Mainframe-to-Snowflake CDC: The Tooling Lane Exists, the Snowflake Partner Motion Does Not
The question
"What is the current tooling landscape for mainframe-to-Snowflake CDC, and does Snowflake have a named partner motion here (Precisely, Qlik, Informatica, Fivetran HVR)?"
Derivative of [[2026-07-26-cobol-mainframe-ai-migration-phdata-pipeline]], whose recommendation was coexistence (offload mainframe transaction data into a governed cloud fabric rather than attempt a z/OS exit). This brief tests whether that recommendation is deliverable through tooling that already exists, and whether it is routable through a lane Snowflake itself publishes.
What we already know (from the vault)
- The parent brief's load-bearing inference: every mainframe program, exit or not, generates the same prerequisite - DB2/VSAM/IMS to CDC to Snowflake plus semantic modeling plus governance - and the offload phase succeeds even when the application exit fails. It explicitly flagged this tooling question as the "can we actually deliver it" gap. See [[2026-07-26-cobol-mainframe-ai-migration-phdata-pipeline]].
- The parent brief also named the tools to check by name in its follow-up list: Attunity/Qlik Replicate and Precisely Connect. That naming was the starting hypothesis, not a verified finding.
- phData's public Snowflake standing is Elite tier, with published solution areas (financial services, healthcare/life sciences, technology, manufacturing/industrial, retail/CPG) and two workload specializations (Data Engineering, Analytics). Snowflake's own directory publishes no mainframe or legacy-replication specialization, and no Precisely/Qlik/Informatica alliance. See [[2026-09-11-phdata-snowflake-partnership-warm-start-map]].
- IBM now sits on both sides of this trade inside the Snowflake ecosystem. IBM owns the mainframe, and IBM (via Hakkoda) took the 2026 AMER Snowflake Services Innovation Partner of the Year. See [[2026-09-07-midmarket-snowflake-si-competitor-set]].
- Slalom's "Zero Legacy" already plants a named modernization flag among Snowflake SIs, which is the competitive context for any phData positioning here. See [[2026-06-28-snowflake-si-cortex-positioning-caf-gap]].
What the web says
On tooling (priority 1):
- Precisely is the most explicit mainframe-to-Snowflake claimant. Its current product naming is Connect plus the Data Integrity Suite - Data Integration service; it names VSAM, IMS, COBOL copybooks, and Db2/z as sources, names Snowflake directly among targets alongside AWS, Azure, Google Cloud, Databricks and SAP, and describes the mechanism as "near real-time replication ... through advanced change data capture (CDC) technology." It cites a Citizens Bank outcome of "reduced mainframe MIPS usage by 80%, cut operating costs by 6%." (Precisely, 2025-10-13) [DOC - this is evidence of what Precisely says, not that it works at scale; it is also a self-comparison page that names zero competitors.] Note the naming trap: Precisely's Connect line is the replication product, while Ironstream is its mainframe/IBM i telemetry-to-observability product. Calling Ironstream a CDC tool in front of a mainframe team is a credibility defect. [INFER]
- Qlik still ships "Qlik Replicate" under the "Qlik Talend Cloud" umbrella, and its mainframe page is weaker than expected. The live page names Qlik Replicate and links a Qlik Talend Cloud trial, names only DB2 mainframe ("non-invasive DB2 CDC"), leads with "reduce costly MIPS cycles", and - the surprise - does not name Snowflake as a migration target anywhere in the body; Snowflake appears only in the site's technology-partner navigation. IMS, VSAM, IDMS, Adabas, CICS and copybook formats are not mentioned on that page. (Qlik mainframe modernization, fetched 2026-10-07) [DOC]. Broader IMS/VSAM support and Snowflake as a target appear in Qlik-adjacent material and AWS co-marketing (AWS APN blog on Qlik Replicate for mainframe) but I did not primary-fetch Qlik's connector documentation this pass. [UNKNOWN] at doc level. Lineage to get right: Attunity Replicate became Qlik Replicate after the 2019 acquisition, and Qlik absorbed Talend in 2023, so "Attunity" and standalone "Qlik Data Integration" are both stale names. [INFER]
- Treehouse Software's tcVISION is the specialist that markets Snowflake hardest, moving DB2, VSAM, IMS, Adabas, IDMS, Datacom and flat files into Snowflake and AWS, frequently with Apache Kafka as the intermediary (meshIQ / Treehouse) [DOC at search-snippet level, not primary-fetched]. It is a small independent vendor, which cuts both ways: deepest source coverage, thinnest procurement comfort for a risk-averse enterprise.
