locations mean hundreds of local contexts, demarcations, contacts and failure possibilities. Nearly 100 hubs imply aggregation points whose health can affect multiple downstream users. Connections among major facilities create high-consequence links that deserve careful diversity and restoration planning.
The public internet sees AS3464, not the whole service
AREN has an externally visible network identity through AS3464, but two different kinds of evidence describe it. The American Registry for Internet Numbers, or ARIN, associates AS3464 with Alabama Supercomputer Authority in its official autonomous-system registration. That record establishes resource administration at the registry level. It does not establish current topology, route diversity, security, availability or ownership of every circuit used to deliver service (ARIN RDAP record for AS3464).
Hurricane Electric BGP Toolkit provides a different view. Its page for AS3464 shows publicly observed prefixes and route-security indications and includes descriptions using Alabama Supercomputer Network. This is useful confirmation that the historical name remains visible in internet-routing context. Yet it is an observation from outside the network, potentially delayed and necessarily incomplete. It cannot see internal paths, local access failures, traffic quality, commercial terms or the service experienced at a member institution (Hurricane Electric BGP Toolkit observation).
The difference between registry and observation is more than methodological housekeeping. A registry answers who administers a number resource. A route observer reports what the public routing system appears to announce or learn. Neither is a substitute for the operator’s own telemetry. An autonomous system can be correctly registered while a local circuit is down. Public routes can remain visible while an application performs badly. Conversely, a change visible in an outside route collector may be benign and planned.
For an insourced operations team, AS3464 is part of the control surface. Staff need to understand route policy, upstream dependencies, address administration and route-security practices. They also need monitoring that goes beyond a single public observer. The packet does not document ASA’s internal tools or staffing coverage in these areas, so it would be speculative to rate them. The public evidence establishes identity and a limited external view only.
The routing evidence nevertheless supplies a useful governance lesson. Technical identifiers often preserve institutional history longer than communications materials. Alabama Supercomputer Network in a route description does not revive a separate operator; it is another strand connecting the historical identity to ASA and AREN. Properly interpreted, AS3464 helps make the network legible. Improperly interpreted, it can tempt an analyst to infer physical diversity, cybersecurity quality or end-user performance that BGP data cannot reveal.
As ELEVATE proceeds, a valuable public indicator would be clearer reporting that links external reachability, backbone events and member impact without exposing sensitive operational detail. Such reporting could distinguish a route event from a site outage and an upstream problem from a campus issue. That transparency would demonstrate the value of internal expertise more persuasively than organisational language alone.
Computing turns connectivity into a research service
The network’s meaning becomes clearer when placed beside ASA’s computing role. ASA-X, the Authority’s tenth supercomputer, is described by ASA as having been installed in 2023. Its published inventory lists 5,056 x86-64 processors, 33 TB of memory, 1.4 PB of shared storage, InfiniBand and A100 and H100 accelerators, along with 1,214 academic users at the time the page was prepared. These are first-party, point-in-time figures. They do not reveal utilisation, queue times, workload success or current user numbers (ASA-X hardware inventory).
This hardware makes network reliability tangible. A remote academic user cannot benefit from accelerators or shared storage if authentication, data transfer or the path to the facility is unreliable. Large datasets magnify the consequences of interrupted transfers. Interactive development and support sessions depend on latency and stability even when a batch computation later runs entirely within the centre. The statewide network and the supercomputer are institutionally distinct capabilities, but users experience them as parts of one research environment.
ASA’s software catalogue lists areas including programming, artificial intelligence, fluid dynamics, materials science, mathematics, molecular dynamics, bioinformatics and quantum chemistry. It says choices respond to faculty requests and describes options for user-installed software. A catalogue demonstrates breadth of intended capability, not continuous availability of every package or suitability for a given workload (ASA HPC software and tools catalogue).
The support model is equally revealing. ASA asks users seeking HPC help to provide commands, error messages, cluster details and job information, and it describes specialist training, consultation and research support. This indicates a troubleshooting practice that depends on structured context. It does not disclose staffing depth, response commitments or measured outcomes (ASA HPC support practice).
The presence of A100 and H100 accelerators can invite an easy but misleading story in which raw computing capacity defines the institution. The deeper public value lies in accessibility. Hardware, InfiniBand, storage and software matter when researchers across Alabama can reach them, obtain support and move data with confidence. AREN is part of that access system. Its reliability influences whether statewide high-performance computing is a shared resource or a facility that is technically available but practically distant.
