spring webflux的dataBuffer怎么理解？是否像流一样？

Question

请看一段代码，这个是用来做代理的：

ServerHttpRequest frontEndReq = exchange.getRequest();
        ServerHttpResponse frontEndResp = exchange.getResponse();
        String path = frontEndReq.getPath().pathWithinApplication().value();
        HttpMethod httpMethod = frontEndReq.getMethod();
        RequestBodySpec reqBodySpec = webClient.method(httpMethod).
                uri(backendServiceUrlPrefix.concat(path)).
                headers(httpHeaders ->
                {
                    httpHeaders.addAll(frontEndReq.getHeaders());
                    httpHeaders.remove("HOST");
                });

        RequestHeadersSpec<?> reqHeadersSpec;
        if (requireHttpBody(httpMethod)) {
            reqHeadersSpec = reqBodySpec.body(BodyInserters.fromDataBuffers(frontEndReq.getBody()));
        } else {
            reqHeadersSpec = reqBodySpec;
        }

        //调用后端服务
        return reqHeadersSpec.exchange().timeout(Duration.ofMillis(backendServiceTimeoutInMillis)).
                onErrorResume(ex ->
                {
                    //调用后端服务异常(超时、网络问题)时，转发到后端异常-后控制器
                    //此处也可不用转发，用frontEndResp.writeWith(...)直接将异常响应写回给调用方
                    Map<String, Object> forwardAttrs = new HashMap<>();
                    forwardAttrs.put(Constant.BACKEND_EXCEPTION_ATTR_NAME, ex);
                    return WebfluxForwardingUtil.forward(Constant.BACKEND_EXCEPTION_PATH, exchange, forwardAttrs)
                            .then(Mono.empty());
                }).flatMap(backendResponse ->
                        {
                            //将后端服务的响应回写到前端resp
                            frontEndResp.setStatusCode(backendResponse.statusCode());
                            frontEndResp.getHeaders().putAll(backendResponse.headers().asHttpHeaders());
                            return frontEndResp.writeWith(backendResponse.bodyToFlux(DataBuffer.class));
                        }
                );

请留意最后一句：

return frontEndResp.writeWith(backendResponse.bodyToFlux(DataBuffer.class));

它的原理是，把目标服务器请求得到的响应转成DataBuffer，再把这个DataBuffer返回给客户端。我看了DataBuffer相关文章，都说这个类是把java nio的ByteBuffer与netty 的ByteBuf相结合操作的类。也就是说，这跟io里面的byte[] 是差不多的，都是把数据先写到内存里面。但万一这个是大文件上传或下载，不是消耗过多内存了吗？
我们下载文件，传统的httpServlectRequest是拿到它的Outputstream以流的方式写回，webflux对应应该怎么做？