- Informatica's mainframe capability is PowerExchange CDC (DB2 z/OS, IMS, VSAM, Adabas) feeding IDMC, but the corporate situation makes the product name a live risk: Salesforce announced its acquisition of Informatica in May 2025, and I did not verify close status or the post-close mainframe-connector roadmap in this pass. [INFER on product shape / [UNKNOWN] on current roadmap.] Do not assert Informatica's mainframe roadmap in front of an account without checking release notes.
- Fivetran HVR is probably the wrong tool for this question, and the reason matters. Fivetran's log-based CDC strength (HVR, acquired 2021) is enterprise relational sources, and its Db2 coverage is strongest for Db2 for i (IBM i / AS/400), which is not z/OS. Conflating IBM i with z/OS is the single most common and most expensive credibility error in these conversations. Whether Fivetran supports Db2 for z/OS as a source today is [UNKNOWN] - it needs a connector-catalog check, not an inference.
- Software AG publishes a page literally titled "Move VSAM and Db2 data to Snowflake" (softwareag.com) [DOC at search-snippet level]. But after IBM's 2024 purchase of Software AG's StreamSets and webMethods lines and the Silver Lake split of the Adabas/Natural business, who actually sells and supports that path in 2026 is [UNKNOWN]. Treat it as a stale-name hazard until verified.
- The streaming intermediary pattern is well-attested as an architecture, with Kafka between the mainframe log reader and the warehouse (Kai Waehner, 2025-06) [PRACT, but note the author is Confluent's field CTO, so it is interested practitioner material, not neutral analysis]. Ververica markets purpose-built CDC connectors reading transaction logs from DB2, IMS, VSAM and IDMS (Ververica) [DOC, vendor].
- The Snowflake-side last mile is solved and is Snowflake's own. Snowflake Openflow (Apache NiFi-based managed ingestion) is Snowflake's named ingestion product, with connectors across SaaS, OLTP databases, event streams and file stores (Snowflake Openflow) [DOC]. No DB2 for z/OS, IMS or VSAM connector surfaced in either of my probes - [UNKNOWN], leaning absent. The practical implication: the capture side is the ISV's job, the landing side is Snowflake's.
On a named Snowflake partner motion (priority 2 - the half the question turns on):
- Snowflake's named migration motions are "Snowflake LiftOff" and the "Migration Accelerated Partner" program, plus SnowConvert for code conversion - and all of them are aimed at legacy data warehouses, Hadoop and ETL, not mainframes. (Snowflake migrate-to-the-cloud, LiftOff) [DOC]
- I fetched Snowflake's partner directory page for the Migration Accelerated Partner program (1,529 partner results) and mainframe is not mentioned anywhere on it - no mainframe, DB2 for z/OS, IMS, VSAM or COBOL as a named migration source, and no CDC or replication ISV named in the visible listing. (Snowflake partner directory, fetched 2026-10-07) [DOC - absence of evidence, from a single page]
- A domain-scoped search of snowflake.com for "mainframe modernization partner solution" returned no mainframe-specific program, category, or solution page. The two places mainframe surfaces at all: (a) the Federal Government industry page, where Snowflake itself writes coexistence language about connecting "legacy databases, mainframes and on-premises systems while maintaining current operations"; and (b) partner-authored profile text, notably IBM Consulting's profile describing end-to-end mainframe application modernization services. [DOC - absence of evidence, two independent probes]
- That (b) is the distinction the question was built to catch. IBM's mainframe-modernization claim is hosted on snowflake.com but authored by IBM. Snowflake renting a partner a paragraph is not Snowflake naming a motion.