This connection raises the standard for ELEVATE. Network operations should not be measured only by whether a link is “up.” The relevant service question is whether members can reliably complete the activities the infrastructure exists to support. ASA’s public materials do not yet provide enough end-to-end measures to answer that. They do show why internal network knowledge could improve the connection between infrastructure events and research impact.
Member experience depends on the last mile and the local room
Statewide architecture is only one half of a member’s experience. The other half lies at the institution: local fibre, power, routers, firewalls, wireless systems, identity services, staffing and the ability to describe a fault. The University of South Alabama’s statement that its AREN gateway is located at the Computer Services Center is a useful reminder that statewide connectivity becomes real at a local handoff.
This limits what backbone claims can mean. ASA may design the AREN backbone without a single point of failure, yet an individual site may still have one entrance, one router or one path to its hub. The available public evidence does not provide institution-by-institution designs, so no general assertion about member redundancy is justified. More than 900 connected sites is a scale claim, not proof of equal service at each one.
Internal operations can still improve the experience even where ASA does not control the local equipment. It can establish clear demarcations, document known dependencies and help members collect useful evidence. It can identify patterns across sites: recurring carrier faults, maintenance collisions or configuration issues following a standard change. A central team that remembers those patterns can reduce the time spent rediscovering them.
Communication is part of reliability. When a member cannot reach a service, it needs to know whether the problem is local, regional or statewide; what the operator is doing; and when the next update will arrive. A technically correct but vague status message shifts diagnostic work onto under-resourced institutions. An internal team with direct access to telemetry and authority may be able to communicate more precisely. Again, that is an opportunity, not a measured result in the current sources.
The variety of members makes standardisation tempting and exceptions unavoidable. Schools and libraries may have limited local networking staff. Universities may run complex routing and security environments of their own. Selected agencies may have strict operational constraints. A durable support model needs common intake and escalation while preserving enough flexibility for those differences.
This is where the project’s labour dimension matters. “Local support labour” is not simply the number of people on ASA’s payroll. It is the accumulation of Alabama-specific knowledge and relationships inside a public institution. Staff turnover can still dissipate that knowledge unless runbooks, mentoring and review practices capture it. The benefit of insourcing appears when knowledge survives individual departures and informs both daily operations and the next procurement.
E-Rate makes administration part of network engineering
ASA says it manages E-Rate work for AREN consortium members and files for school and library support covering internet access, data transmission and eligible internal broadband categories. That is a first-party description of the Authority’s role. It does not independently verify awards, filing success or member satisfaction (ASA E-Rate programme management).
The programme role matters because public connectivity is shaped by administrative deadlines and eligibility rules as well as engineering. A technically sound service can become unaffordable if its procurement and filing structure does not meet programme requirements. A late or mismatched process can affect institutions that have little capacity to absorb the difference. E-Rate therefore sits on the critical path even though it is not a routing protocol or a fibre strand.
Project ELEVATE’s procurement notices explicitly mention E-Rate considerations. That link means the internal team must coordinate technical scope, sourcing timing and consortium administration. The desired network design has to be expressible in eligible, competitively procured services. Supplier registration and public purchasing rules affect which firms can participate and how transitions can be scheduled.
Insourcing could strengthen this coordination if technical staff and programme staff share a common picture of member needs. Engineers can explain why a transport requirement matters; programme specialists can identify constraints before a design becomes difficult to procure. The opposite risk is organisational separation, in which technical plans arrive too late for filing windows or administrative assumptions produce an awkward network contract.
The public record supports ASA’s management role but not a conclusion about its effectiveness. Future evaluation would benefit from aggregate reporting that respects member confidentiality while showing filing coverage, service transitions and avoidable disruptions. Until then, E-Rate should be recognised as part of the operating system around AREN, not treated as a peripheral funding footnote.
Availability needs a boundary, a denominator and a date
The strongest numerical performance result in the available record is historical. Alabama’s Fourth Quarter FY2022 state-agency performance report records 99.88 per cent network availability for Alabama Supercomputer Authority in the reported period. That figure situates network performance within the Authority’s state mission, but it is aggregate, dated and dependent on the report’s measurement boundary. It does not establish current availability or equal results at every member site (Alabama FY2022 fourth-quarter performance report).