On the practical shape (priority 3):
- No list pricing exists anywhere in this category. All three pages I fetched - Snowflake's, Qlik's, Precisely's - contain zero pricing or licensing statements. [DOC by absence] Everything here is enterprise-quoted, typically capacity- or source-count-based. Assume a six-figure annual license conversation for any of the three majors. [INFER]
- Copybook and EBCDIC conversion is the real engineering difficulty, and the vendor pages confirm it by how they handle it. Qlik's mainframe page does not mention copybooks or EBCDIC at all; Precisely's says it handles "VSAM, IMS, COBOL copybooks, Db2/z" "without requiring specialized skills" - a reassurance claim is a reliable tell that this is where projects die. The specifics that bite: the COBOL copybook is often the only schema definition that exists, packed-decimal COMP-3 fields, EBCDIC to UTF-8 code-page handling, REDEFINES and OCCURS DEPENDING ON variant records with no clean relational mapping, and no concept of null. [INFER - engineering judgment, not primary-sourced in this pass]
- MIPS is the pitch and also the catch. Both Qlik and Precisely lead with MIPS reduction, and Precisely cites an 80% MIPS reduction at Citizens Bank. But log-based CDC itself consumes mainframe CPU to read the DB2 log, so the headline saving is net of the offload's own cost, and the number that matters is whichever MLC/Tailored-Fit-Pricing arrangement the client is on. [INFER]
- The slowest gate is organizational, not technical. Putting a log reader or agent on z/OS means a software install into a production LPAR, a RACF conversation, a change-control window, and the mainframe team's own capacity planning. The mainframe team is a separate org from the Snowflake buyer in almost every one of these accounts. [INFER]
Convergences and contradictions
- Convergence - the tooling half of the parent brief's coexistence recommendation is real and does not have to be invented. At least three vendors ship named, documented mainframe CDC products covering the full source set the parent brief listed (Precisely Connect, Qlik Replicate, Treehouse tcVISION), and at least two of them name Snowflake as a target in their own material. Nobody in this lane is building a DB2 log reader from scratch.
- Contradiction - the vendors claim Snowflake far louder than Snowflake claims mainframe. Precisely names Snowflake; Treehouse markets Snowflake specifically; Software AG has a page titled for it. Snowflake's own migration programs name Teradata-and-Hadoop-shaped sources and never mention mainframe, and its partner directory names no replication ISVs. This is exactly the asymmetry the question was testing for, and it resolves against a named motion.
- Contradiction inside Snowflake's own surface - and it lands on the parent brief's weakest vertical. The single place Snowflake publishes mainframe coexistence language in its own voice is the Federal Government page. The parent brief judged government to have high COBOL concentration and the worst procurement fit for a mid-market SI. So the one place Snowflake has said the coexistence thing out loud is the one place the parent brief told us not to go.
Synthesis for RDCO
Verdict: PARTIALLY. A delivery lane exists; a Snowflake-named partner lane does not. Split the two, because the parent brief's recommendation depends on them differently.
On deliverability, the parent brief's coexistence recommendation gets stronger. The hard part of mainframe-to-Snowflake CDC is a bought capability, not a built one: Precisely Connect and Qlik Replicate are mature, enterprise-procurable products that read DB2 for z/OS, IMS and VSAM and land to cloud targets, tcVISION covers the widest source list including Adabas and IDMS, and the Snowflake-side landing is now Snowflake's own product (Openflow, or Snowpipe Streaming behind Kafka). What remains for a data-and-AI SI is precisely the work phData already does: copybook-to-relational modeling, EBCDIC and packed-decimal handling, semantic layer, governance, and the analytics and agent workloads on top. The offload phase is not a mainframe-skills engagement wearing a data-engineering costume; it is a data-engineering engagement with one bought connector at the edge.
On routability, the recommendation gets weaker, and this is the finding worth carrying. There is no Snowflake mainframe program to attach a deal to. Snowflake's named motions - LiftOff, Migration Accelerated Partners, SnowConvert - are shaped for warehouse and Hadoop migrations, and their account-mapping, funding and co-sell mechanics are built around that shape. A mainframe-source opportunity has to be re-framed into one of those existing containers to get Snowflake-side attention, or routed through the ISV's own partner organization (Precisely and Qlik both run their own SI channels), which is a different and slower motion. The honest version of the parent brief's claim is therefore: coexistence is a deliverable engagement that currently has to be sold as a generic migration, not as a named mainframe play. That is a meaningful downgrade from "routable deal path" to "deliverable but unlabeled."
The competitive read is the part that should change behavior. IBM owns the mainframe, owns IBM InfoSphere Data Replication, describes mainframe application modernization on its own Snowflake partner profile, and - per [[2026-09-07-midmarket-snowflake-si-competitor-set]] - owns Hakkoda, which just won an AMER Snowflake Services award. If Snowflake ever does name a mainframe motion, IBM is the overwhelmingly likely anchor partner, and the lane arrives pre-occupied. Combine that with Slalom's "Zero Legacy" flag from [[2026-06-28-snowflake-si-cortex-positioning-caf-gap]] and the strategic conclusion is: this is a capability to be quietly competent at when an account surfaces it, not a category to claim. That also fits the founder's actual role constraint - a DSA pointing an existing account at a known-good tool chain is exactly the right size of move; building a named mainframe offering is not his job and would collide with the no-clients, no-selling-services rule.
What a DSA can actually use from this, as a one-screen routing note: if a Snowflake account names a mainframe source, the tool conversation is Precisely Connect or Qlik Replicate for DB2 z/OS, IMS and VSAM (tcVISION if Adabas, IDMS or Datacom appear), the landing is Openflow or Snowpipe Streaming behind Kafka, the first technical question is "do you have current copybooks," the first commercial question is "who owns the mainframe MIPS budget," and the first program question is "which Snowflake migration motion can this be framed into," because there is no mainframe one. phData's own Snowflake partner standing on mainframe or on any of these ISVs is [UNKNOWN] - nothing in Snowflake's public directory or phData's public material establishes a mainframe specialization or a named Precisely/Qlik alliance, and it should not be asserted. Confirming it is an internal action, not a research question.
Why this is in the vault
It resolves the fifth open follow-up of [[2026-07-26-cobol-mainframe-ai-migration-phdata-pipeline]] with a split answer that changes that brief's confidence in opposite directions: it upgrades the coexistence recommendation's delivery feasibility (the CDC capability is buyable, from named and dated products) and downgrades its go-to-market claim (there is no Snowflake-named mainframe lane, so the parent brief's hoped-for "routable deal path" does not exist as such). It also supplies the three stale-name hazards that would cost credibility in an account conversation - Attunity, Ironstream-as-CDC, and Db2-for-i mistaken for Db2-for-z/OS.
Open follow-ups
- Does Qlik's current product documentation (not marketing) list Snowflake as a supported Qlik Replicate target and IMS/VSAM as supported mainframe sources, and under which product name after the Qlik Talend Cloud consolidation? The marketing page named neither.
- Did the Salesforce acquisition of Informatica close, and does PowerExchange CDC for mainframe survive in the post-close IDMC roadmap and release notes?
- Does Fivetran's published connector catalog support Db2 for z/OS as a source, or only Db2 for i? This is a binary doc check that decides whether Fivetran belongs in this conversation at all.
- Does Snowflake Openflow's published connector catalog contain any mainframe source, and has Snowflake published a reference architecture or quickstart for mainframe offload?
- After IBM's 2024 acquisition of Software AG's StreamSets and webMethods lines, who sells and supports the CONNX / Adabas-to-Snowflake path in 2026, and under what product name?
- Is there any independently published (non-vendor) account of a mainframe-to-Snowflake CDC deployment at production scale - conference talk, engineering blog, regulated-industry writeup? Every concrete claim found in this pass is vendor-asserted, which is the single biggest evidence gap here.