An availability percentage becomes meaningful only when its definition is known. Does it cover the backbone, managed customer edges or all member access? Are planned maintenance periods excluded? Is the denominator measured continuously? Does a partial degradation count as downtime? Is an institution unavailable when one upstream path fails but traffic continues over another? Without those details, 99.88 per cent is a useful official historical indicator, not a complete account of user experience.
The date is equally important. FY2022 predates the publicly described ELEVATE transition. It can serve as context or a possible baseline only if later measures use a comparable boundary. A higher future percentage calculated differently would not prove improvement, and a lower one during expanded coverage might not prove deterioration. Comparability is the discipline that turns reporting into accountability.
ASA’s published service design adds another measurement challenge. A backbone designed without a single point of failure is an architectural property, while availability is an observed outcome. Both matter, but neither guarantees the other. Redundant components can share hidden dependencies or fail during an unsuccessful switchover. A non-redundant access path can remain stable for a long period. Design review, failover testing and incident statistics answer different questions.
ELEVATE should eventually be evaluated on more than aggregate uptime. Measures could include incident acknowledgement, time to isolate the failing domain, restoration, frequency of recurring faults, success of planned changes and accuracy of member communications. Procurement acceptance could report whether promised paths and monitoring were delivered. Workforce measures could show role coverage and training without implying that headcount equals competence.
None of those measures is present in sufficient detail in the packet, so they are recommendations for scrutiny, not claims about ASA’s current practices. The available evidence permits a narrower conclusion: Alabama has a historical official performance figure, a set of current operator design claims and a transition narrative. It does not yet provide a public, like-for-like before-and-after demonstration of ELEVATE’s effect.
That gap is not unusual during a multi-year programme. It is still consequential. Insourcing asks the public to accept transition risk in exchange for longer-term capability. Transparent measures are how an operator shows that the exchange is working and catches problems before the organisational story hardens into conventional wisdom.
What could still break the transition
The first risk is knowledge loss at the handoff. Documentation may describe the intended network while omitting years of practical exceptions. If supplier access ends before internal staff can operate confidently, the state can own the responsibility without owning the necessary understanding. Mitigation requires overlap, demonstrations, controlled exercises and verified records, but the public sources do not state how much of that has occurred.
The second risk is role dilution. ASA operates a network, a supercomputing centre, support services and administrative programmes. Doubling staff for ELEVATE sounds substantial, yet a larger total workforce does not reveal whether network engineering, security, service management and field coordination each have adequate depth. Key-person dependence can move from a contractor to one internal expert.
The third risk is confusing purchased diversity with actual diversity. Procurement categories and supplier descriptions may imply multiple routes, but shared conduit, facilities, power or local entrances can preserve a common failure point. AFN’s supplier-authored figures do not resolve physical-path questions, and public BGP observations cannot see them. Acceptance testing and current records are essential.
The fourth risk is transition overload. Site migrations, internet-connection procurement, hiring and daily service all compete for attention. The September and December updates show multiple streams moving at once. A delayed component can force temporary arrangements that become permanent. A team focused on project milestones can also neglect routine maintenance or incident review.
The fifth risk is measurement drift. If ASA reports only broad availability, improvements in one layer could obscure deterioration in another. Members with a fragile local path may not recognise themselves in a healthy statewide number. If metrics change during ELEVATE, comparison becomes difficult. The historical 99.88 per cent figure demonstrates the need to publish boundary and method with any later result.
The sixth risk is assuming that a public operator and its suppliers have opposing interests. That framing can damage cooperation precisely when fast escalation is needed. GDIT should not be assigned blame or credit from a notice that merely says planned services would not be a renewal. AFN should not be treated as either the saviour or owner of AREN based on its partnership announcement. ASA’s job is to manage each relationship with evidence.
The seventh risk is neglecting members’ local capacity. A sophisticated central network cannot compensate for every under-resourced site. If support processes demand data a small institution cannot collect, escalation will stall. ASA’s HPC support materials show the value of precise commands and error information, but network intake may need graduated assistance for members with different skills.
Finally, insourcing can become an end in itself. Organisational control is valuable only when it improves continuity, adaptability, accountability or cost over time. A state team should still ask whether a function belongs inside, whether a specialist supplier is better placed to perform it and how to preserve leverage either way. The durable achievement would be an intelligent operating model, not the maximum possible payroll.