- Does Snowflake's Partner Network taxonomy contain a technology-partner category covering legacy or mainframe replication, and do Precisely, Qlik or Treehouse appear in Partner Connect? A per-vendor directory-profile check would settle whether a motion exists one level below the pages I probed.
Related
- [[2026-07-26-cobol-mainframe-ai-migration-phdata-pipeline]] - the parent brief this derivative tests; its coexistence recommendation is the thing being graded
- [[2026-09-11-phdata-snowflake-partnership-warm-start-map]] - phData's public Snowflake tier, solution areas and workload specializations; the basis for the [UNKNOWN] on mainframe partner standing
- [[2026-09-07-midmarket-snowflake-si-competitor-set]] - IBM/Hakkoda's 2026 AMER Snowflake award, which is why IBM is the likely anchor if a mainframe motion ever gets named
- [[2026-06-28-snowflake-si-cortex-positioning-caf-gap]] - Slalom "Zero Legacy" and phData's competitive band; the context for "be competent, do not claim the category"
- [[2026-07-15-stratechery-ibm-mainframe-ai-problems]] - the IBM moat thesis that started this thread, and the still-unreconciled mainframe revenue figure
- [[2026-04-13-joe-reis-ai-hard-parts]] - the vault's standing skeptical prior on legacy modernization, which the vendor-asserted-only evidence base here reinforces
Sources
Vault
06-reference/research/2026-07-26-cobol-mainframe-ai-migration-phdata-pipeline.md06-reference/research/2026-09-11-phdata-snowflake-partnership-warm-start-map.md06-reference/research/2026-09-07-midmarket-snowflake-si-competitor-set.md06-reference/research/2026-06-28-snowflake-si-cortex-positioning-caf-gap.md
Web - primary-fetched 2026-10-07
- Snowflake partner directory / Migration Accelerated Partners - https://www.snowflake.com/en/migrate-to-the-cloud/migration-accelerated-partners/ (mainframe absent; no CDC ISVs named)
- Qlik, Mainframe Modernization - https://www.qlik.com/us/mainframe-to-cloud/mainframe-modernization (Qlik Replicate under Qlik Talend Cloud; DB2 only; Snowflake not named in body; MIPS claim)
- Precisely, "Who Is the Best Vendor for Mainframe Data Modernization?" (2025-10-13) - https://www.precisely.com/blog/data-integration/who-is-the-best-vendor-for-mainframe-data-modernization/ (Connect + Data Integrity Suite; VSAM/IMS/copybook/Db2-z; Snowflake named; Citizens Bank MIPS claim)
Web - search-result level, not primary-fetched (treat as leads)
- Snowflake Openflow product page - https://www.snowflake.com/en/product/features/openflow/
- Snowflake LiftOff - https://www.snowflake.com/en/migrate-to-the-cloud/snowflake-liftoff/
- Snowflake Federal Government (the one Snowflake-voiced mainframe coexistence line) - https://www.snowflake.com/en/solutions/industries/public-sector/federal-government/
- meshIQ / Treehouse Software tcVISION to Snowflake via Kafka - https://www.meshiq.com/blog/treehouse-software-and-meshiq/
- Software AG, "Move VSAM and Db2 data to Snowflake" - http://www.softwareag.com/en_corporate/resources/mainframe-modernization/video/move-vsam-and-db2-data-to-snowflake.html
- AWS APN blog, Qlik Replicate for mainframe data - https://aws.amazon.com/blogs/apn/how-to-unleash-mainframe-data-with-aws-and-qlik-replicate/
- Ververica, mainframe offloading for banking - https://www.ververica.com/banking/mainframe-offloading
- Kai Waehner (Confluent field CTO), mainframe modernization with data streaming - https://www.kai-waehner.de/blog/2025/06/13/mainframe-integration-with-data-streaming-architecture-business-value-real-world-success/
Not consulted
- No Gartner or Forrester reprint was opened; the market-guide material in this category sits behind vendor lead-gen forms. Flagged, not retried.