The evidence is strongest on direction, weaker on outcome
Taken together, the sources form a coherent but uneven picture. Historical and university evidence connect Alabama Supercomputer Network to AREN. ASA’s first-party materials identify the Authority as operator and describe a broad research, education and public-service mission. Registry and routing sources associate AS3464 with ASA and preserve the older name in a technical context. Procurement notices, newsletters and a supplier announcement show a transition in motion.
The evidence is strongest where the question is identity or declared activity. It is reasonable to say that ASA oversees AREN, that the historical Alabama Supercomputer Network name is linked to it, that Project ELEVATE was presented as a multi-year insourcing programme and that network-services procurement formed part of the work. It is also reasonable to say that AFN publicly described a large support role and that staff expansion was reported.
The evidence is weaker where the question is outcome. The sources do not independently verify every site count or route. They do not establish that intended physical diversity was delivered. They do not compare incident response before and after insourcing. They do not show that the doubled staff has complete operating readiness. They do not support a judgment about GDIT’s past performance. They do not turn a public BGP view into an end-to-end availability measure.
This asymmetry should shape both reporting and management. It would be unfair to dismiss ELEVATE because final proof is not yet visible during a multi-year transition. It would be equally wrong to declare success by converting goals and supplier claims into results. The defensible position is that Alabama has undertaken a strategically credible capability transfer whose public evidence of implementation is growing, while evidence of realised reliability benefits remains incomplete.
The distinction between “credible” and “proven” is useful. Credibility comes from alignment: hiring supports insourcing, procurement addresses transport layers, site work extends the programme and the operator already has a longstanding mission. Proof requires outcomes that survive scrutiny: accepted infrastructure, capable staffing, repeatable operations and comparable service measures.
That is also why the story matters outside Alabama. Public research and education networks often sit between public accountability and highly specialised markets. Outsourcing can provide expertise and scale, but it can also distance institutional memory from the accountable body. Insourcing can restore knowledge and agency, but it can also concentrate operational risk in a team still being built. Alabama’s experience will offer a useful case only if observers preserve these distinctions instead of treating the programme as an ideological referendum.
A practical standard for success
Project ELEVATE should be judged by whether ASA can exercise informed control across a network it will never own in every physical detail. Five practical tests follow from the evidence.
The first is continuity. Members should experience planned migrations with clear notice, tested fallback and minimal avoidable interruption. Incidents during the transition should produce documented learning. Continuity does not mean no failures; it means that transition work does not create unmanaged risk and that restoration improves with experience.
The second is retained capability. ASA should be able to explain its architecture, dependencies, route policy, escalation paths and member demarcations without relying on a departing supplier’s memory. Staff expansion matters when knowledge is distributed, rehearsed and captured. The test is whether the organisation can act safely across shifts and personnel changes.
The third is verifiable supplier integration. AFN and other providers should deliver against technically meaningful acceptance criteria. Counts of locations and hubs should reconcile with operational inventory. Claimed diversity should be tested at the physical and logical levels appropriate to each service. Supplier monitoring and ASA monitoring should allow quick agreement about where a fault lies.
The fourth is member-centred performance. A statewide aggregate should be accompanied by measures that reveal the experience of different service classes without exposing sensitive details. Availability, restoration and communication should be defined consistently. The network exists for schools, colleges, universities, libraries and agencies, so their ability to use it is the ultimate outcome.
The fifth is adaptive procurement. Lessons from incidents and migrations should change future requirements. E-Rate and state purchasing obligations should be incorporated early rather than treated as obstacles discovered after a design is complete. An expert public buyer can preserve competition while requiring the observability and accountability its operations team needs.
These tests avoid a false choice between state capacity and external expertise. ASA can deepen its own knowledge while continuing to purchase transport, transit and specialist services. It can value a supplier’s contribution without ceding operator identity. It can report progress without presenting first-party plans as independent proof.
Alabama’s network has already outlived one name and several generations of technology, from the T-1-era backbone described in the 1990s to a service associated today with multi-gigabit access and an HPC environment containing A100 and H100 accelerators. That long arc is evidence of institutional continuity, not continuous modernity. Each era has required the operator to renew skills and infrastructure.
ELEVATE is the present renewal. Its most important product will not be a particular circuit, provider or organisational chart. It will be the state’s ability to understand the network well enough to keep public institutions connected, to challenge suppliers with evidence and to recover intelligently when systems fail. The record shows Alabama moving in that direction. The next stage is to make the outcome as visible as the intent.